Risk-aware service level agreement modeling in smart grid

Abstract

As advanced Smart Grid environments grow from a simple grid towards a complex provider ecosystem, there is an uncertain challenge on those grid environments that need to manage a risk paradigm. We present an automated risk-aware service level agreements modeling to the grid provider for speed automated pricing and getting better performance to program as an agent-oriented platform. The key idea of our novel approach is proposing a risk level agreements contract and a pricing model to decrease the complexity of previous methods from an off-line service level agreement to an on-line risk level agreement for managing the risk lifecycle of contracts used to record the rights and obligations of the services and their consumers. Based on a risk level agreements contract the model optimizes resource management according to the business objective level of the provider with an online risk-aware rendezvous to define the penalty level of the cost model. The corresponding quality of service criteria is defined based on multi-class risk-aware service level agreements between Smart Grid providers and their power consumers which include the tail distributions of the per-class costs in addition to the more standard quality of service metrics such as throughput and mean delays. Our empirical experiments show the benefits of the proposed approach.

1 Introduction

One way for companies to distinguish themselves from other companies is through innovation such as technological development. Developments in the environment of companies come at a rapid pace [7]. The amount of decision-making information increases, as a consequence the time to make a decision increases as well. But for companies, this is business as usual. All expert companies found that it is hard to make the right decision concerning the matter of implementing a new technology or not. Smart Grid, being rapidly introduced to the power industry is a new way of delivering power resources for many companies that had emerged [3]. The basic idea is employing networked components within the traditional power grid to attain better control and greater reliability for managing the energy resources of the Internet of Things in Smart Grid [28]. This technology refers to the different services that are provided for a consumer over a network of power lines for low power and lossy networks towards the perspective vision of Internet of Multimedia Things [13, 38]. This network involves different hardware and software of the services in an integrated network [10] or in a distributed network [4], utilizing information and monitoring systems to improve the grid performance [18], offering a wide range of additional services for the consumers [7].

1.1 Service-level agreement

Providing new services may be joint, with risky decisions that need to be negotiated due to those power consumers which are based on Service Level Agreements (SLA) contract [8]. The provider agreement to the consumers, quantifies the minimum acceptance service. The Smart Grid provider may not always fulfill agreed SLAs because of outages in the power center and linkage in transmission line. Thus, the fulfillment of agreed SLAs would not lead to economic penalties [22]. The lack of a fault-tolerant system in high-reliability requirements can cause consumers to request that their power come from other providers [14]. In nano-technology, minimizing the Failure Probability (FP) of the allocated tasks to Smart Grid infrastructure can not be economically feasible [26]. Fault-tolerance or overprovisioning resources increase consumer costs and is economically inefficient due to the use of redundant overbooked resources. Smart Grid can offer many types of services and the key considerations in contracting with this type of service will differ on risks and benefits associated with each model. Further, SLA negotiations in grid computing are currently also under investigation [22].

1.2 Risk-aware service-level agreement

It is important that there is a distinction between risk and uncertainty. A risk is measured by impact probability and unlike the uncertainty, risk involves exposure to impact. The uncertainty must be modeled on risk analysis that is distinguished from parametric uncertainty. This uncertainty includes either aleatory or epistemic uncertainty or both [29]. This process identifies risk events, assesses the probability of each event, makes a cost-benefit analysis of alternative responses, deicides on response, re-assesses probability and impact, and monitors the risky behavior [35] based on Business Level Objectives (BLOs). Note, in grid operations, there are two main control objectives: BLO based on a business perspective and Security Level Objectives based on a security strategy. The BLO is related to consumer satisfaction or economic loss risk. Security Objectives are the targets the provider establishes for their security program or security incident risk. The existence of a clear statement mapping business objectives with security objectives should exist.

There are many situations which can increase the risk of operations in Smart Grid networks. Overloading resources or activities in old resources impacts Quality of Service (QoS) directly, and consequently, on providers’ revenue due to penalties payment of non-fulfillment of SLA. A new model is introduced for measuring uncertainty in typical Smart Grid network services and how it affects QoS access. This model outlines how risk propagates through Smart Grids, which is usually a combination of grid resources that are interconnected. This information is used when evaluating SLA negotiations and Risk-aware Service Level Agreements (RSLA) implementation and provides strategies for reducing risk with lower economic costs than with full redundancy. Risk-aware policies are provided to maximize economic efficiency for increasing the fulfillment rate of SLAs by the Smart Grid Infrastructure Provider (SGIP). The provider negotiates with the consumer by introducing a Petri-based risk graph tool for intelligent assessing an economic model. The consumer must decide about any service risk and pay proportionally based on the lower the risk, the higher the price to make a risk-aware or risk-averse decision. Next, the provider agent can decide on the service risk automatically based on the agreed RSLA in a new model.

1.3 Proposed approach and evaluation overview

The main problem of this work is presenting a business-driven Smart Grid framework to allow a grid provider to maximize its BLOs while providing the risk-aware agreed rate of the consumer QoS based on the Risk Level Agreements (RLA) (as a new SLA contract) for an automatic dynamic pricing and initiating the third party. Focus is placed on two stages of a Smart Grid service provisioning: negotiation on RLA and resource management on the risk level agreement. The negotiation of RSLAs occurs between the power consumers and the Smart Grid providers based on the risk level of BLOs. The provisioning of resources is managed to fulfill the terms of agreements or to transfer it to an agreed third party (see Fig. 1a). At a glance, the model presents an agent-driven risk-aware grid management to introduce a new type of SLA as RSLA (see Fig. 1b). The proposed architecture has three layers. Provider Layer (containing the basic Smart Grid infrastructure), Service Layer (providing all Smart Grid services with different level services based on RLA) and the Consumer Layer (consuming all of the services generated by the previous layers). The experimental results of a case study (i.e. a simple model of an energy storage system) show the benefit of the risk-aware intelligent automated agent-oriented model to differentiate SLAs in Smart Grid market for cost and risk minimization at individual node and graph levels. The model optimizes the resource management according to the business objective level of the provider to an online risk-aware rendezvous to define the penalty level of the cost model based on a new risk level agreements contract. Also, the presented accounting model allows provider agent adjust prices to risk as a function of resources amortized.

Fig. 1

Our Proposed RSLA Elements

1.4 Contribution

Our motivation in this paper is extracted from a perfect model of a grid provider to different service costs that is called a different service level agreement [9]. For this, we think a risk model can be applied to a standalone risk middleware in a power grid network. This model distributes the risk of each level of the SLA to reduce the risk impact based on the service cost for preventing the economic penalties and achieving better operational decisions. The main objective of our paper is increasing the fulfillment rate of SLAs in provider side to apply risk-aware policies for maximizing economic efficiency. The aimed is to overcome the SLA risk management problems on a risky power grid environment that will lead companies to detect, rank and validate risks when the business processes are placed in a complicated grid. Our contribution is a new move toward an automatic agent-oriented model for Smart Grid risk-aware SLA which increases the fulfillment rate of SLA on the infrastructure provider side while maximizing the economic efficiency. This modeling is a view of an intelligent SLA risk modeling and monitoring [18]. Also, we present a graph-based propagation risk model economically modeled to provide consumers risk levels (moderate, low and very low) proportionally paid (the lower the risk the higher the price). The contributions can be summarized from approach capability for automation risk management requirement by proposing a formal model to speed automatic risk calculation (i.e. Petri Net modeling), defining a novel contract to help online agent-oriented negotiation (i.e. RLA contract) and formulating a pricing model to value the risk-aware management (i.e. defining a pricing relation) through risk lifecycle designing, real-time concurrent negotiation and smart grid revenue modeling for a simple electrical Energy Storage System (ESS) in a simulation environment to minimize the node risk and the graph risk.

The remaining sections of this paper is organized as follows: first, we present previous related works (Section 2). Second, we propose the problem definition and its requirements (Section 3). Next, we expose problem of modeling and useful background around the assessment and management of the risks to identify the most important Smart Grid related risks with Petri Net modeling; too, a revenue model is offered to apply different risk levels at different prices and assessment of the SLA risk (Section 4). Also, we detail the empirical experiments and the evaluation of proposed RSLA approach to several risk treatment of facing the risks (Section 5). Finally, we conclude the work and expose future works (Section 6).

2 Related work

Cost reduction is always a big issue in power grid business; therefore, companies might find the grid network solution proves to be very interesting. Also, flexibility is very important to power companies in order to follow market changes, because they must be able to react rapidly to developments (such as a credit crisis) [9]. Since the credit crisis, power companies get a lot fewer requests and invoices from consumers or other links to its network. The company can adapt to these the fast changes, they can survive in critical times. This will have an effect on the environments of companies because the usage of the different providers is lower which results in large overhead costs [29, 35]. Those companies that switch to Smart Grid can save money and many other resources because operational tasks, such as operational costs for the maintenance of hardware or the employees retained for performing these tasks, are taken over by the infrastructure provider [13]. They only have to pay for the services that they use with added risk assessment service, which culminates in a decrease in capacity entailing fewer costs. By spreading the grid on-demand services and the dynamic nature of the grid makes for traditional, more static risk management and their related configuration ineffective and need to be online as well [27]. We think that the way that grid risks should be presented, is the same the way grid resource is delivered as a service.

We live in an uncertain information world, thus an important challenge to power grid is risk subject [5]. Satisfying the consumer by fulfilling the agreed SLA according to RLA would not lead to economic penalties, consumer task failure [13] or loss of consumer [13]. This problem leads to overprovisioning with high cost or risky decision made on resource allocation. The risk assessment is a very important problem because there is an uncertainty pertaining to achieving the objectives that lead to risky behavior or imaging the risky environment [8]. We have to increase the satisfaction rate of consumers by automatically applying the RSLAs at the Smart Grid side to choose risk-aware policies for minimizing both consumer and provider costs. The RLA is a formal business record, persisting the context and tradeoffs of critical business decisions, across changes in the organization that related to the consumers, until such time as any decision needs to be revisited. Each RLA includes discussion of many key topics: the SLA’s Risk Tolerance, Risk Scenarios, Inherent Risk, Negotiable Laws, Recommended Controls to mitigate risk, Risk Mitigation Decisions, and remaining Residual Risk that is either accepted, transferred, or avoided. We only use a simple risk tolerance, reduced risk scenarios, and inherent risk that the complete definitions are out of the article scope as a new paradigm.

Traditional grid technology is responsible for satisfying the consumer requirements without any knowledge about the dynamic power consuming [24]. The provider does not know how much and when the power is required by the consumers. Thus, it has programmed consumer needs based on the worst case of peak power usage, for example, during the noontime of summer days [21]. In fact, peak usage might be more than that which normally required but consumer behavior may be changed at the time. Nowadays, the amount of power is more than the average required, which results in wasted generated electrical power. This needs a dynamic programming or negotiation between producers and consumers. Smart Grid technology has arisen as a successful commercial solution so that power can be sold as a utility: consumers usually use power according to their tasks at hand and are charged and pay for what they consume [36]. The sold power as a virtual resource usually can be generated by different networks and scales at consuming time.

Now, in Smart Grid business, providers define different power prices, therefor their consumers may decide to buy them or not and there is no speed online (automatic) negotiation between them [29]. Our research, however, is framed in risk pricing of Smart Grid market. In Smart Grid markets, both the consumer and the provider are smart independent agents that negotiate the terms of QoS. The price that consumer will pay to the provider is according to Risk Level Agreements (RLA) awards and penalties [24]. The term of the contract established according to the risk-aware manner of RSLA upon the finalization of negotiations. This novel contract keeps contractual information about the terms of the QoS as well as pricing information on different risk level agreements. All paying price and penalty in SLA violation case is defined and formulated in a new complement contract as RLA as our contribution. This needs moving to an automatic online risk calculation tool to monitor the risk of power consuming variation in a specific time to consumer side [34].

As a complementary effort, we must provide a more intelligent and efficient coordination between the provider and the consumer with the power demands of loads. Thus, the power generation utility will not need to support the peak power demand and may save the additional power in ESS or balance with consumers on/off, down/up consuming or down/up electrical generating [12]. The status of traditional Smart Grid designs is far from this service because of the lack of a basic infrastructure. We try to place leverage on the popularity of a risk-based SLA to fulfil the shortage of the Smart Grid infrastructure. We propose an unequipped middleware for the risk as ”Risk Middleware” to interface with the risk-aware application and operating system to support risk-averse customer delivery services. A sample risk middleware consists of five layers including application risk, components risk service, risk service (for providing a distributed service), sharing risk service management (for transferring a step of a risk service to a client environment) and service provider risks (for internal management and monitoring, transferring risk to the client for monitoring and decision making). The risk service provider can be transformed to a risk provider center. The middleware platform manufacturer must be provided sharing risk service management for risk task management, interfaces providing, independence from the platform, high-efficiency by hiding the algorithm complexity, especially for our proposed risk analysis service. This can especially be applied well to the multi-Grid scenarios or the opportunity in risk transferring. The design of a risk middleware is an important research Smart Grid security puzzle [11], however it is not related to the subject of my paper and it is only being presented for the realization of offered RLA management.

3 Problem definition

The main objective of our paper is to come up with the risk classified or modeled, for risk management that will lead companies to detect, rank and validate it by learning to monitor its behaviors when a business process is placed into a Smart Grid. We model a risk as an object that enters a Smart Grid system for an objective. After the analysis process on the object, the system generates a risk agent (Ragent) as a bot entity that flows in the Smart Grid risk graph roads. A Risk object can be defined as an uncertain entity that is of interest for our future system analysis. Normally, the objects can be represented by their threats with the help of probability, primitive geometric behavior, quantitative models and qualitative models. We can apply some used algorithms as creating risk object, basic evaluating risk object, analyzing the risk object, applying the benefit assessment method, understanding a risk object, remembering and naming all Ragents as below:

• (1) Creating: We trig it with an event in an uncertainty graph model. We recognize a threat and remember it to create a risk object to birth the risk process.

• (2) Evaluating: After creating the threat object, we evaluate and reinforce the basic probability of the occurrence and the object risk level.

• (3) Analyzing: After the evaluation, we analyze the status and evaluate the risk environmental conditions for risk level recognition.

• (4) Applying: Then, we apply some infernal assessment model for certitude and measure some risk parameters such as risk level.

• (5) Understanding: After applying, we understand this status based on RLA for the all services under the influence of the threat/risk occurring.

• (6) Remembering: Finally, we remember the risk object and naming this like a tyke for flowing in the presented risk Middleware and growing this like a serpent.

For our agent-based model, we think that a multi-level structure can be applied to a standalone risk middleware for supporting the all needed infrastructure, actors, stakeholders, controls, and the others entities of the system control. For the independent evaluating section of a risk, middleware is regarded as a think tank (continuous thinking) for a crucial program and technical issues of securely management, especially for future governmental security. For this, the architecture needs to be constructed with an imaginary Smart Grid for continuous thinking with our proposed layer as a ”risk infrastructure” such as risk population generator, risk population resource, risk population monitor, risk population control, risk resource control, risk cost control, risk think tank, risk assessment process, risk impact control, and risk assurance task to future complicated power networks.

A business-driven Smart Grid approach is contributed which allows providers to deal with the complexity of maximizing BLOs while providing the agreed rate of QoS. The inclusion of risk management in the Smart Grid paradigm can be visible from both sides, namely the consumer and the provider. The risks are divided into two main categories based on the effectiveness on high-level objectives (i.e. business organization view) and low-level management decisions (i.e. Smart Grid organization view). These are the external impact of business process and internal threat or impact of grid providers. A revenue risk-aware model is presented to provide different risk levels while adapting these prices to their current value of the resources (that is rate resources decreases in their value over time) to model and compute risk propagation in Smart Grid services. This aids in the evaluation of SLAs negotiation and RLA enforcement of triggering strategy to mitigate risk without increasing prices proportionally.

A secured SLA solution for Smart Grid must manage the SLA to allow the terms of negotiation in according to consumer needs, assessment and measurement of the agreements. Additionally, automatically providing and assessing a secured negotiation based on parametric features of risk level agreement through secured SLAs is a challenging task in Smart Microgrids [7]. The secured automated SLA life cycle consists of following phases. A provider sets service and bounds to announce the services by using some templates. After providing these, the consumer searches the published offers to find a suitable provider and wishes to use services according to the agreed level for satisfying their needs. Next, they use the negotiation protocol to support risky iterative where in the client lays the desired QoS and the provider analyzes the issue to see if it can commit itself to the agreed request. After the SLA is signed, the operation process starts SLA monitoring, accounting and RLA enforcement. Renegotiation and termination of these negotiations are performed automatically in the Smart Grid [16]. We propose an RLA to accelerate an automatic decision making process of the SLA contract to speed up the agents negotiation toward real-time Smart Grid operations for seeking better service to best fit the needs of the consumer of presumed request services [6].

4 Problem modeling

This section describes the Risk Negotiation Model and its infrastructure (Section 4.1), proposes Risk Assessment (Section 4.2) with a Mathematical Graph Model (Section 4.3), gives Smart Grid operation (Section 4.4) and brings the Revenue Modeling (Section 4.5) forward for consideration.

4.1 Negotiation risk-aware model

In considering the infrastructure, the consumer agent can get three types of Smart Grid resources (get a power node, store power on a storage node, free a power node) and consider how to connect using network interfaces and storage links. Therefore, provider agent, aimed to make decisions on consumer agent service without human intervention, then calculates the all needed resources with the amount of each required resource based on an agreed temptation on each resource as a risk parameter. Finally, the provider agent can program and send a request to the consumer agent for online negotiation. Its provider agent has a set of different N sites to satisfy the received request which includes several different Physical Resources (PRs).

The RLA of a set of PRs for consumer agent i is formulated as $RLA_{i}(t)= \{\overrightarrow {R^{i}}, \overrightarrow {A_{i}}, {\Delta } t, \gamma , \tau , Reven(vt)\}$ where $\overrightarrow {R^{i}}=\{{R_{1}^{i}}, {R_{2}^{i}}, \ldots , {R_{k}^{i}}\}$ is RLOs that models a number of resources to be purchased by the consumer i; the $\overrightarrow {A_{i}}$ describes the resource allocation according to its standard; each R#i term presents the amount of voltage, frequency, storage, load, and so on; the γ denotes the risk level (high, medium, low, very low) of selected service and temptation degree τ for the interest consumer price; Δt describes the needed time period during PR allocation. If the Smart Grid has enough free resources to allocate the RLA_i or fulfill the tolerate τ then allocate else reject. The Reven(vt) proposes a revenue function to the Smart Grid provider. Current commercial Smart Grids are not enforced to specify Δt and sell resources at a fixed price/hour value. A new middleware to Smart Grid marketing needs to specify Δt to carry out negotiations of consumer i and provider j at a certain price P_i with an agreed temptation degree τ to distribute their workloads across time.

The Reven(vt) function proposes the revenue of Smart Grid provider after RLA_i operation that has a penalty as a negative revenue and is defined as $Reven(vt, \hat {\alpha }, \hat {\upbeta }, \hat {\alpha \upbeta })$ step function where α is the maximum penalty; β describes the maximum revenue; $\hat {\alpha }$ defines the upper bound of maximum penalty; $\hat {\upbeta }$ proposes the upper bound of maximum revenue; $\hat {\alpha \upbeta }$ presents a decision risk point for getting revenue or paying penalty; and the vt (violation time) is the amount of time in which Smart Grid has not been completely satisfied the agreed QoS (see Fig. 2). The provider can violate the RLA without any penalty, we need to allow a penalty as a grace period. if ($vt\leq \hat {\upbeta }$), then the provider will get a maximum revenue (β); if ($vt\geq \hat {\alpha }$) then the provider will get a negative revenue or pay all the maximum penalties (α); if ($\hat {\alpha }\geq vt\geq \hat {\upbeta }$) then the provider system failed to enter a risk state and will earn the money ($\hat {\upbeta }\leq vt\leq \hat {\alpha \upbeta }$) or will pay penalty ($\hat {\alpha }\geq vt\geq \hat {\alpha \upbeta }$). To avoid infinite penalties, the maximum penalty α is defined. We assumed only the three decision points for the provider to decide easy (see Fig. 2a) or more points to decide accurately its accuracy (see Fig. 2b) and may as a function, a table or a graph be complicated in the risk area [35].

Fig. 2

RSLA revenue as a function of violation time

Provider and consumer agents negotiate on $\hat {\upbeta }$, β, β₁, $\hat {\alpha }$, α, α₁ and $\hat {\alpha \upbeta }$ values (in Fig. 2a) to establish different QoS ranges for describing different revenues and penalties. The consumer agent assumes that vt should normally be close to zero and have to pay β most of the time. When the consumer agent begins requesting resources from the SGIP, it commences the next series of negotiations it sends: the RLA template to the SGIP that has $\overrightarrow {R^{i}}, \overrightarrow {L_{i}}, {\Delta } t$, γ and τ values. According to its prediction to satisfy the request, it returns a complete RLA_i to specify $\hat {\upbeta }$, β, β₁, $\hat {\alpha }$, α, α₁ and $\hat {\alpha \upbeta }$ values. If the SGIP has enough resources, they can be returned as a function of the risk level of the consumer agent interest, the number of resources and the Smart Grid market status. However, the result of lower risk level interest will be higher prices (β), but there are greater penalties with lower α and τ tolerance to RLA_i violations, that is a lower $\hat {\upbeta }$ and $\hat {\alpha }$. Assuming the program of SGIP consumer agent is accepted after the negotiations produce an agreement then the provider will automatically confirms the RLA_i, otherwise it rejects it and looks for another agent or program in its market.

4.2 Grid risk and RLA assessment

Risk assessment, as an automatic service, is a new paradigm for real-time risk measuring [1]. This follows the on-demand, automated, multi-agent view of the Smart Grid continuous risk score of the grid environment. We visualize such assessments as being made available by one or more of the grid risk entities. A provider can enforce continuous assessments through the evaluation of its own runtime environment. We present a new type of service called RaaS with a vertical layer or middleware that isolates all other services to the risk middleware rendering the risk assessment as a service. The service can be presented for both the consumer (as risk monitoring) and the provider sides (as risk evaluating) or others for decision making support. We present the SLA and RLA definitions as below.

• Service Level Agreement: SLA includes an agreement based on these service levels required by a consumer. Although consumers need high performance service level demands of a provider that is difficult to provide.

• Risk level Agreement: RLA contains an agreement with a determination of the risk levels required by a consumer based on an SLA contract. Although consumers have different demands of services from a grid provider, the required risk level is always different based on the budget or risk freedom degree of the RLA. The grid provider can define the risk bound to each different time service of an SLA and change it to an RLA contract.

We perpend risk as the effect of the uncertainty on both BLOs and the consumer’s budget which depends on the probability of an undesirable event and the impact of the desired results to its degree of variation or risk freedom. In a certain time frame, the occurrence of an event may impact on desired outcomes. For example, if one storage fails within the redundant storage system, an event has occurred, its impact is the high cost of replacement and the power loss with variation on the penalty. The impact of risk is determined economically by penalties and defined in the SLA. The calculation of risk is a difficult process which depends on calculating the FP of a complex system of the sequence events and how this failure could affect the SLA satisfaction terms. This is completed with agreements on the penalty variation or the degree of risk freedom degree in a negotiation paradigm.

Our proposed RSLA lifecycle model is such as the ITIL framework with a phase different on negotiation (see Fig. 3) [15]. General service lifecycle management is comprised of four phases with six steps: Construction Phase (Discover Service Provider and SLA Definition) for service construction to start an online agreement, Negotiation Phase (service negotiation) for beginning an online negotiation to an agreement, Operation Phase (service operation) for online service representation to provision, and Destruction Phase (service termination and service enforcement penalty) for terminating an open negotiation and the penalty calculating. Our RSLA model is added RLA Definition in Construction Phase, Setup RLA in Negotiation Phase and Monitor RLA in Operation Phase. RSLAs also play a main role in the design, usage and delivery of the service lifecycle, because the risk-aware service driven occurs by automatically securing service expectations, online calculation, online reconciliation and revenue responsibilities (see Fig. 3). Once an RLA negotiation has been identified and assessed, we apply the method to manage the risk factor so that, based on BLOs, it will fall into one of the eight under mentioned categories (see Fig. 3a):

• Avoidance (withdraw the activity).

• Mitigation (optimize or reduce the uncertainty).

• Retention (accept the uncertainty and budgeting).

• Transfer (transfer to reassess based on RLA).

• Killing (remove the source of uncertainty).

• Tribulation (valuate the uncertainty).

• Forwarding (forward the uncertainty and budget to third parties).

• Accepting (accept the activity and budget).

The negotiated risk process takes place after identifying, analyzing and evaluating risky events to threats minimized and opportunities maximized. This process will involve full management, usually including the appropriate responses, benefit controls, and needed actions of risk-aware strategies which are very important in risk management, since the impact of penalty variation of any event may depend heavily on chosen risk cost strategies. It should be noted that the best response when addressing the changing risk penalty over time depends on the current status of the Smart Grid provider, utilization of its in-house infrastructure or received workload. This step involves predicting the risk level estimation for each potential risk response for making risk response better upon alignment with its BLOs, given that its current negotiation status which is a main challenge for the automation process in order to reduce the penalty variation of the SLA to define a safe negotiation period.

Fig. 3

Seven steps evolution of RSLA in three levels lifecycle management

The day-to-day presentation of uncertainties can cause the consumer to consider both the positive and negative impacts to the Smart Grid business. Traditionally businesses have been proposed strategies geared to dealing with risks that represent threats. However, it also has value in exploiting events that may represent opportunities. For instance, two status could lead us to higher revenue: (1) the number of consumers could be higher than expected, if we ensure that we can attend to all of them; (2) consumers could finish their request earlier than expected, if we use free resources to attend to the number of consumers. Depending on the risks nature, we distinguish between two types of mitigation and adoption strategies to deal with negative events (threats) and positive events (opportunities). Overall, the discipline of a risk-aware management evolves from the RLA assessment. This is a central process of risk-aware management for providing outputs in the form of Risk-Level Estimations (RLEs) to be defined individually for each risk. The task is a kind of probability analysis to define the correlation between specific threats to specific resources for describing the cost or benefit of resource allocation. A calculated assumption to consider many things includes constant as well as possible threats and the probability of an undesired event; from causing cost variations to presenting a risk-aware status.

Risk analysis is the main key element of Smart Grid management; therefore, management of risk that is a process of comprehensive identification of threats. The analysis model is an important structure for analyzing all existing object risks in a secured system and the aim is the provisioning of information which is essential for decision making of application of specified methods to the security resources in the enterprize. However, the needs for control or acceptance of determining measures of the assets of the Smart Grid system must be determined previously. Our Freedom-based Risk Analysis Model (FRAM) is represented in Fig. 4 which is a quantitative method (e.g. Petri Net model) beside the parameters measurements. The proposed risk analysis process includes five tasks as seen below:

• Resource Evaluation: Valuation of physical and non-physical resources, where this resource value is the purchase value with the changes of goods value and is needed for insurance,

• Consequence Assessment: Consequence analysis must define the degree of destruction or losses, which can apparently occur,

• Threat Identification: threat analysis must determine its probability of occurrence and resource destruction possibility as well as a freedom degree possibility,

• Probability Assessment: the frequency of threat occurrence, which should determine the duration time and strength of the threat beside the protections effectiveness as well,

• Freedom Assessment: the definition of the freedom degree of treats or behavior changing, based on RLA, which can apparently occur and change.

Fig. 4

FRAM Risk Analysis Model

We propose a risk analysis method by using the Petri Net model as a meta model, in order to carry out the first phase of the Freedom-based risk-level estimations (FRLEs) with a quantitative analysis approach. This provides a meaningful perspective on how to analyze risk at all levels and all types of risks in Smart Grid technology. Note that the above given values to quantitative measures of risk are probability occurrence, impact factor and freedom degree.

4.3 Petri net modeling of risk

Petri Net is a formal model to represent and analyze asynchronous, distributed, parallel or stochastic processes that is a complicated process (see Fig. 5) [2]. To support and fulfill the needs of these different processes, various classes of the strong model have been proposed such as timed PetriNet, stochastic PetriNet, hybrid PetriNet, functional PetriNet, and Hybrid Functional PetriNet. This mathematical model consists of three tuple classes (places/conditions, transitions/events, and arcs/connections) to model and simulate a system [23]. In Petri graphical representation, a place is shown as a circle, transition as a rectangle or bar, token by a black dot and arc by an arrow (see Fig. 6a). To understand the graphical notation of the model see Fig. 6b and the text notations see Table 1.

Fig. 5

Petri Net Process Modeling and Analysis

Fig. 6

Petri Net Graphical Modeling

Table 1 Notations and its’ descriptions of petri net modeling

The simple Petri Net model can be presented as a directed bipartite graph to connect its places and transitions by arcs. A place p is called an input node of transition t if an arc directly links p (t or $p\rightarrow t$) and is called an output node of transition t if it directly links t to p ($t\rightarrow p$). The arc is labeled with its weight as a positive number and the default value is equal to 1 that is not shown as a label that can link two nodes of the different classes. The components of the system are modeled by places for passive components to execute the Petri system model and transitions for active components to change the system’s states during the Petri execution. Places can contain tokens that the position and number of the tokens may vary and transitions model the system activities that can change the system states by firing the enabled transitions. Firing an enabled transition deletes tokens from input places and inserts to output places based on the cardinality of each connected arc. A transition is specialized in inference/aggregation operations of logic rules e.g. Fuzzy rules (see Fig. 6c).

Our risk model is a graph of nodes and links to show the dependency between them (see Fig. 7). The n_i node fails when it does not satisfy the agreed QoS (e.g. inability of storage to charge or discharge, inability of a computational resource to satisfy defined efficiency, failing network links and etc.). Failure propagation (FP) of n_i is notated as P(n_i) to be measured according to the previous description. Let n_i and n_j be two connected nodes that work together as a total system. We define n_i as a risky link to n_j with weight w_ij when the failure of n_j prevents n_i from working correctly (for example, n_i is a power generator to charge n_j storage) where w_ij ∈ [0,1] is the probability and the failure of n_j propagated to n_i. As a consequence, failure of n_i can be caused by internal failure in n_i or a propagated failure of n_j to n_i with probability w_ij. $\acute {P}(n_{i}$) is defined as the propagated probability failure of n_i in (1). Equation (1) is modeled on the basis of the Union Formula of probabilities in (2), where it assumes that P(n_i) and P(n_j) are dependent (unlike, P(n_i) and P(n_j) which are independent in (2)). The graphical modeling of such a risk relationship is shown in Fig. 7a.

Fig. 7

Risk Graphical Modeling

We can use the above notations as an initiative to calculate risk in large and complex systems. For example in Fig. 7b, let cha be a power generator to handle requests from consumers and store power in storage system sc. We assume that 30% of the power is consumed by the converter con and $w_{scon}=\overline {w_{scon}}=0.3$. If con fails, the failure will be propagated to sc and to cha respectively. In the above mentioned experiment we posit that P(cha) = 0.05, P(sc) = 0.01, and P(con) = 0.03, and the weights of the output arc from cha and the input arc sc are equal to 1 (see Fig. 7b). On the consumer side, if the node cha fails, the total power system will fail. The FP of the complete system is $\acute {P}(cha)$ and calculated in (3). In resolving this problem, the probability that the complete system will fail (that is consumer cannot access cha) is equal to $\thicksim 0.068$ and is true for $\acute {P}(n_{i})\geq P(n_{i})$. In this exercise, the probability of the failure of node sc that was propagated to cha is really equal to the probability of the system’s failure organized by sc and con. It is for this reason that $\acute {P}(cha)$ is calculated as a function of $\acute {P}(sc)$ rather than P(sc) in (4).

We can model a complex system in a multi-level modeling to allow the grouping of many nodes to a single meta node as a subsystem in other levels to explain the risk dependencies. Multi-level modeling of the Petri Net explains how a node can have risk dependencies to other nodes in its own level or that others which is appropriate to union and intersection modeling of a risk graph. Therefor we propose two types of meta nodes (nodes that contain sub workflows, i.e. in the workflow they look like a single node, although they contain many nodes for defining unions and intersections between risk probabilities. Finally, our model explains that as the system is headed by n_i, it will fail because of the failures in n_i or in n_j (with probability w_ij) or in n_k (with probability w_ik) (see Fig. 7c).

$$ \begin{array}{@{}rcl@{}} \acute{P}(n_{i})=P(n_{i})+ w_{ij} P(n_{j})- w_{ij} P(n_{i})P(n_{j}) \end{array} $$

(1)

$$ \begin{array}{@{}rcl@{}} P(n_{i} \cup n_{j})=P(n_{i})+ P(n_{j})- P(n_{i})P(n_{j}) \end{array} $$

(2)

$$ \begin{array}{@{}rcl@{}} \acute{P}(cha)=\acute{P}(cha)+ \acute{P}(sc)- \acute{P}(cha)\acute{P}(sc) \end{array} $$

(3)

$$ \begin{array}{@{}rcl@{}} \acute{P}(sc)=\acute{P}(sc)+ 0.3\acute{P}(con)- 0.3\acute{P}(sc)\acute{P}(con) \end{array} $$

(4)

$$ \begin{array}{@{}rcl@{}} \acute{P}(n_{i})=\acute{P}(n_{i})+ \acute{P}(\cup)- \acute{P}(n_{i})\acute{P}(\cup) \end{array} $$

(5)

$$ \begin{array}{@{}rcl@{}} \acute{P}(\cup)=w_{ij}{P}(n_{j})+ w_{ik}{P}(n_{k})- w_{ij}{P}(n_{j}) w_{ik}{P}(n_{k}) \end{array} $$

(6)

$$ \begin{array}{@{}rcl@{}} \acute{P}(Master)= P(Master)+ \acute{P}(Slaves)- P(Master)\acute{P}(Slaves) \end{array} $$

(7)

$$ \begin{array}{@{}rcl@{}} \acute{P}(Slaves)=P(x\cap y)+P(x\cap z)+ P(y\cap z)- 2P(x\cap y\cap z) \end{array} $$

(8)

We introduce the union operator ”∪” as a virtual transition (P(∪) = 0) to form the subsystem n_j and n_k and treat it like a meta node when calculating the risk propagation to n_i in order to find $\acute {P}(n_{i})$ in (5) to compute it with $\acute {P}(\cup )$ in (6). For example, imagine that n_i is a virtual storage that executes a storage intensive task against total storage system which distributes a load within two storages (n_j and n_k) to improve performance. If only one storage fails, the system will fail because there is no redundancy for the consumption of power. As a storage intensive task, we assume w_ij and w_ik equal one. If the failure probability of a single storage for a given time period is 6% then the system fails with $\acute {P}(\cup )= 0.1164$ probability. This model introduces an intersection operator ”∩” to the model redundancy in fault-tolerant Smart Grid systems as well (see Fig. 7d). The failure probability of a subsystem operated by the node ∩ is the intersection of failure probabilities of nodes n_j and n_k. The result for these independent storages is $\acute {P}(\cap )=w_{ij}w_{ik} P(n_{j})P(n_{k})$. In a master-slave system it could be supposed that a master node which is represented by ”M” sends power to three slave nodes which is represented by ”S”(Slave_x, Slave_y, and Slave_z), if one of the slave nodes (S) fails then the other two remaining nodes will handle the work and if these two slave nodes fail, the system will fail. This system failure probability is calculated as (7) and (8).

4.4 Risk-aware smart grid operation

In the Smart Grid we aimed to keep focus on the risk and revenue model to simplify its graph risk model by experimenting only with one application template. Risk must be considered during resource allocation and other Smart Grid operations. For example, a consumer, needing high availability, would negotiate the SLAs with a high penalty to the provider in case of SLA violation. In such a scenario, the Smart Grid provider would have to minimize the FP of its resources according to two of the complementary risk minimization strategies that entail an increment in the operation cost. Those power companies that enforce risk minimization policies always apply the same action. A graph is used to add redundancy to nodes whose failures would result in a failure to the rest of the system. The two above mentioned scenarios are explained below.

• For node n_i, the minimizing P(n_i), the risk in the node is not propagated. Hardware resource lifetime and workload are two main factors that influence P(n_i). The failure rate of hardware is high at both the beginning and the end of resource’s life time. This failure is directly correlated to the workload which is higher during peak hours and lower during off-peak hours. The statics in the statistical analysis of P(n_x) are computed based on historical data.

• For node n_i, the decreasing $\acute {P}(n_{i})$, is propagated in the graph. Risk graph analysis and redundancy provided in the critical paths and nodes of the graph would remarkedly reduce the total risks of system with reasonable economical performance. The risk propagation model is an important model in the field of research that machine learning, pattern recognition or new models to this problem be applied [33].

4.5 Smart grid revenue modeling

We describe a revenue model in an automatic SLA negotiation and in other operations, as well, that would allow providers to consistently contribute differentiated risk levels according to their business objectives. The price of a requested service for a consumer or a maximum revenue for a provider is defined as MaxReven in Eq. (9) and the price of Smart Grid resources is set for a constant time frame (see Table 2). Where the consumer can reserve a resource with a minimum price that a provider can sell that resource without losing money is shown as ϖ with an award ξ as a function of the negotiation time which is max for an online negotiation. Another constant award is paid to the consumer for the RLA contract based on the agreement level and the critical time sheet called ζ. A consumer demand/offer overprice can be shown as ψ and he pays more when there are more requests than offers (if the request is much lower than the offer then ψ = 0). A business value is the amount of money that consumer must pay for extra QoS unit called χ. We calculate the lower amortization cost of all resources that the consumer is using during a certain time period as ϖ and the addition of the amortization costs of all used resources for Smart Grid as (${\sum }_{i=1}^{k} \varpi _{i}$). The calculation of amortization cost of a single allocated resource is modeled as (10).

$$ \begin{array}{@{}rcl@{}} MaxReven=\varpi -\xi +\psi + \chi -\zeta \end{array} $$

(9)

$$ \begin{array}{@{}rcl@{}} Cost_{amort}=(T_{cost}- I_{amort})\frac{Delta_{time}}{(L_{Trg}-L_{Tnow}) P_{usage}^{\gamma}} + {P_{sac}} \end{array} $$

(10)

The cost of initial investment plus general maintenance expenses during the entire lifetime of a resource is called total cost of ownership shown as T_cost. I_amort is the sum of all incomes associated with the provisioning of virtual resources for given resources. Delta_time is the amount of time that a consumer is willing to use a resource. L_Trg is the planned life time for a given resource group: the time from whence a resource is provisioned until it is disengaged from the Smart Grid. L_Tnow is the lifetime since a resource is provisioned until now. Also, γ =[0,1] is a weighted function that shows the ratio of a resource’s group to which the cost is being calculated. For example, if any given consumption requires 9 batteries from a node with 45 butteries, γ = 0.20. The P_usage parameter is the percentage of the use of resources as anticipated by the provider to this time. If P_usage = 1, all the assigned Smart Grid resources to consumption are at full capacity during this time. The value P_usage would increase proportionally the reservation price that is needed to actually amortize a resource completely at the end of its lifetime if the resources are underutilized. Finally, for a resource, P_sac is a percentage of the subscriber acquisition costs that is incurred by Smart Grid to convince a potential customer and involved in research, marketing, and accessibility costs. However, to avoid inequality when amortization in each resources having the same type and age, an account is marked for them. The values of T_cost, I_amort, γ, L_T and P_sac are applied to every resource not just those in individual resources.

Table 2 Notations and its’ descriptions of revenue modeling

Equation (10) is different from traditional calculation of the amortization cost, T_cost/L_Trg, because this formula takes full load usage into account; however, it does not consider the resource value that decreases over the time. The ψ overprice must be low when the offer ratio is high enough to let consumers choose from a large enough set and it is high only during peak hours in which most resources are being utilized and at a time when the Smart Grid is probably in a risky status. An accurate estimation of χ depends on those hidden variables related to consumer behavior, market status, the reputation of the provider, etc however this issue will not be discussed in this paper. In experimental results, a fixed overprice for SLAs were applied according to the level of QoS and Risk. We mention that ψ and χ as being equal to the total amortized cost and represent overprices. The acceleration of the amortization of the resource causes decreasing of Cost_amort over time for different pricing which depends on the age of used resources.

5 Simulation results

We present a Case Study to evaluate all applied models and their impacts on the proposed SLOs when incorporating an offered risk treatment into its self-managed operation (Section 5.1). Also, we have experimented tasks to enable a provider to fulfill the profit maximization and consumer satisfaction for two critical Smart Grid related risks as the risks of provisioning and resource failures, with different workloads and consumer demands as in Applied Models (Section 5.2). We have computed the fulfillment of these two service level objectives to the different risk treatment detailed above as Results Evaluation (Section 5.3). Three different SLA allocation strategies Medium, Low and Very Low are used to offer three levels of SLA risks.

5.1 Case Study: Electrical Energy Storage Systems

As more renewable energy is developed, energy storage is increasingly important and attractive, especially grid-scale electrical energy storage; hence, finding and implementing cost-effective and sustainable ESS storage and conversion systems to reduce risk as a case study is vital [32]. The ESSs not only participate in the backup power supply but also have the potential to provide various distributed ancillary services. The aim of this section to show the suitability of the proposed risk management process for explained Smart Grid organizations to assist a risky decision-maker to improve the achievement of their tasks. In this case study, we experiment with several risk response of some critical infrastructure-level risks to risk management infrastructure. The Smart Grid can deploy several types of topology to charge circuit graphs. In our experiments, providers deploy a simple charge circuit graph arrangement according to the structure in Fig. 8: a Charger gets power from the power source to charge the supercapacitors across a set of n homogenous ESSs, as Supercapacitor₁, Supercapacitor₂, Supercapacitor₃, and Supercapacitor₄, which is connected to other ESSs with a flat topology via power converters namly a charger and a DC/DC converter, respectively. The increased penetration of renewable energy sources has prompted the integration of battery energy storage systems in active distribution networks. We link a direct connection among ESS elements based on the compatibility of supercapacitors, voltage in the terminals and the power ratings of the same elements [17, 20, 30]. The application of ESS elements in a hierarchy where energy is stored in addition to a cost-effective policy is a promising means of improving the outstanding features of the system reducing its amortized costs. Supercapacitors enjoy a exceptional cycle efficiency that reaches roughly 100% efficiency which can be utilized in an extended life cycle. These are used extensively to assuage the fluctuations of the load current in batteries that when compared with other batteries, show a considerably higher volumetric power density but exhibits lower energy density [25]. This of capacitor may use up or lose more than 20% of the energy it has stored per day although no load is connected to it and the voltage in the terminal increases or decreases linearly proportional to the charging or discharging of the capacitor [31].

Fig. 8

Simple Electrical Energy Storage System

5.2 Applied models

In this section we propose all that applied models be shown as a graph (Applied Graph Model), node risk (Node Risk Minimization (NR_min)), graph risk (Graph Risk Minimization (GR_min)), pricing (Risk-aware Pricing Model), cost (Cost Minimization (CO_min)), stakeholder (Stakeholder Model) and supercapacitor failure (Supercapacitor Failure Model) as below.

Applied Graph Model

The consumer can model its suitable arrangement by deploying several types of capacitors that, in our experiments, the consumer deploys a simple Direct Acyclic Graph (DAG) structure according to Fig. 8: a charger (P_ch) balances power across a set of n supercapacitors (P_sc1,⋯ ,P_scn) that use a converter (P_c0) as a persistence loader. The supercapacitor set varies from 2 to 4.

Node Risk Minimization

Given the two weighted criteria, low consolidation and resource age, the power provider prioritizes the allocation of those capacitors. The first is devised to decrease the risks that emanate from overloaded resources which would not be necessary to supply the concurred QoS and the second would try to evade the selection of new resources that nearsighted the end of lifetime.

Graph Risk Minimization

The link is analyzed by the provider to detect the single point of failures and therefor applies NR_min. Given the model in Section 4.3, a detected failure in the node would result in the failure of the entire provider system; thus it resolves to replicate it. The consumer selects the sort of risk minimization strategy as a function of the risk needs of its requirement. Where as the provider calculates the price, it utilize a fixed overprice of 50% to the NR_min SLAs and 100% to the GR_min SLAs in order to determine β (see Fig. 2). Lower values for α, $\hat {\alpha }$ and $\hat {\upbeta }$ suggest that there is less tolerance to failures for low-risk SLAs (see Table 3), due to the fact that vt will reach $\hat {\upbeta }$ sooner (see Fig. 2a). The risk level determines $\hat {\alpha }$, α, α₁ = 50%α, $\hat {\upbeta }$ and β₁ = 50%β in order to evaluate the model with regards to the relative results and tendencies. The all needed parameters setting is set up based on grid environments.

Table 3 Revenue parameters for each group of SLAs

Risk-aware Pricing Model

The proposed Risk-aware Pricing Model (RPM) is contracted between the power producer and consumer. It is υ% agreed tolerance among the producer reservation price (PRP) and the consumer reservation price (CRP). The PRP is registered as the maximum price and CRP is registered as the minimum price in the RLA document. Despite the fact that the provider and the consumer know their own RLA content about the risk tolerance (υ%), the consumer will not convey its request price to provider; consequently it can only be approximated in function to historic prices as (11) and other markets.

$$ \begin{array}{@{}rcl@{}} Price_{RPM}=PRP + (CRP - PRP) \times \upsilon <percent> \end{array} $$

(11)

Cost Minimization Model

Given two weighted criteria: (1) high consolidation to save energy costs which are loading tasks even now and keep idle switches off; (2) amortization to allow lower prices in the models presented in (9) and (10) the Smart Grid provider prioritized the capacitors. For SLAs allocation in higher risk, the high consolidation and resources age are chosen.

Stakeholder Model

The stakeholder consists of a power provider, a consumer, a network provider and a third-party. The first one wants to use the provided power services in the SmartGrid; the second tries to deploy their services on the SmartGrid; the third offers consuming environments to Smart Grid services by a middleware; and the fourth outsources consuming services to other providers.

Supercapacitor Failure Model

Failure is defined as: (1) Capacitance loss greater where 20% that the lack of ability of the capacitor may be the result of one or many faults, (2) Series resistance increase bigger than 100%, (3) Leakage power and (4) Cell opening [19]. This can cause Aging (normal, high voltage or high temperature) with Weibull distribution, Overvoltage (abnormal), Fire (abnormal), and Shocks and Vibrations. For example, the failure mode is a cell opening consequences as Grid stopping, Grid damaging and Performance losing.

5.3 Results evaluation

We present a setup (Simulation Setup) and evaluate the results of cost, node risk and graph risk to risk scenarios (Risk Minimization), revenue calculation (Revenue Modeling) and risk treatment (Risk Management) as below.

Simulation Setup

For the sake of approving our proposed QoS model, and its interrelationship with risk-aware business strategies, we choose Zabbix as our Tool simulator which adopts the simulation environment and methodology to our work [37]. The adapted platform is able to model monitoring and tracing the energy consumption and SLA performance with the automatic provider over/under-loading detection. Also, we can reduce the preparation time of the simulation environment mainly to focus on the decision-making of SLA violations and the corresponding tasks. We simulated the Smart Grid power center which included 9 power providers to 172,800s simulation time with 10 min interval time (for the utilization measurements). Initially the number of services deployed oscillates between 3 to 40 services/hour (as a function of a specified hour of the day and a specified the day of the week). Each provider includes minimum 4 physical supercapacitor (Psc). Tool environment capability allows us to design and define a more novel stronger storage than the available technologies.

Risk Minimization

The graphics of this part exhibit average values weekly so that they will be more transparent and understandable, as hourly or daily averages are profoundly influenced by the workload oscillations. The weekly granularity the attained value is accurate as well as since the simulation is period is long enough (30 months) to exhibit the tendencies of the metrics used to evaluate the efficacy of the scenarios discernably. The behavior of the scenarios vis-a-vis the age of selected resources is displayed in Fig. 9a. All the resources are the same age in the first half of the experiment. When a new agglomeration of resources is introduced in month 10 (week 42), the CO_min will still select older resources having the highest amortization rates. After a brief period in which new resources have higher risks in comparison to older resources, NR_min and GR_min will move their workloads to the new resources, progressively. As explained before, the number of requests increases over time, linearly. In requests around week 82 resources are highly loaded and due to the fact that the provider has a weaker possibility of choosing resources for the various risk levels, resources are highly loaded around the week 82. As shown in Fig. 9b and c this will influence the risk of the SLAs. All extracted results of this section are set based on the setup and parameters of the Smart Grid environment and we move the setting to nanotechnology behavior (e.g. short failure rate) to view the grid organization view [26].

Fig. 9

Evaluation of Risk Metrics for Different SLA Scenarios

Figure 9b displays the weekly average PoF, of the SLAs, which are differentiated from one another via three diverse allocation policies (CO_min, NR_min and GR_min). Although CO_min is close to a constant over time, NR_min maintains a much lower risk than CO_min while it is indubitably influenced by the load of the above mentioned resources. The PoF for the NR_min of the SLAs rises linearly overtime as the amount of capacitors also increases because the prospect of choosing is reduced. When the amount of resources is doubled, the PoF of the NR_min will be reduced again; however, the PoF of the CO_min is kept constant for the reason that the allocation policy is picked the older resources. GR_min SLAs is also sensible to the load of resources for the reason that the allocation policy is the same as NR_min. On the other hand, the elimination of the single point of failure compels the system to maintain much lower risk rates. The PoF has a direct effect on economic penalties as a result of infractions by the SLAs. Figure 9c shows the rigorous correlation between these economic penalties and the probability of failure. The economic percussion of these failures is SLAs higher in allocated with low-risk policies (NR_min and GR_min) owing to the fact that both prices and penalties are higher for these SLAs (see Table 3). In our evaluation Fig. 9c, as well as the remaining figures processing economic information does not show absolute economic values; rather it shows values divided by Storage hours in order to assist in the comparison of data from appliances having different sizes and times, respectively.

Revenue Modeling

Figure 10a displays the average Storage price/hour for various kinds of SLAs. During the initial step, the price decreases due to the fact that resources are being amortized which, when the pricing model related to (11) is applied, reduces ϖ. On the other hand, Fig. 10b demonstrates that during this period the profit (revenue minus reservation price minus penalties) for the CO_min policy can be improved as the market permits a higher profit margin: ϖ decreases; however ψ and χ might increase at a lower relative amount which may decrease prices while increasing the profit margin. On the other hand, the profit for risk minimization scenarios remains close to the constant during this period of time (despite the reduction in the price ) since the business value (χ) of low-risk SLAs is higher and son users are willing to pay more for this additional QoS). As as consequence the provider has more leeway to raise its margin profit with risk minimization scenarios. Figure 10a demonstrates that when the middle of the experiment is reached, prices will increase as well because they have a tendency to designate and or allocate new resources with a lower amortization rate where resources are doubled. The CO_min policy keeps prices at a constant rate due to policy which allocates services in older resources. Despite the fact that the penalty rate increases for minimization in the first half of the experiment (from week 1 to week 42) as shown in Fig. 9c, the effect of these penalties on the net profit is relatively low due to the various scaling of Figs. 10b and 9c. This is exactly why the profit for NR_min and GR_min decreases slightly over time. This having been side, when new resources are added, the increment of incurred price and the reduction of SLA penalties, would have some positive effect on the net profit. The impact of the penalties in CO_min SLAs continue to be quite stable during the entire simulation.

Fig. 10

Modeling of the Revenue

Risk Management

We present the effect of our risk treatment method on the providers’ profits and consumer satisfaction. We correlate this method with a risk-averse strategy and present experimental results with different inbound consumer demands in order to express the maturation of both provider and consumer BLOs, in any case in a given month (See Fig. 11). The results present a risk-aware strategy that is well-defined and a treatment response or response by taking the current usage of the Smart Grid into consideration(see Fig. 11a and b). Just as shown along the x-axis of the diagram, both figures start a given month with a low demand that permit it to turn off several physical nodes of a minimum number of online nodes which permit the provider to maximize the profit due to violating of roughly 11% of SLAs which are shown in Fig. 11b. Presuming that consumer satisfaction is more important than profit maximization, the Smart Grid provider may choose a different risk strategy in order to favor the above mentioned BLO. Then it would experiences an increase in demand which would lead to it being completely filled. A risk adoption response for sharing risk under-provisioning is therefor applied and an out-source to the third-parties do not fit into the Grid resources. Despite consumer satisfaction (see Fig. 11b) risk-aware response guarantees SLA fulfillment of all services that will lead to maximized profit earned by the provider (see Fig. 11a). Risk-aware policy causes half of these nodes to be turned off in order to match the time-varying demand with the number of online nodes due to dropped incoming demand and the over-provisioned grid. Risk-averse response policy forces the provider to pay a lot of penalties to its consumers for very high SLA violations number. Finally, the provider receives a three days peak in consumer demand which is again addressed by means of out-sourcing deals with a third-party.

Fig. 11

Effects of Risk Managing

6 Conclusion

This paper introduces a risk-aware intelligent automated agent-oriented model to differentiate SLAs in the Smart Grid market for cost and risk minimization at individual node that decreases the risks emanate from overloaded resources and tries to evade the selection of new resources that nearsighted the end of the lifetime and graph levels that does less tolerance of failures for low-risk SLA. The model optimizes the resource management according to BLO of the provider to an online risk-aware rendezvous to define the penalty level of the cost model based on a new risk level agreements contract that automates the agreements calculation to smart agents. Also, an accounting model is presented to allow the provider agent to adjust prices the risk as a function of the resources amortized that presents the intelligent adjustment. The novelty of this method is proposing a new RLA contract and a pricing model to decrease the complexity of previous methods from an off-line method to an on-line risky method. The simulation of the presented real-time concurrent negotiation and smart grid revenue modeling minimize the risk of a graph on a case study that decreases the price and increases the profit margin, however it is a time-consuming and costly process. Our future risk propagation model will be produced to support bidirectional dependencies and cycling in the risk graph to allow the provider agent to automatically minimize risk in required nodes or paths of the risk graph to prevent the costs from soaring, due to the excess of redundancy. Regarding revenue modeling, our future producer-consumer model is proposed to discover the business value of Smart Grid resources to accurately estimate ”how consumers are willing to pay additional QoS to maximize profit of the provider” without losing the consumers.