Deep Ensemble Technique for Short-Term Load Forecasting Using Smart Meter Data

There has been a growing interest in using the smart meter data for accurate and effective energy management with the availability of smart meter data sourced from homes. Personal load profiles are more complex and hence difficult to forecast compared with aggregate loads. To handle the above tasks, a deep ensemble network is proposed using Spike Neural Network and Long Short-Term Memory to accurately model these predictions. This system utilizes the profiles of different users that are incorporated into the task’s understanding. The consumer profiles are subjected to anomaly detection using Spike Neural Network, which detected the anomalies in the data in an online fashion, then the data without anomalies are fed into the Long Short-Term Memory neural framework. The experimental results of the proposed method proved to be superior when compared with the result of Spike Neural Network and Long Short-Term Memory model separately.