基于模糊聚类与改进BP算法的日负荷特性曲线分类与短期负荷预测

摘要

提出了一种将模糊聚类技术与人工神经网络中的BP网络相结合的日负荷特性曲线分类与短期负荷预测的方法.通过模糊聚类技术将不同用户的负荷特性曲线进行分类,建立出不同的典型负荷曲线.然后利用同预测曲线相同类型的典型曲线,结合温度、日类型、湿度等对短期负荷预测影响较大的因素作为学习样本建立相应的BP网络模型.针对传统BP算法的不足,利用变学习速率和附加动量来改进BP算法并预测日负荷曲线.通过对实际日负荷曲线样本进行分类和对短期负荷进行预测证明该方法预测精度较高,在实际应用中具备可行性.%A classification of daily load characteristics curve method and forecasting of short-term load which combines fuzzy clustering with one of the artificial neural network named BP neural is put forward. Different typical load curves are created by means of fuzzy clustering technology which classifies the load characteristics curves of different customers. Then this paper takes use of the typical curve which is similar to the predicted curve, and associates with some factors which exert greater influence on short-term load forecasting, such as temperature, day type and humidity and so on to build relevant BP neural model. Aiming at the shortage of BP algorithm, the variable learning rate and the additional momentum are adopted to improve BP algorithm and predict daily load curve. The classification of practical daily load curve samples and forecasting of short-term load prove that the proposed method possesses higher forecasting precision, having the feasibility in practical application.