Stochastic optimization approaches in solving energy scheduling problems under uncertainty

This paper focuses on evaluating the performances of various approaches to solve a day-ahead energy scheduling Mixed Integer Linear Programming (MILP) cost minimization problem, with hybrid energy generation system consisting of solar photo voltaic (PV), waste to energy (WTE) and main grid. The system under consideration is a highly energy-intensive industrial facility that can be generalized to any industrial process regardless of the domain. Out of the three case studies considered in this paper, the first case study deals with adopting unsupervised machine learning approach to generate a reduced number of probabilistic scenarios of the uncertain parameters and implements the Here & Now and Wait & See algorithms to solve the resulting stochastic optimization problem. While the second case study directly collects and assumes a finite set of historical data of the uncertain parameters as scenarios, the third case study uses an evolutionary algorithm called Limited Evaluation Evolutionary Algorithm (LEEA) to solve the energy scheduling problem with associated constraints. Various performance metrics such as mean absolute error (MAE) and root mean square error (RMSE) are utilized to compare the performances of the various approaches presented through case studies. The values of these evaluation metrics showcased the enhanced performance of evolutionary algorithm based approach compared to the other two approaches. Copyright (C) 2022 The Authors.