Energy Price Prediction Considering Generation Bids Variation: A Two-Stage Convolutional Long Short-Term Memory Approach

Abstract

Locational Marginal Price (LMP) is composed of energy, congestion and loss price components. All these price components are comprehensively determined by locational demands and locational generation bids. Due to difficulties of accessing updated generation bid information, previous price predictions from market participants' perspective focused only on learning the spatio-temporal correlations among historical price and load data without using generation bid information. In this paper, a two-stage convolutional long short-term memory (CLSTM) approach is proposed to incorporate historical generation bids into energy price prediction from market participants' perspective. Historical generation bids are organized into a 3-dimensional (3D) tensor, and taken as output of first stage and input of second stage in the training process. The implicit correlation among locational bids, demands and energy price is learned to improve price forecasting accuracy. Verification of the proposed approach is performed on the IEEE 30-bus system with publicly available historical market data from ISO-New England (ISONE). Comparisons between the proposed approach and other state-of-art prediction approaches are conducted to demonstrate the improvement of the two-stage CLSTM approach.