基于小波变换与优化BP神经网络的超短期光伏发电功率预测
Ultra-Short-Term Photovoltaic Power Prediction Based on Wavelet Transform and Optimal BP Neural Networks
夏晓荣 1胡鹏飞 1王飞 1张明晨 2赵洁 2王波2
作者信息
- 1. 国网湖北省电力有限公司荆门供电公司,湖北荆门 448000
- 2. 交直流智能配电网湖北省工程中心,武汉大学电气与自动化学院,湖北武汉 430072
- 折叠
摘要
光伏发电功率的精确预测可以帮助电网实现更精细的管理,提高能源利用率;但光伏发电功率受到多种环境因素的影响,且具有较大的随机波动性,故挖掘光伏发电的效率特性非常困难.该文提出一种新方法,通过使用小波变换和优化BP神经网络来预测超短期光伏发电功率.该方法基于皮尔逊系数,可以获得与气象因素相关的预测结果;基于离散小波变换(discrete wavelet transform,DWT),将原始功率一阶差分序列分解为若干个不同频段的分量,提取光伏出力波动的频域特性;利用K-means聚类方法对功率一阶差分值进行聚类,并建立相应的神经网络预测模型,通过重组所得预测结果,得到初始预测功率差分值;利用气象因素通过GA-ACO-BP神经网络修正预测所得功率差分值,得到最终预测功率序列.利用某光伏电站所记录的实际功率数据进行验证,结果表明:DWT-GA-ACO-BP预测模型能提供较为精确的预测结果.
Abstract
Accurate PV power forecasting can assist the grid in achieving more leaned management and enhance energy utilization efficiency.However,the PV power output is affected by many environmental factors and exhibits significant stochastic volatility,making it challenging to mine the efficiency characteristics of PV generation.In this paper,we propose an ultra-short-term PV power prediction method based on wavelet transform and optimized BP neural network.Based on Pearson coefficients,the proposed method can obtain the highly correlated meteorological factors;and using discrete wavelet transform(DWT)it can decompose the original power first-order difference series to extract the frequency domain characteristics of the PV output;and the K-means clustering method is used to cluster the first-order power difference values and to build a neural network prediction model,and the initial predicted power differential value is predicted through recombining the predicted results.The meteorological factors are employed to correct the predicted power difference values through GA-ACO-BP neural network,acquiring the the final predicted power sequence.The actual power data recorded by a certain photovoltaic power station is utilized for verification,and the results indicate that the DWT-GA-ACO-BP prediction model can provide more accurate prediction results.
关键词
光伏出力预测/小波变换/优化BP神经网络/K-means/功率差分序列/超短期预测Key words
PV output forecast/discrete wavelet transform/optimized BP neural network/K-means/power differential sequences/ultra-short term forecast引用本文复制引用
出版年
2024
NETL
NSTL
万方数据