A Multi-Objective Generation Expansion Planning with Modeling Load Demand Uncertainty by a Deep Learning- Based Approach

Abstract

In recent years, with population rise and electrification of the transportation fleet, the need for electricity demand has grown, which increased the importance of Generation expansion planning (GEP). Most of the literature investigated GEP by considering one objective function (minimizing the cost), whereas other objectives also have a high priority. For this reason, in this paper, a multi-objective GEP with aims of minimizing cost, minimizing emission, maximizing reliability, and flexibility is presented. GEP studies' foundation is based on the amount of annual peak load demand, which shows the superiority of considering load uncertainty in GEP studies. We used a deep learning method based on long short-term memory (LSTM) networks, which have a high ability in the time series forecasting for modeling load uncertainty. The optimization problem is also considered as a mixed-integer linear programming (MILP) that guarantees the optimal global solution. The forecasted peak load for the year 2020 as a test day shows the deep LSTM network's robustness for annual peak load forecasting (5.23% error with real data).