| 引用本文: | 师浩琪,郭力,刘一欣,王成山.基于多源气象预报总辐照度修正的光伏功率短期预测[J].电力自动化设备,2022,42(3): |
| SHI Haoqi,GUO Li,LIU Yixin,WANG Chengshan.Short-term forecasting of photovoltaic power based on total irradiance correction of multi-source meteorological forecast[J].Electric Power Automation Equipment,2022,42(3): |
|
| 摘要: |
| 针对单一气象预报源可能存在的误报和偏差问题，提出一种基于多源气象预报总辐照度修正的光伏功率短期预测方法。根据功率序列特征，采用自组织映射神经网络聚类算法实现历史数据广义天气类型划分。按照晴朗程度实现广义天气类型与公共气象服务天气类型预报的匹配对应，并计算不同广义天气类型总辐照度各等级之间的折算系数。在计算各广义天气类型系统误差的基础上，如果预测日数值天气预报广义天气类型分类结果与公共气象服务天气类型预报的一致，则叠加修正总辐照度系统误差；否则，采用树扩展朴素贝叶斯算法计算2种广义天气类型的转移概率，在修正系统误差后利用折算系数计算公共气象服务天气类型预报对应广义天气类型的总辐照度序列，并根据转移概率设定权重系数进一步修正总辐照度序列。建立预测模型，基于遗传算法优化的反向传播神经网络获得光伏功率短期预测结果。利用某光伏电站的实际运行数据和气象预报数据验证了模型的有效性。 |
| 关键词: 多源气象预报 转移概率 数值天气预报 功率预测 短期预测 光伏发电 |
| DOI：10.16081/j.epae.202201019 |
| 分类号:TM615 |
| 基金项目:国家重点研发计划项目(2020YFB1506804)；国家自然科学基金资助项目(51907140) |
|
| Short-term forecasting of photovoltaic power based on total irradiance correction of multi-source meteorological forecast |
|
SHI Haoqi, GUO Li, LIU Yixin, WANG Chengshan
|
|
Key Laboratory of Smart Grid of Ministry of Education, Tianjin University, Tianjin 300072, China
|
| Abstract: |
| Aiming at possible misstatement and deviation problems of a single weather forecast source, a short-term forecasting method of photovoltaic power based on total irradiance correction of multi-source meteorological forecast is proposed. According to the characteristics of power series, the self-organizing map neural network clustering algorithm is adopted to divide the generalized weather types of historical data. The matching correspondence between generalized weather types and weather type forecast from public weather service is realized according to the degree of sunshine, and the conversion coefficient between each level of total irradiance of different generalized weather types is calculated. On the basis of calculating system error of each generalized weather type, if the generalized weather type classification result of numerical weather prediction is consistent with that of weather type forecast from public weather service on the forecast day, the system error of total irradiance is corrected by superposition, otherwise the tree augmented naive Bayes algorithm is adopted to calculate the transfer probability of two generalized weather types, the conversion coefficient is used to calculate the total irradiance sequence of generalized weather type corresponding to the weather type forecast from public weather service after the correction of system error, and the weigh-ting coefficient is set by transfer probability to further modify the total irradiance sequence. The forecasting model is built, and the short-term photovoltaic power forecasting results are obtained based on back propagation neural network optimized by genetic algorithm. The validity of the model is verified by the actual operation data of a photovoltaic power plant and the weather forecast data. |
| Key words: multi-source meteorological forecast transfer probability numerical weather prediction power forecasting short-term forecasting photovoltaic power generation |
