基于改进的BP神经网络的光伏组件发电量预测模型设计

Design of Photovoltaic Module Generation Power Forecasting Model Based on Improved BP Neural Network

针对光伏并网发电系统发电量的不稳定性对电网的冲击和污染,分析了光伏发电的影响因素,建立了基于改进的BP神经网络的发电量预测模型。该模型采用Fletcher-Reeves共轭梯度算法,克服了传统BP神经网络算法收敛较慢的缺点,在保持训练过程稳定的前提下,具有更快的学习速率。结合光伏发电的历史数据和当天天气情况,对该模型进行训练、测试和评估,并应用于光伏系统发电量的预测中。结果表明,该模型具有较高的精度,提高了光伏并网发电的安全和稳定性。

Since the power generated by grid-connected photovoltaic power generation system is unstable, it has great impact on the power grid. This paper analyzes the influencing factor of photovoltaic power generation, and the back-prop- agation neural network is adopted to establish the power prediction model. Using the Fletcher-Reeves conjugate gradient algorithm, this model overcomes the defect of slow convergence rate for traditional BP neural network. On the premise of keeping stable for the training progress, this model has faster learning rate. Combination of historical photovoltaic power generation and weather data, this model is trained, evaluated and applied to predict the photovoltaic power generation. The results indicate that the proposed model has high precision and improves safety and stability of grid-connected photo- voltaic power generation.