Cooling,Heating And Electrical Short-term Load Forecasting In Parks Using Distributed Energy

The accurate prediction of the users’ cooling,heating and electrical load is the key technology to make the park’s integrated energy system give full play to its advantages.For parks with distributed energy sources which includes renewable energy,the prediction of cooling,heating and electrical load is particularly important for increasing the utilization proportion of renewable energy and for enhancing the stability of the operation for energy supply system.The commonly used load characteristic indexes is introduced.These indexes can also be applied to analyzing cooling and heating load.Curve method and index method is used to analyze the daily,monthly,annual and weekend load characteristics of the three kinds of parks and draw some conclusions according the their similarities and differences.The correlation and sensitivity analysis are used to summarize the main factors affecting load characteristics.Cooling heating and electrical short-term load forecasting based on improved entropy method is introduced.The improved entropy method is used in the feature point weight and the similar day weight determination separately and this method is suitable for load forecasting in the days without sudden weather changes.A case study show that the proposed method has higher accuracy and efficiency.Cooling heating and electrical short-term load forecasting based on empirical wavelet transform is introduced.The load sequence is resolved into three components with different characteristics through the process of decomposition and identification.And then linear regression model,generalized regression neural network and auto-regressive and moving average is selected to predict the components separately.The load forecasting value is reconstructed by using generalized regression neural network again.This method is suitable for load forecasting in the days with sudden weather changes.A combined forecasting model which takes into account both the computational speed and the load forecasting accuracy is established combining the two proposed methods.