Temperature prediction of battery energy storage plant based on EGA-BiLSTM

Abstract

Battery energy storage plants (BESPs) are more and more important in the future power systems. The industry desires a credible temperature prediction method to deliver a safe temperature range of the BESPs. This paper develops a bespoke methodology that combines the elitist preservation genetic algorithm (EGA) and bidirectional long-short term memory network (BiLSTM) to deliver accurate battery temperature predictions. First, this paper applies the EGA to obtain the optimal segmentation strategy of time-series data. Second, the BiLSTM is used to predict both the highest and the lowest temperature of the battery pack within the energy storage power plant. In this step, an improved loss function is proposed to improve the prediction accuracy of the BiLSTM. This paper applies the real operation data, from January to February 2020, of one certain BESP to justify the developed an EGA-BiLSTM method. Case studies reveal that compared to the LSTM and the LightGBM, the developed method significantly reduces the prediction error by 12% and 26%, respectively. Lastly, we also conduct ablation experiments to prove that the EGA-BiLSTM method also accommodates short-term scenarios, where the R2_score of short-term temperature prediction for BESP could achieve up to 0.86. All these can provide an effective method of temperature prediction, ensuring safe operation of the BESPs.

Keywords

• Battery energy storage plants
• Temperature prediction
• Elitist preservation genetic algorithm
• Bidirectional long-short term memory network