基于蜂群算法改进的BP神经网络风电功率预测

Improved BP neural network based on Artificial Bee Colony algorithm for wind power prediction

由于风能具有随机性和间歇性的特点,造成了其功率输出的不稳定,而大规模的风电接入给电力系统的正常稳定运行和调度带来影响。详细分析影响风电场输出的因素,确定风速、风向正弦和余弦为影响风电输出最主要的关联因素,采用统计预测方法将历史实际输出功率、风速、风向正弦和余弦作为BP神经网络的输入矢量,并采用人工蜂群算法优化得到神经网络的权值和阈值,构建ABC-BP神经网络风电功率预测模型。通过对某实测风电功率进行预测验证,结果表明:基于蜂群算法改进的BP神经网络风电功率预测,可以克服BP神经网络易于陷入局部极小的缺陷和不足,极大地提高了全局搜索能力以及预测的稳定性和精度;同时,将自适应的选择策略引入到蜂群算法优化适应度的选择中,减少了网络层参数的训练时间,提高了收敛速度。

Due to the randomness and intermittence of wind,the output of wind power is instability.The large-scale access of wind power has an impact on the operation and scheduling of power system.Firstly,the influence factors of output for a wind farm is analyzed in detail.The most important factors including the wind speed,wind direction,sine and cosine are also determined.According to the statistical method,the history of actual output power,wind speed and direction of sine and cosine are selected as input vectors of BP neural network.The weights and threshold of neural network are optimized by the artificial bee colony.On this basis,the ABC-BP neural network model for wind power prediction is then constructed.A practical case is selected for verification of prediction.It is shown that the BP neural network improved by Bee colony algorithm can overcome the defects and shortcomings of traditional BP neural network by which the calculation is easy to fall into a local minima in wind power prediction.The global search ability,prediction accuracy and stability are greatly improved.In addition,the adaptive selection strategy is introduced into the fitness selection of bee colony algorithm optimization,which reduces the training time of network layer parameters and improves the convergence speed.