Abstract:In order to predict the heat load of heating system, the weather, user''s indoor temperature and time lag are taken as the data basis. In this paper, a new method is provided to solve the problem of the lack of thermal control theory in thermal power industry by using statistical theory and big data structure and by studying and revising a lot of data samples. The core of the Algorithm is pattern recognition algorithm and correlation analysis of time series, which makes it possible to solve the influence of weather and time delay on user heating. In this paper, we take n typical thermal power stations and their tropical users as the experimental objects, and take the local weather forecast and real-time weather data as the basis, the samples were trained by big data pattern recognition machine learning algorithm based on heating station heating data, weather forecast data and typical heating user''s indoor temperature, a set of forecasting model for Dynamic Energy Consumption Index of thermal power station is formed.