• Aucun résultat trouvé

Presentation of the paper Probabilistic Forecasting of Imbalance Prices in the Belgian Context.

N/A
N/A
Info
Télécharger
Protected

Academic year: 2021

Partager "Presentation of the paper Probabilistic Forecasting of Imbalance Prices in the Belgian Context."

Copied!
22
0
0

Chargement.... (Voir le texte intégral maintenant)

Texte intégral

(1)

Probabilistic Forecasting of

Imbalance Prices

in the Belgian Context

Jonathan Dumas

(2)

Introduction

2

Focus on imbalance prices

Probabilistic forecasting

A two-step probabilistic forecasting approach in the Belgian context

(3)

Summary

• Problem formulation

• Belgian Context

• A two-step probabilistic approach

• Case study description

• Numerical results

(4)

Problem formulation

4 Resolution Horizon Frequency Inputs

(5)

Summary

• Problem formulation • Belgian Context

• A two-step probabilistic approach

• Case study description

• Numerical results

(6)

Belgian context

6

(7)

Belgian context

(8)

8

Belgian context

Marginal and imbalances prices on 08/01/2018

(9)

Summary

• Problem formulation

• Belgian Context

• A two-step probabilistic approach

• Case study description

• Numerical results

(10)

A two-step probabilistic approach

10 NRV transition matrices estimation NRV Historic data ARC table Imbalance prices forecast Assumptions Learning phase Operationnal forecaster Step 1:

computing the NRV state transition probabilities

Step 2:

forecasting the imbalance prices

(11)

Summary

• Problem formulation

• Belgian Context

• A two-step probabilistic approach

• Case study description

• Numerical results

(12)

Case study description: Belgium

12

Dataset: imbalance prices, NRV and ARC table

• Learning Set = 2017 • Validation Set = 2018

We compare to two approaches:

- Deterministic: Multi Layer Perceptron (MLP) - Probabilistic: Gaussian Processes (GP)

Forecasting horizons: from 15 minutes to 6 hours ahead Metrics:

- Probabilistic: CRPS & PLF

(13)

Summary

• Problem formulation

• Belgian Context

• A two-step probabilistic approach

• Case study description

• Numerical results

(14)

Numerical results: one-shot forecast

14

Horizon = 6 hours ahead 10/01/2018, 12h00 UTC

Hours

(15)

Numerical results: last value forecasted

Horizon = 6 hours ahead, 10/01/2018

(16)

Numerical results: scores

16

Mean scores for horizon from 15 to 360 minutes

PLF CRPS NMAE NRMSE % min

(17)

Summary

• Problem formulation

• Belgian Context

• A two-step probabilistic approach

• Case study description

• Numerical results

(18)

Conclusions

18

A first step:

- outperforms other approaches on probabilistic error measures - is less accurate at predicting the precise imbalance prices

Future work:

- add input features to better describe the market situation

(19)

Thank you, any question?

(20)

20 NRV transition matrices estimation NRV Historic data ARC table Imbalance prices forecast Assumptions Learning phase Operationnal forecaster

(21)

NRV transition matrices estimation NRV Historic data ARC table Imbalance prices forecast Assumptions Learning phase Operationnal forecaster 1. NRV forecast

2. Selection of the most

probable MIP/MDP into the ARC

table.

(22)

22 2018-01 2017-12 LS VS 2018-01 2018-01 LS VS 2018-12 2018-11 LS VS 2018-01 2017 LS VS 2018-02 2017 -> 2018-01 LS VS 2018-12 2017 -> 2018-11 LS VS MLP & Two-Step Probabilistic Approach GP

Références

Télécharger maintenant ( PDF - 22 Page - 1.21 MB )

Documents relatifs

Some improvements of the spectral learning approach for probabilistic grammatical inference

Keywords: Probabilistic grammatical inference, Rational series, Machine Learning, Spec- tral learning, Structured low-rank approximation, Maximum-Likelihood, Baum-Welch Al-

The Effect of Biofuels on the Link between Oil and Agricultural Commodity Prices: A Smooth Transition Cointegration Approach

Relying on a smooth transition cointegration approach, we show that biofuel development has led to an increase in the long-term price effect of oil on agricultural

Probabilistic Forecasting of Regional Wind Power Generation for the EEM20 Competition: a Physics-oriented Machine Learning Approach

• A physics-oriented machine learning model for day- ahead direct forecasting of regional wind production is proposed: a QRF model learns from a combination of

Short-range probabilistic forecasting of convective risks for aviation based on a lagged-average-forecast ensemble approach

Figure 6: As in Figure 5, this time for the reflectivity maximum in the column for 30, 27, 18 and 15 member blended ensemble for 1, 2, 5 and 6 h forecasts based on

Forecasting Cryptocurrency Prices Time Series Using Machine Learning

The modified model of Binary Auto Regressive Tree (BART) is adapted from the standard models of regression trees and the data of the time series.. BART

On the existence of shadow prices

Abstract. For utility maximization problems under proportional transaction costs, it has been observed that the original market with transaction costs can sometimes be replaced by

A probabilistic approach to the estimation of regional photovoltaic power production

As previously mentioned, one issue with the upscaling method is that the meteorological fields are only assessed by a limited number of points (set of reference PV plants). This

A Tale of Two Diagnoses in Probabilistic Systems

In order to formalise problems related to fault diagnosis, we partition the set of events Σ into two disjoint sets Σ o and Σ u , the sets of observable and unobservable