基于QPSO参数优化的WLS-SVM短期负荷预测

Short-term Load Forecasting Based on WLS-SVM Method with Parameter Optimization by QPSO

为了解决负荷非线性特性导致的预测模型难以准确建立的问题,提出一种基于量子粒子群优化(QPSO)参数选择的加权最小二乘支持向量机(WLS-SVM)的短期负荷预测模型和方法。首先,利用量子粒子群优化方法来对模型进行训练,从而选出最优超参数。其次,采用具有良好泛化性能的WLS-SVM回归模型弥补损失的鲁棒性。文中以黑龙江电网短期负荷预测为例,将该方法与一般LS-SVM模型的预测结果进行了对比分析,结果表明此方法能明显提高预测精度。

In order to solve the problem that load forecasting model is difficult to accurately establish because of the nonlinear characteristics of load,the weighted least squares support vector machine（WLS-SVM） method based on quantum-behaved particle swarm optimization（QPSO） is introduced to build short-term load forecasting models.Firstly,the quantum particle swarm optimization method is used to train the model to obtain the optimal hyper parameters.Then the WLS-SVM regression model with good generalization performance is used to strengthen robustness.In addition,the forecasting results by this method are compared with that by general LS-SVM model,which show that this method can significantly improve the prediction accuracy.In the end,the short-term load forecasting of Heilongjiang power grid is simulated by using of this method and the results verify its validity.