改进人工免疫网络算法在电力短期负荷预测中的应用

Application and research of improved artificial immune network to power short-term load forecasting

针对现有的短期负荷预测方法易陷入局部极值以及预测精度不高等缺陷,文中提出了一种基于改进免疫算法优化BP神经网络的短期智能负荷预测方法。通过利用改进的矢量距优化免疫网络,从而达到优化网络的目的。融入免疫调节原理,引入抗体浓度的概率选择式,采用自适应变化策略重新设计变异算子,利用新的变异尺度设计种群抗体,采用新的神经元适应度函数,并结合免疫网络调节的进化算法进行网络学习。实例分析表明,基于改进免疫网络优化的BP网络短期负荷预测算法比混沌算法优化BP网络算法精度更高,更具实用性。

According to the deficiencies of load forecasting model at present,a short-term load forecasting model based on optimized clone immune and BP neural network（BPNN） is presented.In the design of artificial immune network（AIN）,the principle of immune network regulation is used in a creative way and the method of immune programming is used to evolve the network structure.The probability of selective antibody concentration,a new fitness function of neurons,a new mutation operator and a new self-adaptive chaos mutation operator are adopted in the AIN.The excitation function controls the BP algorithm which greatly accelerates convergence of BP training,the self adaptable strategy based on clone immune optimizes the controlled BP algorithm,and it improves its global searching ability better than the BP algorithm optimized by chaos and avoids the algorithm to be trapped in local minimum and improves the prediction accuracy.