A second-level ultra-short-term photovoltaic power prediction method

Abstract

The invention provides a second-level ultra-short-term photovoltaic power prediction method, which belongs to the technical field of photovoltaic power generation. The present invention uses the LSSVM algorithm to establish a prediction model, because the hyperparameters of the LSSVM algorithm have a great influence on the prediction performance, and the chaotic cube mapping is used to generate more random chaotic numbers, the calculation of the initial population position is optimized, and the population is initialized in combination with the chaotic numbers; The first iteration of the gray wolf optimization algorithm optimizes the hyperparameters of LSSVM to obtain a better solution; the second optimization uses the improved Griewank function as the disturbance function to search for a better solution through iterative local search and obtains a better solution; the third optimization The local adaptive difference algorithm LSaDE is used to search for a better solution, and the optimal solution is obtained; the hyperparameters are determined, and the prediction model is trained to obtain the predicted value. Determine the model accuracy according to the accuracy evaluation results, and re-optimize the parameters if it fails to meet the standard. The invention realizes the prediction effect of short prediction time and high prediction accuracy.

Description

technical field [0001] The invention relates to the technical field of photovoltaic power generation, in particular to a second-level ultra-short-term photovoltaic power prediction method. Background technique [0002] In recent years, the newly installed capacity of photovoltaics has increased dramatically year by year, and the penetration rate of photovoltaic power generation in the entire power grid system has become higher and higher. It has the characteristics of randomness and discontinuity, which makes the photovoltaic output power fluctuate greatly and has a great impact on the power grid. Prediction and energy storage technologies are key technologies to stabilize fluctuations in photovoltaic output power. [0003] In the process of photovoltaic power generation, power fluctuation is a normal phenomenon. However, under normal circumstances, the average change rate of power fluctuation does not exceed 3% / s. However, in complex weather conditions, due to the movement...

AI Technical Summary

Problems solved by technology

The prediction algorithm with a short prediction time has a serious loss in prediction accuracy; the prediction algorithm with a high prediction accuracy has a long running time and cannot meet the requirements of second-level ultra-short-term power prediction

[0008] 2. It is difficult to set key parameters

The existing technology either uses manual experience or an optimization algorithm with a long optimization cycle, which takes a long time and has no obvious effect

[0009] 3. Difficulty obtaining sample data

Most of the existing technologies use comprehensive weather, cloud information and historical output power as input samples, and the acquisition accuracy of weather and cloud information is not high and expensive, which does not meet the requirements of economic operation

Images