Prediction Markets for Probabilistic Forecasting of Renewable Energy Sources

Mahdieh Shamsi,Paul Cuffe

doi:10.1109/tste.2021.3112916

Mahdieh Shamsi, Paul Cuffe

Open Access

https://doi.org/10.1109/tste.2021.3112916

Copy DOI

Abstract

This paper demonstrates how a binary prediction market is capable of achieving a probabilistic renewable energy forecast. In prediction markets, participants trade shares associated with the outcome of unknown future events (here, the renewable production, as a random variable), and the instantaneous price of shares represents the probability of the outcome. The focus of this study is to exploit this informational value of the prediction market price in renewable energy forecasting. To this end, in this paper three different methods of renewable probabilistic forecasting have been considered as the trading agents in a binary prediction market, the aggregated probability of the renewable output is elicited from the equilibrium price in this market and finally, the full cumulative distribution function of possible renewable output is extracted through regression analysis. The proposed method is applied to the test cases of three onshore wind farms in Australia. The simulation results suggest that the performance of the proposed method is superior to the individual models and forecasting is improved in terms of reduction in the electricity market imbalance costs.

Highlights

R ENEWABLE energy is typically forecast by deterministic methods which provide a single value for the future production, referred to as point forecasts [1]
Prediction markets use the wisdom of the crowd principle to aggregate information and provide accurate forecasts of unknown future events
The production of a renewable energy source is a random variable for which its probabilistic forecasting can be obtained through such a market

Summary

Introduction

R ENEWABLE energy is typically forecast by deterministic methods which provide a single value for the future production, referred to as point forecasts [1]. The uncertainty associated with renewable energy forecasting poses technical and economical challenges to the operation and management of power systems [2]. At the system operator level, the tasks such as reserve requirement determination [3], unit commitment, and economic dispatch [4] will be affected by the forecast accuracy. From the perspective of Renewable Energy Sources (res), this issue affects their trading strategies in electricity markets and their consequent revenue streams, as investigated in [5] and [6] for wind power, and in [7] for solar power. Within the scope of this paper, it suffices to note that these methods can be categorised into two groups: parametric

Objectives

Methods

Results

Conclusion