基于改进遗传算法的电力网络负载预警模型

The Electric Power Network Load Early Warning Model Based on Improved Genetic Algorithm

在对电力网络负载预警时,电力负荷参数变化过程存在大变化、非线性的特点,预警过程与时间相关性较大,传统算法进行电力网络负载预警过程中,没有负荷变化的模糊性与时间序列相关性,使得网络学习收敛速度慢,易陷入局部极小值,造成了电力网络负载预警精确度较低的问题。为提高预警精度,提出改进遗传算法的电力网络负载预警方法。将遗传算法与BP神经网络算法相融合,组建电力系统的重构相空间模型,通过计算最大Lyapunov指数来找出负载时间序列,具有混沌特性,利用混沌神经网络对负载时间序列进行短期负荷预测,结合灵敏度分析对基本遗传算法的编码、初始种群、适应度函数和交叉、变异策略等进行改进,有效的建立了电力网络负载预警模型。仿真结果表明,该电力网络负载预警模型精确度高,实用性强。

In order to improve the accuracy of forewarning, a power network load forewarning method was pro- posed based on genetic algorithm. The genetic algorithm and BP neural network algorithm were integrated, to estab- lish the reconstructed phase space model of power system. The largest Lyapunov exponent was calculated to find load time series, which possess the characteristics of chaos. Chaotic neural network was used to make short term load fore- casting for load time series. Combined with the sensitivity analysis, we improved the coding of basic genetic algo- rithm, initial population, the fitness function, the crossover and the mutation strategy and so on, which can effective- ly establish power network load forewarning model. The simulation results show that the power network load forewar- ning model has high accuracy and practicability.