Smart home data analytics using AI tech

• Main Author: Li, Kexing
• Other Authors: Xu Yan
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/158072
• Institution: Nanyang Technological University

Smart home is popular because of its intelligence, convenience and other excellent characteristics. Since 2019, global consumers' spending on smart home related devices has been increasing. It is expected that the spending will increase by US $88 billion in 2025. Taking smart home as the platform for load forecasting and improving the accuracy of load forecasting can promote the energy management of smart grid and users, reduce energy waste and reduce users' electricity consumption. Compared with the lack of accuracy of traditional prediction, deep learning based on artificial intelligence has the advantages of high fault tolerance and strong nonlinear learning ability. It is easier to obtain effective information from historical data, and the prediction accuracy is greatly improved. It is one of the hottest research directions in recent years. However, the previous research on load forecasting is still less applied in residential power consumption. This design carries out data analysis and load forecasting of smart home through artificial neural network. Firstly, collected the historical load data of smart home to form a data set. Then the data set was optimized and trained. Finally, the trained neural network model was used for load forecasting to study the application of machine learning in residential power consumption. This design uses the tensorflow machine learning framework in Python program and keras library to build three neural network models and predict the load data. The mean absolute error is used to show the prediction effect. The actual load and the predicted load are drawn on the same image for comparison. The actual load of the first 24 hours is predicted. Among the three neural networks, the accuracy of RNN is the highest, that of CNN is the lowest, and that of LSTM is between the two, though their errors are within the acceptable range. But overall, the results show that the neural networks all have high accuracy and have a good effect on load forecasting. This method has a good prospect in power system.