Adaptive Methods for Short-Term Electricity Load Forecasting During COVID-19 Lockdown in France

Frugal day-ahead forecasting of multiple local electricity loads by aggregating adaptive models

This paper focuses on day-ahead electricity load forecasting for substations of the distribution network in France; therefore, the corresponding problem lies between the instability of a single consumption and the stability of a countrywide total demand. Moreover, this problem requires to forecast the loads of over one thousand substations; consequently, it belongs to the field of multiple time series forecasting. To that end, the paper applies an adaptive methodology that provided excellent results at a national scale; the idea is to combine generalized additive models with state-space representations. However, extending this methodology to the prediction of over a thousand time series raises a computational issue. It is solved by developing a frugal variant that reduces the number of estimated parameters: forecasting models are estimated only for a few time series and transfer learning is achieved by relying on aggregation of experts. This approach yields a reduction of computational needs and their associated emissions. Several variants are built, corresponding to different levels of parameter transfer, to find the best trade-off between accuracy and frugality. The selected method achieves competitive results compared to individual models. Finally, the paper highlights the interpretability of the models, which is important for operational applications.

An evaluation of the impact of COVID-19 lockdowns on electricity demand

The COVID-19 pandemic has given rise to significant changes in electricity demand around the world. Although these changes differ from region to region, countries that have implemented stringent lockdown measures to curtail the spread of the virus have experienced the greatest alterations in demand. Within Australia, the state of Victoria has been subject to the largest number of days in hard lockdown during the COVID-19 pandemic. We conduct an exploratory data analysis to identify predictors of demand, and have built a time series forecasting model to predict the half-hourly electricity demand in Victoria. Our model distinguishes between lockdown periods and non-restrictive periods, and aims to identify a variety of patterns that we show to be influential on electricity demand. The model thereby provides a nuanced prediction of electricity demand that captures the shifting demand profile of intermittent lockdowns.

Adaptive Load Forecasting Methodology Based on Generalized Additive Model with Automatic Variable Selection

For decentralized energy management in a smart grid, there is a need for electric load forecasting at different places in the grid hierarchy and for different levels of aggregation. Load forecasting functionality relies on the load time series prediction model, which provides accurate forecasts. Complex and heterogeneous multi-source load time series in a smart grid require flexible modeling approaches to meet the accuracy demand. This work proposes an adaptive load forecasting methodology based on the generalized additive model (GAM) with the big data estimation method. It is based on a set of GAM terms, constructed for a specific multi-source load forecasting application in the grid and a procedure that dynamically selects the most relevant terms and generates forecasts for particular load time series. Data from publicly available New York Independent System Operator (NYISO) databases are used for testing. The 24-hour-ahead forecasting results for eleven New York City zones, of different sizes and types, indicate the applicability of the proposed methodology.

Impact of COVID-19 on electricity energy consumption: A quantitative analysis on electricity

In addition to the tremendous loss of life due to coronavirus disease 2019 (COVID-19), the pandemic created challenges for the energy system, as strict confinement measures such as lockdown and social distancing compelled by governments worldwide resulted in a significant reduction in energy demand. In this study, a novel, quantitative and uncomplex method for estimating the energy consumption loss due to the pandemic, which was derived from epidemiological data in the beginning stages, is provided; the method bonds a data-driven prediction (LSTM network) of energy consumption due to COVID-19 to an econometric model (ARDL) so that the long- and short-term impact can be synthesized with adequate statistical validation. The results show that energy loss is statistically correlated with the time-changing effective reproductive number (R_t) of the disease, which can be viewed as quantifying confinement intensity and the severity of the earlier stages of the pandemic. We detected a 1.62% decrease in electricity consumption loss caused by each percent decrease in R_t on average. We verify our method by applying it to Germany and 5 U.S. states with various social features and discuss implications and universality. Our results bridge the knowledge gap between key energy and epidemiological parameters and provide policymakers with a more precise estimate of the pandemic’s impact on electricity demand so that strategies can be formulated to minimize losses caused by similar crises.

Entropy-Based Anomaly Detection in Household Electricity Consumption

Energy efficiency is one of the most important current challenges, and its impact at a global level is considerable. To solve current challenges, it is critical that consumers are able to control their energy consumption. In this paper, we propose using a time series of window-based entropy to detect anomalies in the electricity consumption of a household when the pattern of consumption behavior exhibits a change. We compare the accuracy of this approach with two machine learning approaches, random forest and neural networks, and with a statistical approach, the ARIMA model. We study whether these approaches detect the same anomalous periods. These different techniques have been evaluated using a real dataset obtained from different households with different consumption profiles from the Madrid Region. The entropy-based algorithm detects more days classified as anomalous according to context information compared to the other algorithms. This approach has the advantages that it does not require a training period and that it adapts dynamically to changes, except in vacation periods when consumption drops drastically and requires some time for adapting to the new situation.

Integrative Smart Grids’ Assessment System

The COVID-19 pandemic has significantly affected the energy sector. The new behavior of industrial and non-commercial consumers changes the energy consumption model. In addition, the constraints associated with the coronavirus crisis have led to environmental effects from declining economic activity. The research is based on evidence from around the world showing significant reductions in emissions and improved air quality. This situation requires rethinking the energy development strategy, particularly the construction of smart grids as a leading direction of energy development. Evaluating the efficiency of smart grids is a vital tool for disseminating successful experience in improving their management. This paper proposes an approach to a comprehensive assessment of smart grids based on a comparative analysis of existing methods, taking into account the changes that need to be considered after the experience gained from the COVID-19 pandemic. The approach provides an accurate set of efficiency indicators for assessing smart grids to account for the direct and indirect effects of smart grids’ implementation. This evaluation approach can be helpful to policymakers in developing energy efficiency programs and implementing energy policy.

Residential Short-Term Load Forecasting during Atypical Consumption Behavior

Short-term load forecasting (STLF) is a fundamental tool for power networks’ proper functionality. As large consumers need to provide their own STLF, the residential consumers are the ones that need to be monitored and forecasted by the power network. There is a huge bibliography on all types of residential load forecast in which researchers have struggled to reach smaller forecasting errors. Regarding atypical consumption, we could see few titles before the coronavirus pandemic (COVID-19) restrictions, and afterwards all titles referred to the case of COVID-19. The purpose of this study was to identify, among the most used STLF methods—linear regression (LR), autoregressive integrated moving average (ARIMA) and artificial neural network (ANN)—the one that had the best response in atypical consumption behavior and to state the best action to be taken during atypical consumption behavior on the residential side. The original contribution of this paper regards the forecasting of loads that do not have reference historic data. As the most recent available scenario, we evaluated our forecast with respect to the database of consumption behavior altered by different COVID-19 pandemic restrictions and the cause and effect of the factors influencing residential consumption, both in urban and rural areas. To estimate and validate the results of the forecasts, multiyear hourly residential consumption databases were used. The main findings were related to the huge forecasting errors that were generated, three times higher, if the forecasting algorithm was not set up for atypical consumption. Among the forecasting algorithms deployed, the best results were generated by ANN, followed by ARIMA and LR. We concluded that the forecasting methods deployed retained their hierarchy and accuracy in forecasting error during atypical consumer behavior, similar to forecasting in normal conditions, if a trigger/alarm mechanism was in place and there was sufficient time to adapt/deploy the forecasting algorithm. All results are meant to be used as best practices during power load uncertainty and atypical consumption behavior.