Search

In this paper, risk-constrained arbitrage trading strategies that exploit price differences arising across short-term electricity markets, namely day-ahead (DAM), continuous intraday (CID) and balancing (BAL) markets, are developed and evaluated. To open initial DAM positions, a rule-based trading policy using DAM and CID price forecasts is proposed. DAM prices are predicted using both technical indicator features and data augmentation methods, such as autoencoders and generative adversarial networks. Meanwhile, CID prices are predicted using novel features that are engineered from the limit order book. Using the forecasts, the direction of price movements is correctly predicted the majority of the time. To manage open DAM positions while optimising the risk-reward ratio, deep reinforcement learning agents trained using the advantage actor–critic algorithm (A2C) are employed. Evaluated across Dutch short-term markets, A2C yields profits surpassing those obtained using A3C and other benchmarks. We expect our study to benefit electricity traders and researchers who seek to develop state-of-art intelligent trading strategies.

High electricity prices have become the title of several news articles recently in Sweden and the prices have experienced large sudden fluctuations during certain periods. In this thesis work, a smart control model for the electricity usage in three different households has been developed with the main purpose to minimize the electricity cost. This has been implemented by using mixed-integer linear programming (MILP) to optimize the cost 24 hours ahead, and by forecasting two of the main inputs; the load and the electricity spot prices for bidding zone three (SE3) in Sweden. The units included in the model are the photovoltaic system, the batteries, the electricity consumption in the house and the electric vehicles. However, the main task of the smart control was to determine when and in which amount the energy should flow from one unit to another, or to/from the grid. In other words, it decides the charging/discharging of the batteries, the selling/buying of electricity and the charging of the electric vehicle (EV). Different amounts of cost savings/profits have been obtained when applying the smart control on the three houses, which have different annual consumption, capacities of the components, heating systems and more. The results showed that it is most optimal to run the model between the time interval 13.00-00.00, when the spot prices for the next day are known, in order to avoid the remarkable impact accompanied with the use of forecasted electricity prices as input to the model. The forecasting of the load is, on the other hand, required to run the model, but this thesis showed that the effect of the uncertainties in this forecast is relatively small. Three types of machine learning methods were implemented to perform the forecasts, namely linear regression (LR), decision tree regression and random forest regression. After measuring especially the mean absolute error (MAE) to validate the results, the random forest regression showed the least error and the

The importance of electricity in people’s daily lives has made it an indispensable commodity in society. In electricity market, the price of electricity is the most important factor for each of those involved in it, therefore, the prediction of the electricity price has been an essential and very important task for all the agents involved in the purchase and sale of this good. The main problem within the electricity market is that prediction is an arduous and difficult task, due to the large number of factors involved, the non-linearity, non-seasonality and volatility of the price over time. Data Science methods have proven to be a great tool to capture these difficulties and to be able to give a reliable prediction using only price data, i.e., taking the problem from an univariate point of view in order to help market agents. In this work, we have made a comparison among known models in the literature, focusing on Deep Learning architectures by making an extensive tuning of parameters using data from the Spanish electricity market. Three different time periods have been used in order to carry out an extensive comparison among them. The results obtained have shown, on the one hand, that Deep Learning models are quite effective in predicting the price of electricity and, on the other hand, that the different time periods and their particular characteristics directly influence the final results of the model

In this paper we propose a Quantile Regression Deep Neural Network capable of forecasting multiple quantiles in one model using a combined quantile loss function, and apply it to probabilistically forecast the prices of 8 European Day Ahead Markets. We show that the proposed loss function significantly reduces the quantile crossing problem to (near) 0% in all markets considered, while in some cases simultaneously increasing forecasting performance based on classical point forecast metrics applied to the expected value of the probabilistic forecast. The models are optimized using an automated approach with an elaborate feature- and hyperparameter search space, leading to good model performance in all considered markets., Green Open Access added to TU Delft Institutional Repository 'You share, we take care!' - Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public., Water Resources, Intelligent Electrical Power Grids

A model's expected generalisation error is inversely proportional to its training set size. This relationship can pose a problem when modelling multivariate time series, because structural breaks, low sampling rates, and high data gathering costs can severely restrict training set sizes, increasing a model's expected generalisation error by spurring regression model overfitting. Artificially expanding the training set size, using data augmentation methods, can, however, counteract the restrictions imposed by small sample sizes: increasing a model's robustness to overfitting and boosting out-of-sample prediction accuracies. While existing time series augmentation methods have predominantly utilised feature space transformations to artificially expand training set sizes and boost prediction accuracies, we propose using autoencoders (AEs), variational autoencoders (VAEs) and Wasserstein generative adversarial networks with a gradient penalty (WGAN-GPs) for time series augmentation. To evaluate our proposed augmentors, as a case study we forecast Belgian and Dutch day-ahead electricity market prices using both autoregressive models and artificial neural networks. Overall, our results demonstrate that AEs, VAEs, and WGAN-GPs can significantly boost regression accuracies; on average decreasing benchmark model mean absolute errors by 2.23%, 2.73% and 2.97% respectively. Moreover, our results demonstrate that combining AE, VAE, and WGAN-GP generated time series can further boost regression accuracies; on average decreasing benchmark errors by 3.44%. As our proposed augmentors outperform existing augmentation methods, we strongly believe that both practitioners and researchers aiming to generate time series or reduce time series regression errors will find utility in our study.

In this manuscript we explore European feature importance in Day Ahead Market (DAM) price forecasting models, and show that model performance can deteriorate when too many features are included due to over-fitting. We propose a greedy algorithm to search over candidate countries for European features to be used in a DAM price forecasting model, that can be applied to several regression and machine learning modelling methodologies. We apply the algorithm to build price forecasting models for the Dutch market, using candidate countries selected through an integrated analysis based on open-source European electricity market data. Feature importance is visualised using an Auto Regressive and Random Forest model. We explain these results using cross-border flow and DAM price data. Two types of models (LEAR and the Deep Neural Network) are considered for the DAM price forecasting with and without European features. We show that in the Dutch case, taking European market integration into account improves the Mean Absolute Error (MAE) of the best performing DAM price forecasting model by 3.1%, the relative MAE (rMAE) by 3.85%, and the Symmetrical Mean Absolute Percentage Error (sMAPE) by 0.31 p.p., compared to the best forecasting model without European features. Through statistical testing we show that European features improve the accuracy of the forecasts with statistical significance., Water Resources, Intelligent Electrical Power Grids

While the field of electricity price forecasting has benefited from plenty of contributions in the last two decades, it arguably lacks a rigorous approach to evaluating new predictive algorithms. The latter are often compared using unique, not publicly available datasets and across too short and limited to one market test samples. The proposed new methods are rarely benchmarked against well established and well performing simpler models, the accuracy metrics are sometimes inadequate and testing the significance of differences in predictive performance is seldom conducted. Consequently, it is not clear which methods perform well nor what are the best practices when forecasting electricity prices. In this paper, we tackle these issues by comparing state-of-the-art statistical and deep learning methods across multiple years and markets, and by putting forward a set of best practices. In addition, we make available the considered datasets, forecasts of the state-of-the-art models, and a specifically designed python toolbox, so that new algorithms can be rigorously evaluated in future studies., Team DeSchutter

Electricity spot prices are difficult to predict since they depend on different unstable and erratic parameters, and also due to the fact that electricity is a commodity that cannot be stored efficiently. This results in a volatile, highly fluctuating behavior of the prices, with many peaks. Machine learning algorithms have outperformed traditional methods in various areas due to their ability to learn complex patterns. In the last decade, deep learning approaches have been introduced in electricity spot price prediction problems, often exceeding their predecessors. In this thesis, several deep learning models were built and evaluated for their ability to predict the spot prices 10-days ahead. Several conclusions were made. Firstly, it was concluded that rather simple neural network architectures can predict prices with high accuracy, except for the most extreme sudden peaks. Secondly, all the deep networks outperformed the benchmark statistical model. Lastly, the proposed LSTM and CNN provided forecasts which were statistically, significantly superior and had the lowest errors, suggesting they are the most suitable for the prediction task., Elspotspriser är svåra att förutsäga eftersom de beror på olika instabila och oregelbundna faktorer, och också på grund av att elektricitet är en vara som inte kan lagras effektivt. Detta leder till ett volatilt, fluktuerande beteende hos priserna, med många plötsliga toppar. Maskininlärningsalgoritmer har överträffat traditionella metoder inom olika områden på grund av deras förmåga att lära sig komplexa mönster. Under det senaste decenniet har djupinlärningsmetoder introducerats till problem inom elprisprognostisering och ofta visat sig överlägsna sina föregångare. I denna avhandling konstruerades och utvärderades flera djupinlärningsmodeller på deras förmåga att förutsäga spotpriserna 10 dagar framåt. Den första slutsatsen är att relativt simpla nätverksarkitekturer kan förutsäga priser med hög noggrannhet, förutom för fallen med de mest extrema, plötsliga topparna. Vidare, så övertränade alla djupa neurala nätverken den statistiska modellen som användes som riktmärke. Slutligen, så gav de föreslagna LSTM- och CNN-modellerna prognoser som var statistiskt, signifikant överlägsna de andra och hade de lägsta felen, vilket tyder på att de är bäst lämpade för prognostiseringsuppgiften.

Machine learning and agent-based modeling are two popular tools in energy research. In this article, we propose an innovative methodology that combines these methods. For this purpose, we develop an electricity price forecasting technique using artificial neural networks and integrate the novel approach into the established agent-based electricity market simulation model PowerACE. In a case study covering ten interconnected European countries and a time horizon from 2020 until 2050 at hourly resolution, we benchmark the new forecasting approach against a simpler linear regression model as well as a naive forecast. Contrary to most of the related literature, we also evaluate the statistical significance of the superiority of one approach over another by conducting Diebold–Mariano hypothesis tests. Our major results can be summarized as follows. Firstly, in contrast to real-world electricity price forecasts, we find the naive approach to perform very poorly when deployed model-endogenously (mean absolute percentage error 0.40–0.53). Secondly, although the linear regression performs reasonably well (mean absolute percentage error 0.17–0.32), it is outperformed by the neural network approach (mean absolute percentage error 0.17–0.21). Thirdly, the use of an additional classifier for outlier handling substantially improves the forecasting accuracy, particularly for the linear regression approach. Finally, the choice of the model-endogenous forecasting method has a clear impact on simulated electricity prices. This latter finding is particularly crucial since these prices are a major results of electricity market models.

Day-ahead electricity market (DAM) volatility and price forecast errors have grown in recent years. Changing market conditions, epitomised by increasing renewable energy production and rising intraday market trading, have spurred this growth. If forecast accuracies of DAM prices are to improve, new features capable of capturing the effects of technical or fundamental price drivers must be identified. In this paper, we focus on identifying/engineering technical features capable of capturing the behavioural biases of DAM traders. Technical indicators (TIs), such as Bollinger Bands, Momentum indicators, or exponential moving averages, are widely used across financial markets to identify behavioural biases. To date, TIs have never been applied to the forecasting of DAM prices. We demonstrate how the simple inclusion of TI features in DAM forecasting can significantly boost the regression accuracies of machine learning models; reducing the root mean squared errors of linear, ensemble, and deep model forecasts by up to 4.50%, 5.42%, and 4.09%, respectively. Moreover, tailored TIs are identified for each of these models, highlighting the added explanatory power offered by technical features.

Forecasting has been an essential part of the power and energy industry. Researchers and practitioners have contributed thousands of papers on forecasting electricity demand and prices, and renewable generation (e.g., wind and solar power). This article offers a brief review of influential energy forecasting papers; summarizes research trends; discusses importance of reproducible research and points out six valuable open data sources; makes recommendations about publishing high-quality research papers; and offers an outlook into the future of energy forecasting.

[EN] The mainstream of EU policies is heading towards the conversion of the nowadays electricity consumer into the future electricity prosumer (producer and consumer) in markets in which the production of electricity will be more local, renewable and economically efficient. One key component of a local short-term and medium-term planning tool to enable actors to efficiently interact in the electric pool markets is the ability to predict and decide on forecast prices. Given the progressively more important role of renewable production in local markets, we analyze the influence of renewable energy production on the electricity price in the Iberian market through historical records. The dependencies discovered in this analysis will serve to identify the forecasts to use as explanatory variables for an electricity price forecasting model based on recurrent neural networks. The results will show the wide impact of using forecasted renewable energy production in the price forecasting.

In this paper, a novel modeling framework for forecasting electricity prices is proposed. While many predictive models have been already proposed to perform this task, the area of deep learning algorithms remains yet unexplored. To fill this scientific gap, we propose four different deep learning models for predicting electricity prices and we show how they lead to improvements in predictive accuracy. In addition, we also consider that, despite the large number of proposed methods for predicting electricity prices, an extensive benchmark is still missing. To tackle that, we compare and analyze the accuracy of 27 common approaches for electricity price forecasting. Based on the benchmark results, we show how the proposed deep learning models outperform the state-of-the-art methods and obtain results that are statistically significant. Finally, using the same results, we also show that: (i) machine learning methods yield, in general, a better accuracy than statistical models; (ii) moving average terms do not improve the predictive accuracy; (iii) hybrid models do not outperform their simpler counterparts., Erratum: https://doi.org/10.1016/j.apenergy.2018.06.131, Team DeSchutter

Motivated by the increasing integration among electricity markets, in this paper we propose two different methods to incorporate market integration in electricity price forecasting and to improve the predictive performance. First, we propose a deep neural network that considers features from connected markets to improve the predictive accuracy in a local market. To measure the importance of these features, we propose a novel feature selection algorithm that, by using Bayesian optimization and functional analysis of variance, evaluates the effect of the features on the algorithm performance. In addition, using market integration, we propose a second model that, by simultaneously predicting prices from two markets, improves the forecasting accuracy even further. As a case study, we consider the electricity market in Belgium and the improvements in forecasting accuracy when using various French electricity features. We show that the two proposed models lead to improvements that are statistically significant. Particularly, due to market integration, the predictive accuracy is improved from 15.7% to 12.5% sMAPE (symmetric mean absolute percentage error). In addition, we show that the proposed feature selection algorithm is able to perform a correct assessment, i.e. to discard the irrelevant features., Accepted Author Manuscript, Team DeSchutter

[EN] This study develops different univariate and multivariate time series models for spanish electricity load demand and price and compares the predictive accuracy. We have analized the load demand influence on the price forecasting. The models proposed are SARIMA, GARCH, DSARIMA, TF and VAR., [ES] Este trabajo desarrolla diferentes modelos de series temporales univariantes y multivariantes para predecir la demanda y el precio de la energía eléctrica en el mercado español y compara su capacidad predictiva. Se ha estudiado la influencia de la demanda eléctrica en la predicción del precio. Se proponen los modelos SARIMA,GARCH,DSARIMA,TF y VAR.

Clustering is used to generate groupings of data from a large dataset, with the intention of representing the behavior of a system as accurately as possible. In this sense, clustering is applied in this work to extract useful information from the electricity price time series. To be precise, two clustering techniques, K-means and Expectation Maximization, have been utilized for the analysis of the prices curve, demonstrating that the application of these techniques is effective so to split the whole year into different groups of days, according to their prices conduct. Later, this information will be used to predict the price in the short time period. The prices exhibited a remarkable resemblance among days embedded in a same season and can be split into two major kind of clusters: working days and festivities.