A prediction‐based optimization strategy to balance the use of diesel generator and emergency battery in the microgrid

• Published in: International journal of energy research Vol. 44; no. 7; pp. 5425 - 5440
• Main Authors: Liu, Zhonghua, Zhang, Huihui, Dong, Jichang, Yu, Hua
• Format: Journal Article
• Language: English
• Published: Chichester, UK John Wiley & Sons, Inc 10.06.2020
• Subjects: Comparative analysis; Diesel; Diesel fuels; Diesel generators; Distributed generation; Emergencies; Energy demand; Energy management; Fuel consumption; Integer programming; Integers; Linear programming; microgrid; mixed‐integer linear programming; multi‐objective mixed‐integer nonlinear programming; Nonlinear programming; Objective function; optimal operation; Optimization; Predictions; Regression analysis; Regularization; Scheduling; Strategy
• ISSN: 0363-907X; 1099-114X
• DOI: 10.1002/er.5292

Summary This paper presents a prediction‐based optimization strategy (POS) for the Energy Management System to balance the use of diesel generator (DG) and emergency battery (EB) in the microgrid. The POS is developed by combing two operating strategies, the “predictive analysis” and “optimal operation” in each scheduling period for the microgrid. Based on the predicted microgrid state and energy demand, a multi‐objective mixed‐integer nonlinear programming model (MOMINP) is constructed to minimize the fuel consumption and the regularization of battery charge/discharge subject to the practical constraints in the microgrid. This paper proposes a detailed scheme to deal with the multiple objectives and nonlinear constraints in the MOMINP, then the MOMINP is successfully converted into a mixed‐integer linear programming model (MILP). And an adjustment strategy is designed to obtain the near‐optimal solution of the MOMINP based on the optimal solution of the MILP solved by using the CPLEX Optimizer. Experimental results show that in a basic scheduling period, the working time of DG in the POS‐softmax regression strategy is shorter than the current operation, and the fuel consumption reduction ratio is about 15.3% with the same battery SoC value at the end of the scheduling. At the same time, the fuel consumption in the POS‐accurate prediction strategy can be reduced by up to 54.9% compared with the POS‐softmax regression strategy and can be reduced by 61.8% compared to the current operation. Based on the comparative analysis of the actual case data of a micro‐grid in 6 months, it can be seen that on average the POS works better than the current operation, with an approximately 23.6% decrease in the objective function and an additional 16.2% decrease with an accurate prediction.