基于周期性截断灰色系统的电力负荷预测

Power Load Forecasting Based on Periodic Truncated Grey System

摘要:

电力负荷预测是电力系统调度和电力生产计划制定的重要依据;电力负荷时间序列有着明显的周期性特征;传统的电力负荷预测主要侧重于预测方法的研究,而忽略了电力负荷数据周期性特性的分析,影响了预测的准确性;针对电力负荷时间序列的周期性特征,提出了一种基于周期性截断的灰色系统模型来进行电力负荷预测;该模型利用周期性截断来反映负荷数据的周期性特征,提高了预测的精度;仿真采用EUNITE Network的公开负荷数据进行算法性能的测试,并与一些主流的电力负荷预测算法:BP神经网络、极限学习机、自回归模型以及传统的灰色系统模型做比较;仿真结果表明,周期性截断的灰色系统负荷预测的归一化均方误差和绝对平均误差是最小的;周期性截断的灰色系统为电力系统负荷预测提供了一种新的有效方法。

Abstract:

Power load forecasting is an important basis for power system scheduling and power production planning. The power load time series has obvious periodicity characteristics. Traditional power load forecasting mainly focuses on forecasting methods, but ignores the analysis of periodic characteristics of power load data, which affects the accuracy of prediction. According to the periodic characteristics of power load time series, a grey system model based on periodic truncation is proposed to predict the power load. The model uses periodic truncation to reflect the periodic characteristics of load data and improves the prediction accuracy. Simulation uses EUNITE Network public load data to evaluate the performance of the algorithm, and compare with some mainstream power load forecasting algorithms:BP neural network, extreme learning machine, auto regression model and traditional grey system model. The simulation results show that the normalized mean square error and absolute mean error are minimum for the proposed method. The periodic truncated grey system provides a new effective method for power system load forecasting.