Solar Power Generation Prediction And Optimization In Electricity Distribution

Abstract

Fossil fuels, a non-renewable source of energy, are gradually depleted. To cope with the potential energy crisis on a global scale, renewable energy sources such as solar power have been the top candidate to replace fossil fuels in energy mass production. To become a permanent replacement, more and more studies are required to help enterprises further reduce the production expense of solar power plant which is still relatively high nowadays, making solar energy generation become more applicable in many corners of the world. As machine learning algorithm and optimization techniques are becoming more popular and applied widely in many industries, it paves the way for the massive growth of solar power production in the near future. This research aims to achieve two main objectives. Firstly, a predicting model using Linear Regression technique is developed to perform energy generation prediction. Thanks to which, the electricity supplier can make timely decision to combat the power shortage and fulfill the demands. The model is capable of providing reliable results with its average R-squared value of 98% and 82% respectively with two datasets. Secondly, a model under Mixed Integer Linear Programming approach is introduced to suggest an optimal distribution plan for the electricity vendor, with cost savings features in consideration. Some recommendations to avoid the potential cost burden are also made after some data adjustments. For future research, more novel machine learning methods and weather forecast data can be explored to improve the model accuracy and reliability, and more cost components can be added to the optimization model to further lower the expenses.