小波包分析与BP神经网络预测相结合的在线能量管理策略

The Online Energy Management Strategy Combining Wavelet Packet Analysis and BP Neural Network Prediction

针对现有混合储能分配策略无法适应不同储能设备特性以及光伏发电功率信号随机波动特征的问题,提出了将小波包分析,BP神经网络在线预测及模糊控制相结合的分配策略。首先对已有的混储总功率进行小波包分频,得到适合于氢储系统的低频功率,再通过此数据集进行BP神经网络的离线训练,并将训练的权重值用于在线神经网络预测。其次,将在线神经网络训练得到的氢储功率结合超级电容的荷电状态通过模糊控制器得到超级电容功率的修正值,并对氢储设备的功率、在线神经网络的权重进行修正,使得在线神经网络适应实时的混储功率波动。最后根据另一混储功率数据10 s采样数据在MATLAB/Simulink平台进行仿真。与基于规则的混储分配策略和低通滤波的分配策略相比较,结果表明此算法可以很好地改善氢储设备充放电功率,适应实时信号的随机波动,使超级电容的荷电状态运行在合理的区间范围内。

To address the issue that the existing hybrid energy storage allocation strategy cannot adapt to the characteristics of different energy storage devices and the random fluctuation characteristics of PV power signal,this paper proposes an allocation strategy combining wavelet packet analysis,BP neural network online prediction and fuzzy control.Firstly,wavelet packet frequency division is performed on the existing total power of hybrid storage to obtain the low frequency power suitable for hydrogen storage system,and then the BP neural network is trained offline by this data set,and the trained weight values are used in the online neural network.Secondly,the hydrogen storage power obtained from the online neural network training is combined with the charge state of supercapacitor to obtain the correction value of super-capacitor power by fuzzy controller,and the power of hydrogen storage device and the weights of online neural network are corrected to make the online neural network adapt to the real-time mixed storage power fluctuation.Finally,the simulation is performed in MATLAB/Simulink platform based on another 10 s sampling data of mixed storage power data.And compared with the rulebased mixed storage allocation strategy and low-pass filtering allocation strategy,the results show that this algorithm can well improve the hydrogen storage device charging and discharging power,adapt to the random fluctuation of real-time signal,and make the super-capacitor.