Anomalies detection for smart-home energy forecasting using moving average

Abstract

In the past few years, the increase in the relation between the physical and digital world over the internet was witnessed. Even though the applications can enhance smart home systems, it is still early stages and challenges in the field of internet of things (IoT). An extreme level of data quality (DQ) system management is essential to produce a meaningful vision. However, in most home energy management system has no straightforward process of removing abnormal data. Hence, the research aims to propose and validate the model of anomaly detection for power consumption in real-time. The moving average (MA) approach identifies and removes abnormal energy consumption data. The results obtained from the forecasting time series auto regressive integrated moving average (ARIMA) model demonstrated that the proposed heuristics effectively enhanced energy usage forecasting. The selection of optimum parameter values for the MA was comprehended for time-series forecasting error minimization by comparing mean squared error (MSE). These outcomes proved the effectiveness of the existing technique and precision of choice of the appropriate. Therefore, the method can effectively route the cleaned sequence data streams in a real-time environment, which is valuable for spotting the anomalies and eliminating for enhancing energy usage time series.