一种基于最小二乘支持向量机的年电力需求预测方法

针对电力系统年用电量增长的特点,将最小二乘支持向量机LS-SVM(least squares support vector m a-ch ine)回归模型引入年电力需求预测领域,并给出了相应的过程和算法。与常规基于人工神经网络ANN(ar-tific ial neural networks)的智能预测方法比较,该模型优点是明显的:1)将神经网络迭代学习问题转化为直接求解多元线性方程;2)整个训练过程中有且仅有一个全局极值点,确定了预测的稳定性;3)将年电力需求预测的外插回归问题转换为内插问题,提高了预测精度。应用实例表明:该模型实现容易、预测精度高,更适合年电力需求预测。

Annual electric consumption forecasting model based on least square support vector machines

According to the speciality of electric power consumption development,an improved regression model based on the least squares support vector machine(LS-SVM) is introduced into the field of electric power demand forecasting.The design steps and learning algorithm are also addressed. Compared with forecasting methods based on artificial neural networks(ANN),there are some advantages of the proposed method.First,the LS-SVM solution follows directly from solving a set of linear equations instead of an iterative problem.Second,the machine learning in the standard ANN approach are replaced by linear equations in LS-SVM,so the global optimal solution can be uniquely obtained.Third,the forecasting precision can be enhanced due to the model converts extrapolation regression into interpolation regression.The application case proved that the proposed method is of easy realization,accurate prediction and fits for annual electric power demand forecasting.