Investigation and Correction of Anomalies in Smart Home Energy Consumption Data

Abstract

Smart grid evolution is ramping up in the global energy scenario by offering deregulated markets, demand-side management, prosumer culture, demand response, contingency forecasting,outage management, etc., functionalities. Further, the progressive developments in information newlineand communication technologies improve smartness in the traditional power grids. newlineSmart homes/buildings are key sub-categories of smart grids. Especially, smart homes have newlinegained widespread popularity and significance in the present energy sector, which possesses newlinethe communication between various devices/appliances and collect their functional data in terms of energy consumption readings, timestamp, etc. The advanced metering infrastructure newlineconnected to them continuously captures the energy consumption data at predefined rates and stores it as datasets in a file format specified by the utilities. Any deviation in this process is called anomalous tracing, which affects the accuracy of analytics. Usually, the anomalous tracing is due to the malfunctioning of the metering infrastructure, network congestion, unexpected breakdown of communication networks, server station issues, etc. Hence, there is a dire need of identifying the ways of analyzing the captured data to find the data anomalies (e.g., anomalous records, missing data, redundant data, outliers, and garbage data) that usually occur because of aforesaid failures. These data anomalies greatly impact energy analytics newlineviz., load estimation and management, energy pricing, optimizing assets, planning, decisionmaking, etc. Hence, the availability of high-quality data is always desirable concerning all the above-mentioned functionalities. newlineTo comprehend the energy consumption dataset, customer behavior, aforesaid data anomalies, newlineand their behavior, this thesis presents various analytical models. To achieve these objectives, an analytical model is proposed to explore the dataset as well as customer behavior, a 3-step (extraction, comparison, identification) analytical approach is proposed to exp