Research on Heat Load Prediction Model of Secondary Network Based on Improved Neural Network

In this note, a new ICEEMDAN-KPCA-LSTM heat load prediction model is proposed to address the current problem that heat supply enterprises adopt the regulation of crude heat supply, which leads to the mismatch between the heat supply and the heat required by users. The model utilizes ICEEMDAN to decompose the sequence of environmental factors and reduce its non-stationarity; KPCA is used to extract the main influencing variables of the feature sequence and reduce the input dimension. Meanwhile, in order to accurately forecast the heating load of the heating system, the LSTM network is used for dynamic modeling of the multiple characteristic sequences. Experiments indicate that this model has better forecasting precision than conventional methods, and it is possible to optimize system adjustment strategy and energy saving operation. In this paper, a new heat load forecasting model is put forward, which can be used as a reference for the management and operation of thermal power plants.