Photovoltaic power generation prediction method based on improved generalized weather based on k-means clustering

Abstract

A photovoltaic power generation prediction method based on K-means clustering to improve generalized weather. The method clusters the daily historical output data of photovoltaic power plants into K clusters, and attaches digital labels to the data in the clusters; statistics correspond to the daily historical output data. The digital labels and meteorological professional weather constitute an improved generalized weather map corresponding to one or more digital labels of meteorological professional weather; for the historical output data of every two adjacent days, the historical output data of the first day and its corresponding digital labels 1. The digital label corresponding to the historical output data of the second day is used as an input, and the historical output data of the second day is used as an output to establish a BP neural network prediction model of photovoltaic power generation and use the model to predict the power generation. The present invention overcomes the shortcoming of being too absolute in the existing generalized weather mapping, and can not only accurately predict the amount of photovoltaic power generation in sunny weather, but also has good prediction in cloudy to cloudy, cloudy and rainy weather and other changeable weather precision.

Description

technical field [0001] The invention relates to a method capable of accurately predicting the power generation of a photovoltaic power station, which belongs to the technical field of photovoltaic power generation. Background technique [0002] Photovoltaic power is an intermittent energy source, which is affected by various factors such as seasonal characteristics, daily characteristics, weather characteristics and fluctuation characteristics. Under the influence of many complex factors, the establishment process of photovoltaic forecasting model is slightly complicated, but in most cases, the indirect forecasting method is still used, that is, the mathematical model is established based on the historical output data of photovoltaics. This method believes that the data itself covers information such as region and climate, and after statistical and classification processing of the data, a photovoltaic power generation model is established through some algorithms with self-le...

AI Technical Summary

Problems solved by technology

Since the photovoltaic output is easily affected by factors such as wind force and cloud cover of the day, even under the same professional meteorological weather (such as sunny to cloudy weather), there will often be a large difference in the respective power generation curves. Adding a generalized weather state label to historical output data, and using a certain generalized weather state to describe it alone will increase the forecast error, especially in variable weather such as sunny to cloudy and rainy. less accurate

Images