A hybrid granular-evolutionary computing method for cooperative scheduling optimization on integrated energy system in steel industry

https://doi.org/10.1016/j.swevo.2022.101123Get rights and content

Abstract

Integrated Energy System (IES) in steel industry, typically involving gas, heat and electricity, exhibits distributed nature and complex coupling relationship, which demands effective scheduling optimization to cooperatively utilize the multiple energy media. For this purpose, a Granular-Evolutionary Hybrid Computing (GEHC) framework is proposed in this study to solve the scheduling of IES as a Constrained Multi-Objective Problem (CMOP). More specifically, a Collaborative Fuzzy C-Means (CFCM) clustering model is firstly established, where the horizontal and vertical structures are well constructed to obtain information granules regarding the cooperative relationship among the units in different energy subsystems. Then, a Probability-based Granular Computing (PGrC) approach is proposed for the initial optimization of the IES scheduling, which is guided by the expert knowledge compacted in the information granules. To determine the collaborative parameters in CFCM that essentially controls the cooperation among the multiple energy media, as well as satisfying the complex practical constraints, a Multi-Objective Differential Evolution (MODE)-based method is proposed for further scheduling optimization. To evaluate the performance on cooperativity, convergency and diversity, the proposed GEHC and the compared methods that either has partial collaborations or use other optimization frameworks are applied to a real-world optimization problem. The experimental results demonstrate that the proposed GEHC performs superior to the other methods considering both the objective functions and performance indicators, which can be potentially helpful to achieve well-performed unmanned IES scheduling in steel industry.

Introduction

To achieve better multi-energy complementation, lower emissions and higher utilization efficiency, Integrated Energy System (IES) becomes a primary implementation for energy generation, distribution, storage and consumption in steel industry. Comparing with the single energy system, the complex coupling and transformation relationship of multiple energy media in the IES, i.e., gas, heat and electricity, pose great challenges in solving corresponding cooperative scheduling problem, which currently draws great attention in both academia and industry.

A commonly deployed solution for the scheduling optimization of the IES is to establish a Mathematical Programming (MP) model and employ an evolutionary algorithm as the solver. For instance, a Mixed-Integer Linear Programming (MILP) model considering both the actual capacity of the energy devices and the energy conversion procedures was established in [1] for the optimal scheduling of the oxygen and nitrogen in steel industry. A day-ahead scheduling strategy for community IES was proposed in [2], which is also transformed into an MILP by linearization of nonlinear items. [3] reported a chance-constrained programming-based scheduling model of the IES and solved it by CPLEX solver. Due to the fact that the production in industry behaves as a continuous process, such statistic models cannot fully consider the dynamic characteristics of the IES. To overcome this drawback, some scholars have proposed dynamic models to solve the scheduling problem for energy system. For instance, [4] proposed a MILP-based model considering the coupling of power to gas, as well as achieving dynamic utilization and non-fugacity emission of carbon. Aiming to obtain an optimal trade-off between the minimization of the make span of production and time-dependent electricity costs, [5] reported a production scheduling method for improving industrial energy efficiency. All the above literatures generally depend on data or mechanism to model and optimize. However, the human experience from engineering practice has been ignored, which is able to provide prior knowledge naturally satisfied the practical constraints to speed up the convergence of the optimization process.

As a competitive paradigm for processing human-centered information, Granular Computing (GrC) has been widely applied to stock market [6], group decision making [7], etc. With GrC, high-dimensional data involving system dynamic characteristics can be conveniently compacted into a low-dimension unit, i.e., information granule. In the perspective of its applications in steel industry, GrC-based methods have been reported to be very effective for long-term predictions [8,9], energy scheduling [10], etc., As the energy resource reported in these literatures is only single secondary gas rather than multiple media, they do not have effective structures for the considerations of coupling and transformation mechanisms. With regard to the modeling of the relationship among different objects in different feature space, Collaborative Fuzzy C-Means clustering (CFCM) method was proposed in [11]. In the following studies, it has been combined with interval-valued fuzzy set [12], neural network [13], etc. A representative of its application for the IES of steel industry is to forecast the multiple gas tank levels [14]. The hybrid structure in [14] was demonstrated to be capable of giving accurate trend prediction results. Nevertheless, it neither takes any practical constraints into account, nor uses any effective optimization for automatically setting the parameters, which is very desirable for real-world applications.

In addition, it is of great importance to guarantee the satisfactions of constraints during the knowledge exaction and computing process. The constraints considered in our model is fairly complicated, involving not only the equalities defined by the flow equations of each energy network, but also the inequality constraints resulted from technical limits of the equipment, operating limits (e.g., maximum capacity) and performance targets (e.g., reliability, emissions, maximum allowed curtailment) [15]. This complex Constrained Multi-Objective Problem (CMOP) demands an effective and efficient Constraint Handling Technique (CHT). A number of CHTs have been reported in the literatures, such as reference vector-based [16], penalty function strategies [17], decomposition-based [18], descent direction-based [19], etc. [20] proposed a Push and Pull Search (PPS) framework to firstly explore the search space without considering any constraints and then pull the infeasible individuals to the feasible and nondominated regions. A multi-stage evolutionary algorithm for CMOP was also reported in [21], where the complex constraints were added one by one to handle at different stages of evolution. Different from the aforementioned studies where the decision variables explicitly exist in the objectives and constraints, the scheduling scheme in this study needs to be computed via the CGrC process. A simple and effective optimization approach is very desirable to solve this practice-oriented CMOP at hand.

In this study, a Granular-Evolutionary Hybrid Computing (GEHC) framework is proposed for the cooperative scheduling optimization of IES in steel industry. To represent the cooperative structure as well as extract the expert knowledge, a CFCM model is established the practical collaborative relationship among the gas, heat and electricity subsystems. Then, a Probability-based GrC (PGrC) approach is accordingly proposed for a reasonable initial scheduling scheme. To further determine the collaborative parameters in the CFCM model, a Multi-Objective Differential Evolution (MODE)-based method is proposed with the considerations of the practical constraints. The experiments studies on the proposed GEHC show its superiority compared with other commonly deployed solutions or frameworks in terms of cooperativity, convergency and diversity. The main contributions of this study can be summarized as follows.

The CGrC model provides the foundation of cooperative optimization, where the horizontal structure is deployed to formulate the relationship among the same units in different energy subsystem, while the vertical one for the different units in the same subsystem. The probability-based GrC approach that serves as a fast and convenient representation model approach is proposed to efficiently learn the cooperative mechanism in the IES from industrial big data.

The proposed GEHC framework takes the advantage of GrC for knowledge extraction and processing as the initial optimization, and MODE for parameter determination with the consideration of practical constraints. Compared with other models and optimization methods, the proposed GEHC exhibits superiority on both convergency and diversity, in terms of several representative performance indicators.

This paper is organized as follows. In Section 2, the aimed scheduling problem of IES in steel industry and its major challenges are described. Then, the proposed GEHC method are detailed in Section 3. In Section 4, experimental studies are conducted on the real-world applications to show the performance of the proposed method. Finally, Section 5 draws the conclusions.

Section snippets

Problem description

An illustrative structural chart of a typical IES in steel industry is given in Fig. 1. Energy media mainly involve gas, heat and electricity, which are depicted as green, blue and red dotted lines, respectively. To clarify the characteristics of each energy subsystem as well as the inherent collaborations among them, the detailed configurations are separately described as follows, with all the denotations of the variables given in the Nomenclature.

The Proposed Granular-Evolutionary Hybrid Computing Framework for Cooperative Scheduling Optimization

The proposed GEHC framework for solving cooperative scheduling problem of the IES in steel industry mainly involves two stages: problem modeling that also processes the expert knowledge for initial optimization, and parameter determination that also considers practical constraints for further optimization. This framework deploys both the real data and physics-based models given in Section 2, exhibiting as a data-mechanism hybrid driven method.

Experimental studies

In this section, real-world data of a steel plant in China coming from February 1st, 2021 to March 1st, 2021 are employed to verify the effectiveness of the proposed framework. The sampling interval of the data is 1 min. The developing environment and coding language are listed as follows. The detailed configurations of the IES in this steel plant are listed in Table 1.

Conclusions

With the purpose of scheduling optimization for the IES in steel industry, a GEHC framework has been proposed in this study, which utilizes collaborative structure for modeling the cooperative mechanism among the multiple energy media. Based on the designed MODE method, the collaborative coefficients who actually control the cooperativity are optimized under the satisfaction of the practical constraints. The completed experiments demonstrated that the proposed framework perform better in terms

Declaration of Competing Interest

NONE

Acknowledgements

This work is supported by the National Key R&D Program of China (No. 2018AAA0101702), and the National Natural Science Foundation (No. 61833003, No. 62125302 and No. U1908218).

References (27)

Cited by (7)

  • Sustainability of the Steel Industry: A Systematic Review

    2023, Biointerface Research in Applied Chemistry
View all citing articles on Scopus
View full text