Active Operation Strategy of Hybrid Energy Storage in Regional Integrated Energy Systems Based on Long Short-Term Memory Network

The inherent heat and power coupling characteristics of cogeneration units in a regional integrated energy system (RIES) limit their flexibility and operational efficiency. However, the charging and discharging capabilities of energy storage systems depend on their energy states. Conventional "electricity-dictated heating" and "heating-dictated electricity" strategies fail to comprehensively consider the fluctuating characteristics of multi-load demand and renewable energy in the long term, making it difficult to fully leverage the flexibility of energy storage systems. This paper adopts long short-term memory (LSTM) neural network deep learning methods to prospectively predict the long-term trends of load and renewable energy changes. An optimization dispatch model is established considering factors such as carbon emission penalties, renewable energy curtailment penalties and other costs. A strategy is proposed to coordinate heat storage and power storage. Taking a specific RIES as an example, it compares with the traditional "electricity-dictated heating" and "heating-dictated electricity" strategies. The results show that the maximum prediction error of the LSTM neural network is 4.7%. Compared to the other two strategies, the hybrid storage active operation strategy can reduce operating costs by 11.12% and 3.67%, respectively. Additionally, it can smooth the heat-electricity load curve and thus reduce the RIES electricity procurement cost, improve cogeneration efficiency, and further reduce total operating costs.