Probabilistic Forecasting of Regional Net-Load With Conditional Extremes and Gridded NWP

Abstract

The increasing penetration of embedded renewables makes forecasting net-load, consumption less embedded generation, a significant and growing challenge. Here a framework for producing probabilistic forecasts of net-load is proposed with particular attention given to the tails of predictive distributions, which are required for managing risk associated with lowprobability events. Only small volumes of data are available in the tails, by definition, so estimation of predictive models and forecast evaluation requires special attention. We propose a solution based on a best-in-class load forecasting methodology adapted for netload, and model the tails of predictive distributions with the Generalised Pareto Distribution, allowing its parameters to vary smoothly as functions of covariates. The resulting forecasts are shown to be calibrated and sharper than those produced with unconditional tail distributions. In a use-case inspired evaluation exercise based on reserve setting, the conditional tails are shown to reduce the overall volume of reserve required to manage a given risk. Furthermore, they identify periods of high risk not captured by other methods. The proposed method therefore enables user to both reduce costs and avoid excess risk.

Reliable Provision of Ancillary Services from Aggregated Variable Renewable Energy Sources through Forecasting of Extreme Quantiles

“…The Extreme Value Theory (EVT) proposes a more robust framework for the prediction of extremes, which are modelled with a Pareto distribution [9]. Few publications have dealt with EVT in the context of renewable production forecasting, with the exceptions of [5], [6] and [10] who propose EVT forecasts of extremal quantiles of VRE production. In [6], a gradient boosting model predicts non extremal quantiles that are further ranked by a similarity measure.…”

“…Extreme levels of these predictions serve to fit the Pareto distribution modelling the distribution tail. In [10], extreme quantiles of net-load are predicted by integrating EVT into an additive model that is conditioned on a grid of weather prediction for the region of interest.…”

“…The main contributions of this paper are: 1) A methodology to forecast low quantiles of VREbased VPP production, exploring alternative approaches based on statistical and machine learning models. Existing forecasting approaches for low quantiles [6], [10] address problems of low dimensionality (single VRE source, load). However they do not scale well with the size and heterogeneous type of explanatory variables required to predict the production of a VRE-VPP.…”

Adaptive Probabilistic Forecasting of Electricity (Net-)Load

“…Then we apply our approach on two data sets. In Section IV, we consider the regional net-load in Great Britain, and we extend the work of [21] to the adaptive setting, also extending the data set with more recent data (including the coronavirus crisis). Furthermore, we show in Section IV-D that adaptive models have fewer needs for good explanatory variables; indeed, removing two (difficult to obtain) variables from the model has a lower impact on the adaptive variant.…”

Economic-environmental dispatch for the integrated energy system considering off-design conditions

“…Based on the collected historical off-design conditions' data, many representative ML approaches can be used to predict the variable device efficiencies, such as polynomial regression (Li and Yao, 2021), support vector machines (Liu et al, 2020), and deep neural network (DNN) (Ghimire et al, 2019). The DNN has remarkable performance in both the computational accuracy and speed for nonlinear parameter forecast and has been applied in many machines' intelligence fields, such as image recognition (Chen S. et al, 2022), parameter forecast (Browell and Fasiolo, 2021), and system control .…”