Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Toward Value-oriented Renewable Energy Forecasting: An Iterative Learning Approach

Yufan Zhang, Mengshuo Jia, Honglin Wen, Yuexin Bian, Yuanyuan Shi

TL;DR

The paper addresses the misalignment between traditional forecast accuracy and operational value in renewable energy dispatch by introducing a value-oriented forecasting framework for sequential day-ahead and real-time optimization. It formulates forecast parameter learning as a bilevel problem and derives a dual-based loss, enabling gradients to be computed via optimal dual solutions and enabling neural networks to be trained to minimize expected operating costs. An iterative learning algorithm alternates between updating forecast parameters and updating the loss function, yielding forecasts that, while not necessarily more accurate by RMSE, reduce overall operation costs, especially at higher wind penetrations. The approach offers computational advantages over two-stage stochastic programs and extends to problems with binary variables, with demonstrated robustness to operational changes and clear interpretability through dual prices that connect forecasts to marginal costs. The work thus provides a practical, scalable tool for value-driven RES forecasting in sequential dispatch settings.

Abstract

Energy forecasting is an essential task in power system operations. Operators usually issue forecasts and leverage them to schedule energy dispatch ahead of time. However, forecast models are typically developed in a way that overlooks the operational value of the forecasts. To bridge the gap, we design a value-oriented point forecasting approach for sequential energy dispatch problems with renewable energy sources. At the training phase, we align the loss function with the overall operation cost function, thereby achieving reduced operation costs. The forecast model parameter estimation is formulated as a bilevel program. Under mild assumptions, we convert the upper-level objective into an equivalent form using the dual solutions obtained from the lower-level operation problems. Additionally, a novel iterative solution strategy is proposed for the newly formulated bilevel program. Under such an iterative scheme, we show that the upper-level objective is locally linear regarding the forecast model output, and can act as the loss function. Numerical experiments demonstrate that, compared to commonly used statistical quality-oriented point forecasting methods, forecasts obtained by the proposed approach result in lower operation costs. Meanwhile, the proposed approach is more computationally efficient than traditional two-stage stochastic programs.

Toward Value-oriented Renewable Energy Forecasting: An Iterative Learning Approach

TL;DR

The paper addresses the misalignment between traditional forecast accuracy and operational value in renewable energy dispatch by introducing a value-oriented forecasting framework for sequential day-ahead and real-time optimization. It formulates forecast parameter learning as a bilevel problem and derives a dual-based loss, enabling gradients to be computed via optimal dual solutions and enabling neural networks to be trained to minimize expected operating costs. An iterative learning algorithm alternates between updating forecast parameters and updating the loss function, yielding forecasts that, while not necessarily more accurate by RMSE, reduce overall operation costs, especially at higher wind penetrations. The approach offers computational advantages over two-stage stochastic programs and extends to problems with binary variables, with demonstrated robustness to operational changes and clear interpretability through dual prices that connect forecasts to marginal costs. The work thus provides a practical, scalable tool for value-driven RES forecasting in sequential dispatch settings.

Abstract

Energy forecasting is an essential task in power system operations. Operators usually issue forecasts and leverage them to schedule energy dispatch ahead of time. However, forecast models are typically developed in a way that overlooks the operational value of the forecasts. To bridge the gap, we design a value-oriented point forecasting approach for sequential energy dispatch problems with renewable energy sources. At the training phase, we align the loss function with the overall operation cost function, thereby achieving reduced operation costs. The forecast model parameter estimation is formulated as a bilevel program. Under mild assumptions, we convert the upper-level objective into an equivalent form using the dual solutions obtained from the lower-level operation problems. Additionally, a novel iterative solution strategy is proposed for the newly formulated bilevel program. Under such an iterative scheme, we show that the upper-level objective is locally linear regarding the forecast model output, and can act as the loss function. Numerical experiments demonstrate that, compared to commonly used statistical quality-oriented point forecasting methods, forecasts obtained by the proposed approach result in lower operation costs. Meanwhile, the proposed approach is more computationally efficient than traditional two-stage stochastic programs.
Paper Structure (26 sections, 1 theorem, 26 equations, 7 figures, 6 tables, 1 algorithm)

This paper contains 26 sections, 1 theorem, 26 equations, 7 figures, 6 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Preliminaries: Day-ahead and Real-time Energy Dispatch
  3. Deterministic Dispatch
  4. Two-stage Stochastic Program based Dispatch
  5. Improved Deterministic Dispatch
  6. Methodology
  7. Training Phase
  8. Operational Forecasting Phase
  9. Iterative Algorithm for Parameter Estimation
  10. Iterative Learning Algorithm
  11. Interpretating Value-oriented Forecasts
  12. Experiement Setup
  13. Description of Operation Simulations and Data
  14. Benchmark Models
  15. Verification Metrics
  16. ...and 11 more sections

Key Result

Proposition 1

Consider the convex optimization problems 7d-7g and 7h-7k in the form of linear programs (LPs), with the parameterized dual solutions associated with the constraints 7g and 7k are $\lambda_{d,\tau}^*(\Tilde{y}_{d,\tau})$ and $\nu_{d,\tau}^*(\Tilde{y}_{d,\tau},y_{d,\tau})$, respectively. The optimal where the notations $\psi^{D,*}_{d,\tau}$ and $\psi^{R,*}_{d,\tau}$ denote the remaining terms of t

Figures (7)

  • Figure 1: The illustration of deterministic, stochastic, and improved deterministic dispatch, where $k$ is the time interval between $t$ and 1 am on day $d$.
  • Figure 2: Illustration of the proposed bilevel program \ref{['7']} for model training of value-oriented forecasting.
  • Figure 3: The illustration of iterative scheme at the epoch $e$.
  • Figure 4: 4-day wind power forecast profiles issued by the value- and quality-oriented forecasting approaches. Qua-E and Qua-Q stand for quality-oriented forecasts that issue the expectation and $\frac{2}{9}$ quantile of the wind power.
  • Figure 5: Average hourly cost reduction of the proposed approach compared to Qua-E under different levels of difference between DA marginal cost and RT marginal cost/utility.
  • ...and 2 more figures

Theorems & Definitions (3)

  • Remark 1
  • Proposition 1
  • proof