2022 · 15 citations · 22 references
ComputingData ScienceForecastingMachine LearningTime Series AnalysisDemand-side Energy ManagementDemand-side ManagementElectricity Load ManagementEnergy Demand ForecastingLoad ForecastingTime Series ForecastingCharging Demand ForecastingDay-ahead ForecastingElectric Load ForecastingElectricity Consumption ForecastingRandom ForestsResidential Load ForecastingSmart ChargingStatistical Forecasting Models
The development of electric vehicles (EV) is a major lever towards low carbon transportation. It comes with increasing numbers of charging infrastructures which can be smartly managed to control the CO2 cost of EV electricity consumption or used as flexible assets for grid management. To achieve that, an efficient day-ahead forecast of charging behaviours is required at different spatial resolutions (e.g., household and public stations). We propose an extensive benchmark of 14 models for both load and occupancy day-ahead forecasts, covering 8 open charging session datasets of different types (residential, workplace and public stations). Two modelling approaches are compared: direct and bottom-up. The direct approach forecasts the aggregated load (resp. occupation) directly of an area/station whereas the bottom-up approach models each individual EV charging session before aggregating them. This second approach is key to the effective implementation of smart charging strategies. We consider both machine learning models (Random Forests and Gated Recurrent Units) and statistical models (Generalised Additive Models, Poisson Regression, Mixture Regression, Auto-Regressive) in order to maximise the spectrum of our benchmark. We finally propose an adaptive aggregation strategy to assess the variety of forecasts at hand. Overall, we demonstrate that direct approaches reach better performances than bottom-up approaches across all datasets considered. We further show that the different approaches used can lead to an improved performance of direct approaches when using an adaptive aggregation strategy. In fact, our best model produces a forecast which is more than 5 times better relative to the persistence on residential data.
22
Leo Breiman · Machine Learning · 2001 · 126.2K citations · Full text
Sepp Hochreiter, Jürgen Schmidhuber · Neural Computation · 1997 · 98.1K citations
Leo Breiman · Machine Learning · 1996 · 16.4K citations · Full text
On a measure of lack of fit in time series models
Greta M. Ljung, George E. P. Box · Biometrika · 1978 · 6K citations
Time Series Analysis: Forecasting and Control
Beat Kleiner, G. E. P. Box, G. J. Jenkins · Technometrics · 1977 · 3.8K citations
Computing, Data Science, Forecasting +11