Forecasting Multivariate Loads of Integrated Energy Systems Based on LSTM and Weather Variables

Multivariate loads short-term forecasting is a crucial prerequisite for the safe and stable operation and economic optimization management of Integrated Energy Systems (IES). Existing research has shown limited consideration for the interdependencies between electricity, cold and heat loads within the system, as well as the correlations between weather factors and load profiles. This paper employs the Spearman rank correlation coefficient to quantitatively analyze the relationships among weather data and the electricity, cold and heat loads, as well as the correlations among these different load types. The analysis results reveal the presence of strong relationships among various influencing factors and load categories. Building on these findings, this study introduces a comprehensive energy multivariate loads short-term forecasting model based on LSTM networks and compares its predictive accuracy with ANN and GRU models. Case study results indicate that the proposed model provides more precise load predictions, thus offering valuable insights for multivariate loads forecasting within IES.