基于卡尔曼滤波-SVR时刻峰值的短期负荷预测

Short-term Peak Load Forecasting of Every Time Forecasting Based on Kalman Filtering-SVR

以提高短期负荷预测中的时刻峰值精度为目标。为了给电力调度部门提供各时刻负荷分配的极限值,以一天中24个时刻的负荷峰值代替一天96点的负荷作为研究对象,并且在支持向量机的回归拟合(SVR)基本算法基础上,提出了一种SVR预测值经过累积式自回归—移动平均模型(ARIMA)的卡尔曼滤波调整的短期负荷预测模型。该模型利用ARIMA模型建立卡尔曼滤波方程;并将SVR预测值作为观测值,通过卡尔曼滤波的递推方程组,求得最终的负荷预测值,从而实现卡尔曼滤波-SVR预测。经过实例验证该模型可以有效提高短期负荷预测的精度。

Its goal is to improve accuracy in short term peak load of every time forecasting. In order to provide the load distribution limit value to power dispatch department,it uses peak load of 24 times instead of 96 points load as the research object. And it is based on SVR to propose a model which predicted value of SVR is adjusted by Kalman filtering of ARIMA. The model uses ARIMA to built Kalman filtering equation and uses the predicted value of SVR as observation values of Kalman filtering,so the value can be adjusted by recurrence equation of Kalman filtering. This is the KalmanSVR model. The model applier to a typical example has been proved to have improved accuracy in the load forecasting.