Information exchange based clustered differential evolution for constrained generation-transmission expansion planning
Introduction
Generation, transmission and distribution-demand management are three main sections of a typical power system or an electric grid which aims a seamless and reliable delivery of electricity to end users. While it is important to operate different parts of the grid optimally to ensure reliable and economical delivery of energy everyday, it is also important to plan future expansion of these different resources of a grid to ensure the same in a future scenario with an elevated demand that is projected to grow at a rate greater than 30% till 2040 across the globe [1]. Thus optimal expansion planning of power system resources has become a very important problem for all the parties involved in power system along with the widely studied power system operation problems like unit commitment, economic load dispatch, etc.
Expansion planning of the resources in an electric grid is modeled as optimization problem. Traditionally in a vertically integrated system, the generation resources are planned first followed by the transmission resources i.e. the optimization problem is solved sequentially. On the other hand, in co-optimization models of planning both the generation and transmission resources are planned simultaneously to discover potentially better solutions that may be overlooked in sequential planning [2]. In conventional grid planning models the load is assumed to be constant over time which is no longer valid with recent advances in the area of Demand Side Management (DSM) [3,4] and Demand Response (DR) [5]. Investments in form of Internet of Things (IOT) based DSM/DR controllers in the distribution side [6] are making the dispatchable load possible. Such flexible loads can bring down the peak demand levels across the grid, which makes investment in demand management resources are cheaper alternative to investing in generation resources [7]. A co-optimization model with generation, transmission and demand management resources being planned simultaneously is an important problem for the following reasons:
- •
Generation, transmission and demand management resources are complementary. The load can be met by a local generator unit, or by supplying power through transmission lines, or controlled with demand management tools.
- •
Investing in demand management tools are cheaper alternative to the same in generation side.
Co-optimization models have been investigated by researchers [8] and also by Independent System Operators (ISO) who mandate considering the interaction between resources in long term expansion planning [9]. In a vertically integrated system co-optimization is used to expand the scope of planning problem to capture the trade offs by generation, transmission and demand management resources. In restructured power system, co-optimization model is used by transmission planners for anticipative planning [2]. Many off the shelf softwares for planning have used co-optimization models like COMPETES, ReEDS and LIMES but they do not consider the fuel transportation infrastructure to add into the cost [2]. Although consideration of the fuel transportation infrastructure cost while doing a co-optimization based generation-transmission expansion planning could be found in Refs. [10,11]. Some of the recently developed co-optimization softwares like NETPLAN and Prism 2.0 also consider the cost contribution by fuel transportation infrastructure [2]. Recent works on generation-transmission expansion planning have also considered co-optimization with generation-transmission and demand management resources keeping the current trend of smart grid technologies in consideration [12]. Although the authors in Ref. [12] did not consider any cost from the fuel transportation infrastructure in the model. This paper solves a co-optimization problem with generation-transmission and demand management resources while including the cost incurred for new fuel transport infrastructure investment also.
The co-optimization expansion planning of the grid is a complex, large scale, constrained cost minimization problem with many hard constraints like line flow limits, power balance at each node of the grid, corridor limits, etc [10]. Classical solution techniques for solving co-optimization planning problems includes Bender's decomposition [12], backward-forward search [11], Mixed Integer Linear Programming (MILP) [13], etc. Classical techniques have fast convergence and simple form of representation. However, classical techniques suffer from poor quality (PL approach), problems with system dimensionality (linear and dynamic programming), and their execution time for classical techniques may increase exponentially with increase in system dimentionality (branch and bound) [14]. These drawbacks inspired the application of many bio-inspired meta-heuristic techniques to the expansion planning problem. Meta-heuristic techniques are easy to implement as they do not need to convert variables into optimization programming set [15].
Bio-inspired meta-heuristic techniques are developed by imitating natural processes and a large number of works have been reported in literature solving different power system problems using these techniques. Use of such metaheuristic techniques in solving expansion planning problem can also be found in literature e.g. Particle Swarm Optimization [16,17], Bee colony optimization [18], Genetic Algorithm [19], Differential Evolution [20], NSGA-II [21]. It is evident from recent literature surveys [[22], [23], [24], [25]] that improvisation and hybridization in meta-heuristic algorithms can help them achieve better performance. Studies like [[26], [27], [28], [29], [30]] reveal that specially designed heuristics are always welcome and better performing for large scaled, hard-constrained optimization problems.
Differential evolution is a very popular bio-inspired metaheuristic tool used in a large number of works for solving constrained Variants of Differential Evolution (DE) are very successful in solving the constrained optimization problems, which is evident from the fact that most of the entries in competition for constrained optimization benchmarks in IEEE Congress on Evolutionary Computation (CEC) 2010 and 2017 were DE variants. One class of such DE variants is based on clustering techniques and is popular for multi-modal and constrained optimization problems [22,25].
This paper proposes an Information Exchange based Clustered Differential Evolution (IE-CDE) which is mainly inspired by ensemble strategies [31] and refresh gap techniques [32]. The proposed algorithm is tested on CEC 2017 benchmark functions for constrained optimization problem and extensively compared with state-of-the-art algorithms proposed in CEC 2017 [[27], [28], [29], [30]]. IE-CDE is also tested on co-optimization expansion planning problem on a 6-bus and 24-bus system. IE-CDE was also tested on 87 bus North-North Eastern Brazilian system to show the scalability of the proposed algorithm to real systems. The main contributions of this paper are listed below:
- •
A novel IE-CDE optimizer for constrained optimization problems has been proposed
- •
A co-optimization model for generation-transmission-demand management resource expansion has been developed along with consideration of fuel infrastructure cost.
- •
Above mentioned co-optimization model based generation-transmission-demand management resource expansion has been solved with the proposed IE-CDE algorithm.
- •
IE-CDE has been tested on CEC 2017 constrained optimization benchmark problems and its performance on the same has been done with top algorithms of the constrained optimization competition of CEC 2017.
- •
Statistical significance analysis of IE-CDE has been conducted on co-optimization problem and comparison with state-of-art meta-heuristic algorithms has also been done.
The rest of paper is organized in the following sections: 2. Problem Definition and Formulation, 3. Proposed Algorithm: IE-CDE, 4. Analysis on CEC 2017 benchmark functions, 5. Test system for co-optimization problem, 6. Simulation results and discussion, 7. Conclusion.
Section snippets
Generation-transmission expansion planning problem definition
The hybrid generation-transmission- demand management resource expansion planning is done by the utilities as a cost minimization model for the entire power grid while considering the following costs:
- •
Generation Installation cost
- •
Transmission line installation cost
- •
Fuel-transport infrastructure reinforcement/installation cost
- •
Demand management resource installation cost
The objective is to achieve a grid plan that could cater to the future demands and is low on investment cost while satisfying the
Application to restructured power systems
In this section the application of Hybrid Generation-Transmission expansion planning in context of restructured power system is considered. In most of the electricity markets across the world, the transmission grid is owned by a state company which is an Independent System Operator (ISO). The ISO has an objective to minimize the total social cost in the generation expansion game by Generator companies (Gencos) [34]. A general frame work of transmission expansion planning uses a proactive
Constrained optimization using differential evolution
Differential evolution (DE) is a population based meta-heuristic optimization algorithm or evolutionary algorithm originally proposed by Storn and Price in 1995 [37]. Known for its simple nature, easy implementation and capability to find good quality solutions, DE Over the time has become a very popular optimization algorithm among the evolutionary algorithm research community and it has been applied to many real world problems including those of power systems. The original DE algorithm
Benchmarking of proposed IE-CDE on CEC 2017 functions
The proposed algorithm is first tested on a number of synthetic problems to enquire about its ability to handle constrained optimization problems. Hence it is applied on the 28 benchmark problems of CEC 2017 constrained optimization competition [51]. As a relative measure of the performance of the proposed algorithm, the results obtained from the same have been compare to those reported in the top 3 papers of the same competition [41,52,53]. To display the effectiveness of the proposed
Test system
The single-stage Hybrid Generation-Transmission Expansion with fuel infrastructure cost was tested on a Garver System. The Generation-Transmission-Demand management resource expansion planning problem was tested on Garver System and IEEE 24 bus system. For the Garver System the load data, line data and candidate data were taken from Ref. [54]. It was assumed that there is an existing plants in the grid. The Generator Bus details are given in Table 3.
In Table 3, “Limit” is considered as the
Implementation details
The proposed IE-CDE algorithm was applied to solve the Generation-Transmission-demand management resource expansion planning problem. The implementation required two steps:
- •
Encoding the decision variables into a member population vector
- •
Calculating the expansion cost and enforcing the constraints.
The encoding scheme is shown in Fig. 3. The first L elements of the vector are dedicated for
Generation-transmission expansion planning
The Generation-Transmission Expansion planning problem was solved with IE-CDE algorithm to get the best plan. A sequential GEP and TEP was also carried out to see the effectiveness of the hybrid planning process and a trade-off was observed that shows the supremacy of hybrid planning process. The results are tabulated in Table 6. It can be seen that Sequential planning was blinded at the GEP stage and could not figure out a trade off for a better optimum that the hybrid algorithm found and an
Statistical analysis
To test the efficacy of proposed IE-CDE, a non parametric statistical test: Friedman test, is used. IE-CDE is compared against DE, GA and GSA using Minitab 18. The Friedman test results are shown in Table 11:
A low p-value (p < 0.05) in both the cases show that the contender IE-CDE has a significant performance difference as compared to other algorithms, an low rank value indicates it's superiority.
Statistical details for cost outputs over 25 runs are represented in box-plot Fig. 7, Fig. 8.
Conclusion
The paper presents a novel Information exchange based clustered differential evolution algorithm (IE-CDE) to solve the power system expansion planning problem. The problem of power system expansion planning although is not as widely studied as the power system operation problems like economic load dispatch, unit commitment, etc. it is a very essential one for a secure future of the grid. The power system expansion planning problem has been formulated as a single objective minimization problem
References (56)
- et al.
Application of particle swarm optimization technique and its variants to generation expansion planning problem
Elec. Power Syst. Res.
(2004) - et al.
Recent advances in differential evolution–an updated survey
Swarm Evol. Comput.
(2016) - et al.
Noisy evolutionary optimization algorithms–a comprehensive survey
Swarm Evol. Comput.
(2017) - et al.
Computing with the collective intelligence of honey bees–a survey
Swarm Evol. Comput.
(2017) - et al.
Real-parameter evolutionary multimodal optimization—a survey of the state-of-the-art
Swarm Evol. Comput.
(2011) - et al.
Multi-population differential evolution with balanced ensemble of mutation strategies for large-scale global optimization
Appl. Soft Comput.
(2015) - et al.
Coordinated generation and transmission expansion planning in deregulated electricity market considering wind farms
Renew. Energy
(2016) - et al.
A genetic algorithm – differential evolution based hybrid framework: case study on unit commitment scheduling problem
Inf. Sci.
(2016) - et al.
Cluster-based population initialization for differential evolution frameworks
Inf. Sci.
(2015) - et al.
Cluster-based population initialization for differential evolution frameworks
Inf. Sci.
(2015)