The Effectiveness of Deep Learning Methods for Smart Meter Data Analytics

Abstract

Abstract: The introduction of smart grid technology came with the promise of better management of energy and power systems, an essential step toward achieving sustainable energy. Smart meter analytics is viewed as a critical building block for sustainable energy since it improves operational efficiency, enhances power grid reliability, and enables better management of energy resources. This thesis presents a unifying taxonomy to classify the various domains of smart meter analytics and their underlying functions and techniques. This thesis also investigates the deep-learning architectures used for each domain. Current research trends, approaches, and opportunities are introduced and discussed through a the comprehensive updated taxonomy to understand the field of smart meter analytics. This thesis proposes a novel embedded deep clustering base framework for the task of load profiling. The proposed load profiling framework's novelty is twofold. First, using an autoencoder the framework represents data as a hierarchy within the layers of the deep network, allowing for dimensionality reduction and highly nonlinear decision separation between clusters at the autoencoder bottleneck level. This thesis further investigates the forecasting of big temporal data by Seasonal Autoregressive Integrated Moving Average Model (SARIMAX) and Prophet methods. Hence, accurate and computationally cheap load forecasting techniques are a crucial part of the grid system. Both the deep clustering framework and forecasting methods are then evaluated and tested using real-world smart meter data sets, and the experimental results show that the proposed framework produces significantly better load profiles when compared to classical clustering algorithms as well as previous hybrid frameworks proposed in the literature.