An ADMM-LSTM Framework for Short-Term Load Forecasting

Abstract

Accurate short-term load forecasting (STLF) is of paramount importance for ensuring the reliable and efficient operation of power systems. With the continuous increase in volume and variety of energy data provided by renewables, electric vehicles and other sources, long short-term memory (LSTM) has emerged as an attractive approach for STLF due to their superiorities in extracting the dynamic temporal information. Unfortunately, the conventional LSTM training methods rely on stochastic gradient methods that succumb to many limitations. In this paper, we present an innovative LSTM optimization framework via alternating direction method of multipliers (ADMM) for STLF, dubbed ADMM-LSTM. Explicitly, we train the LSTM network distributedly by ADMM algorithm. More specifically, we introduce a novel approach to update the parameters in the ADMM-LSTM framework, using a backward-forward order, which leads to a significant reduction in computational time. Furthermore, the solution of each subproblem in the proposed framework utilizes proximal point algorithm (PPA) or local linear approximation (LLA). As a benefit, the convergence rate is improved by the distributed optimization framework of ADMM-LSTM. Additionally, due to the gradient-free feature, ADMM-LSTM is able to avoid the exploding/vanishing gradient as well. Finally, we demonstrate that our proposed method exhibits superior performance compared to the benchmark methods in terms of both root mean square error (RMSE) and mean absolute percentage error (MSE).

Keywords: Alternating Direction Method of Multipliers, Long Short-term Memory Network, Short-term Load Forecasting, Gradient-free Feature