Smart meter data analytics and power forecasting

• Main Author: Goh, Benedict Wei Zhang
• Other Authors: Xu Yan
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering; Power forecasting; Data analytics; Smart meter
• Online Access: https://hdl.handle.net/10356/176606
• Institution: Nanyang Technological University

With the proliferation of smart meters in households resulting in the ability to consistently obtain household specific energy consumption data at set intervals, there is a large volume of data available to both consumers and utility providers, creating opportunities to gain valuable insights into areas such as demand response and load forecasting. This project explores the energy consumption data from a smart meter enabled household and apply statistical methods and machine learning methodologies such as ARIMA, LSTM and CNN-LSTM to perform accurate forecasting of the household’s energy consumption. The models are evaluated based on commonly used time series forecasting error metrics such as MSE, RMSE, MAE and MAPE to obtain insights into the best performing model. Further hyper parameter tuning is conducted for each model to determine the most optimal configuration and the results show that LSTM models are the most consistent in providing accurate forecasts.