Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Learning-Based Electricity Price Forecast for Virtual Bidding in Wholesale Electricity Market

Xuesong Wang, Sharaf K. Magableh, Oraib Dawaghreh, Caisheng Wang, Jiaxuan Gong, Zhongyang Zhao, Michael H. Liao

TL;DR

This work addresses the volatile ERCOT price-spread relevant to virtual bidding in a two-settlement market. It presents a Transformer-based time-series model that predicts the distribution over five quantized price-spread classes for the next day using time, load, solar, and wind forecasts, with weekly walk-forward retraining to reflect realistic operation. Key contributions include a detailed comparison of training settings (lagging, data size, finetuning) and a suite of trading strategies evaluated by backtesting, identifying hour 19 as the most profitable trading opportunity under practical constraints. The results demonstrate profit potential from forecast-informed bidding while highlighting that traditional classification metrics may not align with end-to-end profitability, underscoring bidder-focused evaluation and the need for robustness enhancements in future work.

Abstract

Virtual bidding plays an important role in two-settlement electric power markets, as it can reduce discrepancies between day-ahead and real-time markets. Renewable energy penetration increases volatility in electricity prices, making accurate forecasting critical for virtual bidders, reducing uncertainty and maximizing profits. This study presents a Transformer-based deep learning model to forecast the price spread between real-time and day-ahead electricity prices in the ERCOT (Electric Reliability Council of Texas) market. The proposed model leverages various time-series features, including load forecasts, solar and wind generation forecasts, and temporal attributes. The model is trained under realistic constraints and validated using a walk-forward approach by updating the model every week. Based on the price spread prediction results, several trading strategies are proposed and the most effective strategy for maximizing cumulative profit under realistic market conditions is identified through backtesting. The results show that the strategy of trading only at the peak hour with a precision score of over 50% produces nearly consistent profit over the test period. The proposed method underscores the importance of an accurate electricity price forecasting model and introduces a new method of evaluating the price forecast model from a virtual bidder's perspective, providing valuable insights for future research.

Deep Learning-Based Electricity Price Forecast for Virtual Bidding in Wholesale Electricity Market

TL;DR

This work addresses the volatile ERCOT price-spread relevant to virtual bidding in a two-settlement market. It presents a Transformer-based time-series model that predicts the distribution over five quantized price-spread classes for the next day using time, load, solar, and wind forecasts, with weekly walk-forward retraining to reflect realistic operation. Key contributions include a detailed comparison of training settings (lagging, data size, finetuning) and a suite of trading strategies evaluated by backtesting, identifying hour 19 as the most profitable trading opportunity under practical constraints. The results demonstrate profit potential from forecast-informed bidding while highlighting that traditional classification metrics may not align with end-to-end profitability, underscoring bidder-focused evaluation and the need for robustness enhancements in future work.

Abstract

Virtual bidding plays an important role in two-settlement electric power markets, as it can reduce discrepancies between day-ahead and real-time markets. Renewable energy penetration increases volatility in electricity prices, making accurate forecasting critical for virtual bidders, reducing uncertainty and maximizing profits. This study presents a Transformer-based deep learning model to forecast the price spread between real-time and day-ahead electricity prices in the ERCOT (Electric Reliability Council of Texas) market. The proposed model leverages various time-series features, including load forecasts, solar and wind generation forecasts, and temporal attributes. The model is trained under realistic constraints and validated using a walk-forward approach by updating the model every week. Based on the price spread prediction results, several trading strategies are proposed and the most effective strategy for maximizing cumulative profit under realistic market conditions is identified through backtesting. The results show that the strategy of trading only at the peak hour with a precision score of over 50% produces nearly consistent profit over the test period. The proposed method underscores the importance of an accurate electricity price forecasting model and introduces a new method of evaluating the price forecast model from a virtual bidder's perspective, providing valuable insights for future research.

Paper Structure

This paper contains 10 sections, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methodology
  3. Model Design
  4. Data Preprocessing and Dataset Split
  5. Model Training
  6. Trading Strategy Design Principle
  7. Results and Analysis
  8. Model Performance Evaluation
  9. Trading Strategy Evaluation
  10. Conclusion and Future Work

Figures (5)

  • Figure 1: Price spread between real-time price and day-ahead price of system lambda in ERCOT market.
  • Figure 2: Model architecture, where the output $P_t$ represent the probability distribution of the price spread of hour 0:00 (In ERCOT, it needs to be converted into Hour Ending 1:00).
  • Figure 3: Annual solar generation Current Operation Plan (COP) trends from 2018 to 2024.
  • Figure 4: Confusion matrix of the best model for each hour. Precision, recall, and accuracy are best viewed when zoomed in.
  • Figure 5: Cumulation profit for the best model using different trading strategies.