Prediction for Photovoltaic Power Generation Based on Multi Feature Gradient Descent Mixture Model

Due to the crisis of non-renewable energy exhaustion and the need of environmental protection of the Earth, solar energy, as a renewable and clean energy, has become a highly recommended energy use. Although government agencies are attempting to achieve self-sufficiency through sustainable energy generation, accurate cost estimation is difficult to determine due to uncontrollable natural factors such as sunlight and wind speed, and the data throughout the year is not average If an intelligent algorithm tool can be designed that can predict the production capacity of solar energy in a specific region by analyzing factors related to annual solar production, it will greatly improve the efficiency of sustainable energy investment. This article designs corresponding photovoltaic power generation prediction algorithms, using support vector regressor (SVR), eXtreme gradient boosting (xgboost), and multilayer perceptron (MLP) neural networks for modeling and optimizing the corresponding models. The key is to compare different algorithms for photovoltaic prediction, and the evaluation shows that the model established using xgboost has the smallest error in the test set, with an accuracy rate of up to 97.5%.