Research on Modeling of Heating Substations Based on LSTM

Abstract

Establish an accurate model of heating substations is the basis for optimal control of heat supply in heating substations. However, it is very difficult to achieve an accurate model by traditional manual control methods as well as neural networks. In this paper, the heat supply problem of heating substations is formulated as a time series-based problem, a new heating substations model is established, which using Long-Short Term Memory (LSTM) neural network by collecting actual meteorological data and heat supply history data of a city heating substations. The experiments show that the model based on LSTM neural network can accurately predict the heat supply at the next moment, and the LSTM neural network model has higher accuracy and stability compared with the traditional neural network models Gradient Boost Regression Tree (GBRT), Random Forest Regressor (RFR) and Feed forward neural networks (FFNN).