Solar Power Plant Monitoring and Maintenance Using Machine Learning to Improving Energy Production

Description

Human sustainable growth needs to transform the use of energy. Additionally, also need to switch to clean and green sources of energy. In this context, the Indian government is making a huge investment using the PPP model for establishing solar power plants. The main issue with this source of energy is to perform timely maintenance and cleaning because the solar power plant’s performance is significantly influenced by dust and other natural factors. Additionally, due to the establishment of a large number of solar panels in solar power generation plants manual monitoring, and maintenance a critical issues. Therefore, it is required to develop an automated system to monitor and maintain the solar power plant for achieving higher performance.

This project aims to design an intelligent model to monitor and maintain solar panels for efficient performance. The simulation includes the analysis of Internet of Things (IoT) enabled power generation plant monitoring data. For analyzing the solar power plant monitoring data Machine Learning based infrastructure has been used. Additionally, by measuring and predicting the performance of solar power plants, the proposed algorithm recommends or notifies the administrator to perform maintenance manually or automatically. The considered dataset can be downloadable from the Kaggle (Download Dataset). The dataset consists of a Time series Problem. Therefore, the time series analysis has been done, with the transformation of data.  The Data transformation is performed to convert the time series to a 2D Vector for performing training of the Machine Learning algorithm. Two popular Machine Learning algorithms namely Long Short Term Memory (LSTM) and 1D Convolutional Neural Network.

After training both the models are tested using validation samples and performance has been measured using different performance metrics. The matrices namely precision, recall, f-score, accuracy, and loss have been considered. The implementation has been done using Python technology and can be executed on Google Colab.