摘要
为了提高具有随机性和复杂性的电力负荷预测精度,提出了一种基于EMD与GA-BP神经网络的短期负荷预测。该方法利用EMD的优点,将原始电力负荷序列分解为若干个IMF分量和余项。针对BP神经网络训练时间长,且容易陷入局部最小的缺点,利用遗传算法优化了BP神经网络,替代了传统的BP算法。最后通过分析各个分量的自身特点,分别构建不同的BP神经网络模型,对各分量分别进行预测,相加各分量预测值,得到了最终预测结果。实例验证表明,与EMD-BP预测方法相比,该方法具有较高的负荷预测精度和较强的适应能力。
In order to improve the load forecasting accuracy of the randomness and complexity, this paper proposes the neural network of short-term load forecasting based on the EMD and GA-BP. This method will divide automatically original load sequence into several independent intrinsic mode functions by using EMD adaptability. In allusion to the BP neural network training for a long time and the shortcomings easy to fall into local minimum, the paper uses genetic algorithm to optimize BP neural network algorithm to replace the traditional BP algorithm. Then the different support vector machine model is constructed by analyzing the characteristics of each component. Finally, with separately forecasting each component and adding forecast values of the components, the final prediction forecast result is obtained. The experiment shows that this method has higher precision and strong ability to adapt than the EMD-BP model prediction.
出处
《电测与仪表》
北大核心
2013年第4期17-21,共5页
Electrical Measurement & Instrumentation
关键词
短期负荷预测
经验模式分解
BP神经网络
遗传算法
short-term load forecasting, empirical mode decomposition, BP neural network, genetic algorithm
