摘要
负荷是电力系统运行和规划的依据,精确的预测可提高系统运行的可靠性和经济性。作者将电力系统负荷预测的三种主要方法,即外推法、灰色预测法和人工神经网络法结合起来,建立了一种线性组合预测模型。在组合模型的权重系数求解中,首先对目标函数和等式约束使用拉格朗日乘子法来求解权重系数。当出现小于零的系数时,改为只使用误差矩阵的对角元素来计算,这种近似对预测精度影响较小,但简化了计算,且保证了组合系数大于零的条件。由于组合模型的总平均误差要小于各单一预测方法的平均误差,这就提高了预测精度,尤其组合模型的最大预测误差要小于单一模型的最大预测误差,从而降低了预测的风险性,实例证明这种组合模型具有较好的实用性。
Load is the foundation of the power system operation and planning. The precise load forecasting can improve the reliability and economy of the power system operation. Combining the three existing methods, i.e., extrapolation, grey forecast and artificial neutral network, a linear combination model for load forecasting is established in this paper. To calculate the weight coefficients of the combination model firstly the Lagrangian multiplier method is applied to object function and equality constrains. When the negative coefficients appear, only the diagonal elements in the error matrix are applied to the calculation. Such approximation may bring less influence to forecasting accuracy, but the calculation is simplified and the condition of positive weight coefficients can be satisfied. Because the total average error of combination model is less than the average error of unitary forecasting method, the forecasting accuracy is improved, especially the maximum forecasting error is less than that of unitary model. Therefore, the risk of forecasting is reduced. The results of actual examples show that combination model possesses better practicability.
出处
《电网技术》
EI
CSCD
北大核心
2002年第10期10-13,共4页
Power System Technology
