基于Kalman滤波和BP神经网络的光伏超短期功率预测模型  

Very Short-Term PV Power Forecasting Model Based on Kalman Filter Algorithm And BP Neutral Network

作　　者：王雨[1] 苏适 严玉廷 

机构地区：[1]云南电网公司研究生工作站,昆明650217 [2]云南电力试验研究院(集团)有限公司电力研究院,昆明650217

出　　处：《电气技术》2014年第1期42-46,共5页Electrical Engineering

摘　　要：随着光伏电站规模不断扩大,提高光伏发电功率预测精度,将对电网的稳定运行有很大帮助。然而,光伏发电最重要的影响因子辐照度受云量影响很大,随机性很强,特别当多云的时候,变化更是很快速、剧烈,这给光伏超短期功率预测带来困难。为此,本文提出一种基于Kalman滤波和反传播(back propagation,BP)神经网络的光伏超短期功率预测模型,采用地外辐射和Kalman滤波估计辐照度,温度和湿度预测值则通过持续预测法获得,再将这三者作为神经网络的输入来预测未来15min的光伏发电功率。最后,采用连续三日的实际数据验证了本文提出模型的可行性。With the expanding of the capacity of photovoltaic power plants, uncertainty and volatility of the photovoltaic power generation will generate a negative impact on the safe operation of Power Grid. Therefore, To improve the PV power power forecasting accuracy and use the forecasting results will provide certain ancillary basis for Power Grid to operate and control PV power plants, which helps a lot for the grid stable operation. However, to be the most important impact factor of PV generation, irradiance is greatly influenced by cloud cover and changes randomly, especially when cloudy, changes rapidly and severely, which brings difficulty to the very short-term PV power forecasting. For this reason, this paper proposes a very short-term PV power forecasting model based on kalman filter algorithm and back propagation(BP) neutral network. This model applies extraterrestrial radiation and kalman filter algorithm to estimate irradiance, uses continuous prediction method to predict temperature and humidity, then input these three to BP neutral network to forecast PV power for the next 15 minutes. At last, apply actual historical data of three forecast days to test and verify effectiveness and feasibility of the proposed model.

关 键 词：KALMAN滤波 BP神经网络 地外辐射 光伏超短期预测 

分 类 号：TM615.2[电气工程—电力系统及自动化]