应用于月度用电量预测的小波分析法

A Wavelet based Method for Monthly Electricity Demand Forecasting

月度用电量预测是中期负荷预测的主要内容,也是制定月度发电规划的基础。文中以美国亚利桑那州为例,采用小波分析法,首先使用小波变换获得若干个采样点减少一半的小波系数;然后分别对各系数插零、重构,恢复到原数据的长度;最后采用RBF神经网络对恢复长度的各系数进行预测。该方法将月度用电量的时间序列分解成趋势项和波动项,分别进行预测,提高了预测精度。

Monthly electricity demand forecasting is the main content of medium-term load forecasting and the base of monthly scheduling for the power grid operation. This paper proposes a wavelet based monthly electricity demand forecasting method. It consists of three steps： first, the wavelet transform converts the original data series into several coefficients which are down sampled; second, inserting zeroes between each coefficient to recover the coefficients to the origin length; finally, RBF neural network is adopted to forecast the coefficients. By decomposing the monthly electricity consumption data series into trend component and fluctuation component and forecasting each component separately, the proposed method greatly improves the precision of monthly electricity demand forecasting.