Photovoltaic power station generating capacity predicting method based on SOM neural network data cluster identification

Abstract

The invention discloses a photovoltaic power station generating capacity predicting method based on SOM neural network data cluster identification. The method includes the steps of dividing data samples into multiple classes according to SOM neural network clustering result in a first step, respectively importing into BP neural networks for network training to obtain various trained BP neural networks, inputting processed prediction daily data into trained SOM neural networks for identification to obtain corresponding data classes, and importing the prediction daily input vectors into corresponding trained BP neural networks for prediction. According to the photovoltaic power station generating capacity predicting method based on SOM neural network data cluster identification, historical data of a photovoltaic power station can be clustered according to self characteristics, and the power generating capacity can be predicted after self identification and classification of data samples. The power generating capacity prediction precision can be effectively improved.

Description

technical field [0001] The invention relates to a method for predicting the power generation of a photovoltaic power station based on SOM neural network data clustering recognition, and belongs to the field of power generation prediction of a photovoltaic power station. Background technique [0002] In the 21st century, photovoltaic power generation technology has been greatly developed. It is clean, non-polluting and renewable, and has been favored by researchers. The power generation forecast of photovoltaic power stations plays a vital role in power stations. This The field has not started late in our country and is not yet mature, and there is a common problem of low prediction accuracy. At present, most researchers will carry out data preprocessing in the process of predicting power generation, including the classification of weather types, and make predictions according to different weather types to improve the prediction accuracy. However, power generation prediction ...

AI Technical Summary

Problems solved by technology

[0002] In the 21st century, photovoltaic power generation technology has been greatly developed. It is clean, non-polluting and renewable, and has been favored by researchers. The power generation forecast of photovoltaic power stations plays a vital role in power stations. The field has not started late in our country and is not yet mature, and there is a common problem of low prediction accuracy

At present, most researchers will carry out data preprocessing in the process of predicting power generation, including the classification of weather types, and make predictions according to different weather types to improve the prediction accuracy. However, power generation prediction based on weather type classification Sometimes there are problems with the method. The principle of improving forecasting accuracy through classification of weather types is that the potential change patterns of data with the same weather type are approximately similar. In fact, the weather conditions are changeable and the nonlinear characteristics are extremely strong. There may be a variety of weather conditions in a day Weather characteristics, such as sunny in the morning, cloudy at noon, rainy in the afternoon, or a rainy day on a certain day, but the data changes of the power station are similar to sunny days, so the forecasting method based on the weather type classification will not improve the prediction accuracy, but will reduce

Images