• Energy Research

This paper introduces the first completely unsupervised methodology for non-intrusive load monitoring that does not rely on any additional data, making it suitable for real-life applications. The methodology includes an algorithm to efficiently decompose the aggregated energy load from households in events and algorithms based on expert knowledge to assign each of these events to four types of appliances: fridge, dishwasher, microwave, and washer/dryer. The methodology was developed to work with smart meters that have a granularity of 1 min and was evaluated using the Reference Energy Disaggregation Dataset. The results show that the algorithm can disaggregate the refrigerator with high accuracy and the usefulness of the proposed methodology to extract relevant features from other appliances, such as the power use and duration from the heating cycles of a dishwasher.

Pattern sequence-based models are a type of forecasting algorithm that utilizes clustering and other techniques to produce easily interpretable predictions faster than traditional machine learning models. This research focuses on their application in energy demand forecasting and introduces two significant contributions to the field. Firstly, this study evaluates the use of pattern sequence-based models with large datasets. Unlike previous works that use only one dataset or multiple datasets with less than two years of data, this work evaluates the models in three different public datasets, each containing eleven years of data. Secondly, we propose a new pattern sequence-based algorithm that uses a genetic algorithm to optimize the number of clusters alongside all other hyperparameters of the forecasting method, instead of using the Cluster Validity Indices (CVIs) commonly used in previous proposals. The results indicate that neural networks provide more accurate results than any pattern sequence-based algorithm and that our proposed algorithm outperforms other pattern sequence-based algorithms, albeit with a longer training time.

Photovoltaic solar energy is booming due to the continuous improvement in photovoltaic panel efficiency along with a downward trend in production costs. In addition, the European Union is committed to easing the implementation of renewable energy in many companies in order to obtain funding to install their own panels. Nonetheless, the nature of solar energy is intermittent and uncontrollable. This leads us to an uncertain scenario which may cause instability in photovoltaic systems. This research addresses this problem by implementing intelligent models to predict the production of solar energy. Real data from a solar farm in Scotland was utilized in this study. Finally, the models were able to accurately predict the energy to be produced in the next hour using historical information as predictor variables.

The energy supply sector faces significant challenges, such as the ongoing COVID-19 pandemic and the ongoing conflict in Ukraine, which affect the stability and efficiency of the energy system. In this study, we highlight the importance of electricity pricing and the need for accurate models to estimate electricity consumption and prices, with a focus on Spain. Using hourly data, we implemented various machine learning models, including linear regression, random forest, XGBoost, LSTM, and GRU, to forecast electricity consumption and prices. Our findings have important policy implications. Firstly, our study demonstrates the potential of using advanced analytics to enhance the accuracy of electricity price and consumption forecasts, helping policymakers anticipate changes in energy demand and supply and ensure grid stability. Secondly, we emphasize the importance of having access to high-quality data for electricity demand and price modeling. Finally, we provide insights into the strengths and weaknesses of different machine learning algorithms for electricity price and consumption modeling. Our results show that the LSTM and GRU artificial neural networks are the best models for price and consumption modeling with no significant difference.

Renewable energies play an important role in our society’s development, addressing the challenges presented by climate change. Specifically, in countries like Spain, technologies such as solar energy assume a crucial significance, enabling the generation of clean energy. This study addresses the critical need to accurately predict photovoltaic (PV) energy demand in Spain. By using the data collected from the Spanish Electricity System, four models (Linear Regression, Random Forest, Recurrent Neural Network, and LightGBM) were implemented, with adaptations for Big Data. The LR model proved unsuitable, while the LGBM emerged as the most accurate and timely performer. The incorporation of Big Data adaptations amplifies the significance of our findings, highlighting the effectiveness of the LGBM in forecasting PV energy demand with both accuracy and efficiency.

This paper presents the implementation and analysis of two approaches (fuzzy and conventional). Using hourly data from buildings at the University of Granada, we have examined their electricity demand and designed a model to predict energy consumption. Our proposal was conducted with the aid of time series techniques as well as the combination of artificial neural networks and clustering algorithms. Both approaches proved to be suitable for energy modelling although nonfuzzy models provided more variability and less robustness than fuzzy ones. Despite the relatively small difference between fuzzy and nonfuzzy estimates, the results reported in this study show that the fuzzy solution may be useful to enhance and enrich energy predictions.

AbstractIn this paper, we present a new fuzzy symbolization technique for energy load forecasting with neural networks, FPLS-Sym. Symbolization techniques transform a numerical time series into a smaller string of symbols, providing a high-level representation of time series by combining segmentation, aggregation and discretization. The dimensional reduction obtained with symbolization can speed up substantially the time required to train neural networks, however, it can also lead to considerable information losses that could lead to a less accurate forecast. FPLS-Sym introduces the use of fuzzy logic in the discretization process, maintaining more information about each segment of the neural network at the expense of requiring more space in memory. Extensive experimentation was made to evaluate FPLS-Sym with various neural-network-based models, including different neural network architectures and activation functions. The evaluation was done with energy demand data from Spain taken from 2009 to 2019. Results show that FPLS-Sym provides better quality metrics than other symbolization techniques and outperforms the use of the standard numerical time series representation in both quality metrics and training time.

Este artículo aborda el problema de la predicción de la demanda eléctrica utilizando redes neuronales y técnicas de simbolización. Las técnicas de simbolización proporcionan una representación simbólica de una serie temporal de menor longitud que la serie temporal original. En nuestra metodología, incorporamos el uso de codificación de la regresión ordinal, preservando la notación de orden entre los símbolos y realizamos una experimentación extensiva con diferentes arquitecturas de redes neuronales y técnicas de simbolización. En nuestra experimentación, utilizamos los datos de la demanda eléctrica total en la red eléctrica de la península española, tomados desde 2009 hasta 2019 con una granularidad de 10 minutos. El mejor modelo encontrado haciendo uso de la metodología de simbolización nos ofreció métricas de calidad ligeramente peores (1.3655 RMSE y 0.0390 MAPE en lugar de 1.2889 RMSE y 0.0363 MAPE del mejor modelo numérico) pero se entrenó 6826 veces más rápido. This paper addresses the electric demand prediction problem using neural networks and symbolization techniques. Symbolization techniques provide a time series symbolic representation of a lower length than the original time series. In our methodology, we incorporate the use of encoding from ordinal regression, preserving the notation of order between the symbols and make extensive experimentation with different neural network architectures and symbolization techniques. In our experimentation, we used the total electric demand data in the Spanish peninsula electric network, taken from 2009 to 2019 with a granularity of 10 min. The best model found making use of the symbolization methodology offered us slightly worse quality metrics (1.3655 RMSE and 0.0390 MAPE instead of the 1.2889 RMSE and 0.0363 MAPE from the best numerical model) but it was trained 6826 times faster. PID2020-112495RB-C21 B-TIC-42-UGR20