The Application Of Combination Forecasting Model In Short-Term Load Forecasting Of Huzhou Grid

The short-term power load forecasting is an important work for power dispatch department. Along with economic development and the continuous improvement of the electricity market, power system load forecasting work directly affect the economic and social benefits, its accuracy is becoming more and more high. The short-term power load is affected by many uncertain factors such as the weather, the change rule is complicated, it is difficult to forecast. With load forecasting accuracy is higher and higher, the single model load forecasting method has been stretched. According to the characteristics of he Huzhou power grid load, which Changes periodicity regularity and the total load is small, the load forecasting accuracy is easily influenced by weather,this paper proposes a divided-period variable weights combination forecasting model,which based on the normalized value of difference between load peak and valley multiple proportion smoothing method and the BP neural network model.The quality of the historical load data is significant to summarizing and concluding the load variation. But the abnormal data and noise is unavoidable in the historical load data, this paper carried out the pretreatment on the historical load data of Huzhou power grid, the abnormal data in the historical load data are classified, then proposes recognition method and according to different types of the abnormal data, puts forward the corresponding correction method. It gives a smooth load curve of the sawtooth wave with Denoising method of wavelet threshold.This paper discusses normalized value of difference between load peak and valley multiple proportion smoothing method base on the load similar day and BP neural network, then separately forecasts and werify the results. Finally the results of the single model are divided-period variable weight combined to establish a combined forecasting model, and the forecast results are analyzed and compared,and finds out a improved combination daily forecasting model of the Huzhou power grid up to the requirement by using joint parameters adaptive optimization.