准确地实现小时负荷预测是实施优化控制和动态安全分析的前提.采用嵌入维最小二乘支持向量机(ELS-SVM)的方法进行建模和预测,对影响负荷的因素进行模糊化处理.采用了粒子群(PSO)优化算法解决ELS-SVM学习过程中多参数难以调整的问题.提出分段小批量学习和更新的在线学习方法,既降低了运算量又能有效地避免积累误差,从而提高预测精度.实验结果表明,该方法有效地将预测精度从2.1%提高到了1.29%.%Hours load prediction is the fundament of doing an analysis of optimization control and dynamic security.In this paper,we use embedding dimension least square support vector machines (ELS-SVM) to make model and forecast.Fuzzy processing of the factors that affect short term load is conducted.And present PSO optimization algorithm is adopted to solve the problem that the parameters are difficult to adjust in the ELS-SVM learning.An approach of segment small quantity learning and updating is proposed to realize online forecasting which not only reduces the computing amount, but also avoids the error of accumulation, thus improves the forecasting precision.The case study shows that the prediction accuracy is improved from 2.1% to 1.29%.
展开▼