Spatial Modeling and Monitoring Electricity Consumption using Generalized Likelihood Ratio Control Chart

Energy monitoring using statistical process control (SPC) methods makes it more straightforward to identify patterns and trends to decrease energy consumption more effectively. The literature review of energy consumption monitoring with SPC techniques generally focuses on the temporal aspect of variation. However, due to the spatial nature of energy data, enhancing these methods to incorporate temporal and spatial aspects would improve the accuracy of the diagnostic information, underscoring simultaneous detection of the time and location of changes. Thus, the main novelty of this work is the spatial modeling and spatiotemporal monitoring of electricity consumption. For this purpose, the study used actual electricity consumption data from eight western cities of Mazandaran province in the north of Iran for spatial modeling using spatial regression models and a geographically weighted regression (GWR) model. The prediction performance evaluation of spatial models showed GWR as an appropriate model, whose coefficients were monitored through a generalized likelihood ratio (GLR) chart in phase II. The GLR chart detected two changes in consumption, and its performance was confirmed based on the statements from electricity experts relying on meteorological information and floating population data. Furthermore, the performance of the GLR chart was evaluated using out-of-control average run length (ARL1) across three different scenarios. The findings indicate that the GLR chart can effectively detect any sizes of shifts (δ), ranging from 5% to 100% of the model's parameter value. Additionally, with larger values of δ, the ARL1 decreases, resulting in faster detection of changes in the model.