Study on Application of Wavelet Neural Network (WNN) Model in City Gas Load Forecasting

To overcome the disadvantages that the neural network forecasting model is prone to local suboptimum, plus the slow training speed in the prediction of nonlinear sequences, the wavelet has, based on the traditional neutral network forecasting model, been introduced as the transfer function of hidden layers, constituting a new forecasting model, which is applied to the forecasting of city gas loads. Based on the historical data of city gas loads in a given area, the fluctuations of weather conditions, dates and other uncertainties are synthesized to train and test models, thereby forecasting the gas load. The results prove that the wavelet neutral network (WNN) forecasting model boasts accurate prediction to the variation trend of city gas loads, characterized by simple structure and quick convergence speed, with high practical application value.