Electricity Demand Time Series Research Articles

Textual data for electricity load forecasting

AbstractTraditional mid‐term electricity forecasting models rely on calendar and meteorological information such as temperature and wind speed to achieve high performance. However depending on such variables has drawbacks, as they may not be informative enough during extreme weather. While ubiquitous, textual sources of information are hardly included in prediction algorithms for time series, despite the relevant information they may contain. In this work, we propose to leverage openly accessible weather reports for electricity demand and meteorological time series prediction problems. Our experiments on French and British load data show that the considered textual sources allow to improve overall accuracy of the reference model, particularly during extreme weather events such as storms or abnormal temperatures. Additionally, we apply our approach to the problem of imputation of missing values in meteorological time series, and we show that our text‐based approach beats standard methods. Furthermore, the influence of words on the time series' predictions can be interpreted for the considered encoding schemes of the text, leading to a greater confidence in our results.

Short-Term Load Forecasting Based on Optimized Random Forest and Optimal Feature Selection

Short-term load forecasting (STLF) plays a vital role in ensuring the safe, efficient, and economical operation of power systems. Accurate load forecasting provides numerous benefits for power suppliers, such as cost reduction, increased reliability, and informed decision-making. However, STLF is a complex task due to various factors, including non-linear trends, multiple seasonality, variable variance, and significant random interruptions in electricity demand time series. To address these challenges, advanced techniques and models are required. This study focuses on the development of an efficient short-term power load forecasting model using the random forest (RF) algorithm. RF combines regression trees through bagging and random subspace techniques to improve prediction accuracy and reduce model variability. The algorithm constructs a forest of trees using bootstrap samples and selects random feature subsets at each node to enhance diversity. Hyperparameters such as the number of trees, minimum sample leaf size, and maximum features for each split are tuned to optimize forecasting results. The proposed model was tested using historical hourly load data from four transformer substations supplying different campus areas of the University of Beira Interior, Portugal. The training data were from January 2018 to December 2021, while the data from 2022 were used for testing. The results demonstrate the effectiveness of the RF model in forecasting short-term hourly and one day ahead load and its potential to enhance decision-making processes in smart grid operations.

A novel outlier calendrical heterogeneity reconstruction deep learning model for electricity demand forecasting

The development of an accurate electricity demand forecasting model is of paramount importance for promoting global energy efficiency and sustainability. Nonetheless, the presence of outliers and inappropriate model training can result in suboptimal performance. To tackle these challenges, this study explores the potential of Convolutional Neural Network (CNN) and active learning theory as forecasting solutions, offering high efficiency and advantages for long time series. In this study, a hybrid model that combines Isolation Forest (IF), Outlier Reconstruction (OR), CNN and Random Forest (RF) is conducted to mitigate computational complexity and enhance the accuracy of electricity demand forecasting in the presence of outliers. IF is employed to detect outliers in electricity demand time series, while OR is used to reconstruct subsequences based on calendrical heterogeneity for training. CNN is applied for both training and forecasting, and the final output is combined using RF. The effectiveness of the proposed IF-OR-CNN-RF model is validated using electricity data collected from recent sources in Australia at different sampling frequency. The experimental results demonstrate that, in comparison with other popular CNN-based electricity demand forecasting models, IF-OR-CNN-RF model outperforms with significantly improved performance metrics. Specifically, Mean Absolute Error (MAE), Root Mean Squared Error (RMSE) and R-squared values are 77.92, 179.18 and 0.9769 in 5-minute frequency; 162.67, 353.96 and 0.9775 in 10-minute frequency; 841.27, 1374.79 and 0.9622 in 30-minute frequency; 2746.01, 3824.00 and 0.9262 in 60-minute frequency; 9106.08, 12269.04 and 0.8044 in 120-minute frequency. IF-OR-CNN-RF model represents a valuable framework for future electricity demand forecasting, particularly in scenarios involving outliers.

Day-Ahead Electricity Demand Forecasting Using a Novel Decomposition Combination Method

In the present liberalized energy markets, electricity demand forecasting is critical for planning of generation capacity and required resources. An accurate and efficient electricity demand forecast can reduce the risk of power outages and excessive power generation. Avoiding blackouts is crucial for economic growth, and electricity is an essential energy source for industry. Considering these facts, this study presents a detailed analysis of the forecast of hourly electricity demand by comparing novel decomposition methods with several univariate and multivariate time series models. To that end, we use the three proposed decomposition methods to divide the electricity demand time series into the following subseries: a long-run linear trend, a seasonal trend, and a stochastic trend. Next, each subseries is forecast using all conceivable combinations of univariate and multivariate time series models. Finally, the multiple forecasting models are immediately integrated to provide a final one-day-ahead electricity demand forecast. The presented modeling and forecasting technique is implemented for the Nord Pool electricity market’s hourly electricity demand. Three accuracy indicators, a statistical test, and a graphical analysis are used to assess the performance of the proposed decomposition combination forecasting technique. Hence, the forecasting results demonstrate the efficiency and precision of the proposed decomposition combination forecasting technique. In addition, the final best combination model within the proposed forecasting framework is comparatively better than the best models proposed in the literature and standard benchmark models. Finally, we suggest that the decomposition combination forecasting approach developed in this study be employed to handle additional complicated power market forecasting challenges.

C3S Energy: A climate service for the provision of power supply and demand indicators for Europe based on the ERA5 reanalysis and ENTSO‐E data

AbstractThe EU Copernicus Climate Change Service (C3S) has produced an operational climate service, called C3S Energy, designed to enable the energy industry and policymakers to assess the impacts of climate variability and climate change on the energy sector in Europe. The C3S Energy service covers different time horizons, for the past 40 years and the future. It provides time series of electricity demand and supply from wind, solar photovoltaic and hydropower, and can be used for recent trends analysis, seasonal outlooks or the assessment of climate change impacts on energy mixes in the long term. This article introduces this service and the resulting dataset, with a focus on the design and validation of the energy conversion models, based on ENTSO‐E energy data and the ERA5 climate reanalysis. Flexibility and coherence across all countries have been preferred upon models' accuracy. However, the comparison with ENTSO‐E data shows that the models provide plausible energy indicators and, in particular, allow comparing climate variability effects on power demand and generation in a harmonized manner all over Europe.

Standardised indices to monitor energy droughts

To mitigate the effects of climate change, energy systems are becoming increasingly reliant on renewable energy sources. Since these energy sources are typically dependent on the prevailing weather, renewable energy systems are susceptible to shortages during certain weather conditions. As renewable sources become larger contributors to the energy mix, the risks associated with these shortages, referred to as energy droughts, increase. Techniques are therefore required that can help policymakers to understand and mitigate the impacts associated with energy droughts. In this paper, two standardised indices are introduced to monitor droughts in renewable energy systems. The indices incorporate energy demand and renewable energy production, and constitute analogues to the standardised precipitation index (SPI) and standardised precipitation evapotranspiration index (SPEI), two indices regularly employed operationally to monitor meteorological droughts. The indices are straightforward to construct, can be defined on any timescale, and facilitate comparisons between regions with different climates and installed capacities. We demonstrate how the standardised energy indices proposed herein can be used to define renewable energy droughts, and illustrate the practical utility of these indices in an application to reconstructed time series of electricity demand and wind and solar power generation across Europe.

A novel distributed forecasting method based on information fusion and incremental learning for streaming time series

Real-time algorithms have to adapt and adjust to new incoming patterns to provide timely and accurate responses. This paper presents a new distributed forecasting algorithm for streaming time series called StreamWNN. StreamWNN starts with an offline stage in which a forecasting model based on tuples of information fusion is created with historical data. In particular, this model consists of the fusion of patterns composed of past values of the time series with the future values of their k-nearest neighbors. Afterwards, streaming data starts to arrive. The model is incrementally updated in the online stage using a buffer with streaming data that more accurately matches the current model patterns. The model can be updated daily, monthly, quarterly or based on error thresholds. The methodology has been applied to Spanish electricity demand time series providing more accurate results when the model is updated incrementally. The best error results are obtained with the daily update of the model, resulting in an error between 2% and 3.5% depending on the prediction horizon. The model provides better error results than other algorithms.

Analysis and Forecasting of Monthly Electricity Demand Time Series Using Pattern-Based Statistical Methods

This article provides a solution based on statistical methods (ARIMA, ETS, and Prophet) to predict monthly power demand, which approximates the relationship between historical and future demand patterns. The energy demand time series shows seasonal fluctuation cycles, long-term trends, instability, and random noise. In order to simplify the prediction issue, the monthly load time series is represented by an annual cycle pattern, which unifies the data and filters the trends. A simulation study performed on the monthly electricity load time series for 35 European countries confirmed the high accuracy of the proposed models.

A Hybrid Residual Dilated LSTM and Exponential Smoothing Model for Midterm Electric Load Forecasting.

This work presents a hybrid and hierarchical deep learning model for midterm load forecasting. The model combines exponential smoothing (ETS), advanced long short-term memory (LSTM), and ensembling. ETS extracts dynamically the main components of each individual time series and enables the model to learn their representation. Multilayer LSTM is equipped with dilated recurrent skip connections and a spatial shortcut path from lower layers to allow the model to better capture long-term seasonal relationships and ensure more efficient training. A common learning procedure for LSTM and ETS, with a penalized pinball loss, leads to simultaneous optimization of data representation and forecasting performance. In addition, ensembling at three levels ensures a powerful regularization. A simulation study performed on the monthly electricity demand time series for 35 European countries confirmed the high performance of the proposed model and its competitiveness with classical models such as ARIMA and ETS as well as state-of-the-art models based on machine learning.

Temporal aggregation of time series to identify typical hourly electricity system states: A systematic assessment of relevant cluster algorithms

Comprehensive numerical models are pivotal to analyze the decarbonization of electricity systems. However, increasing system complexity and limited computational resources impose restrictions to model-based analyses. One way to reduce computational burden is to use a minimum, yet representative, set of system states for model simulation. These states characterize fluctuating renewable generation and variable demand for electricity prevailing at a certain point in time. A review of possible time series aggregation techniques identifies cluster algorithms as most adequate, with k-Means and the Ward algorithm predominating. However, throughout the surveyed literature, the line of reasoning for the selection of these algorithms remains unclear. To support the electricity system modeling community in selecting an algorithm, this paper devises a systematic multi-stage evaluation approach to compare a large variety of cluster analysis configurations, differing in algorithm, cluster representation, and number of clusters. Results show that electricity demand and renewable energy generation time series can be compressed to below one percent while sustaining global characteristics of the original data. Two potent cluster configurations are identified, confirming k-Means and WARD as being prevalent. Beyond electricity market data, the methodology can be applied to various types of fundamental time-dependent input data.

Electricity demand prediction for sustainable development in Cambodia using recurrent neural networks with ERA5 reanalysis climate variables

Sustainable energy development plays a prominent role in energy planning to maintain natural resources and mitigate the usage of fossil fuels. The atmospheric factor is one of the main influencing factors that changed the electricity consumption’s behavior due to global warming. In this study, the recurrent neural network (RNN) models were developed to examine the effects of 66 climate variables, collected from the European Center for Medium-Range Weather Forecast (ECMWF) ERA5 reanalysis, on power demand in Cambodia. The statistically significant climate variables were filtered by considering the cross-correlation between power demand and each climate variable. Moreover, the wide range of feedback delays was computed from the power demand dataset and was defined using the 95% confidence intervals. The comparison between a nonlinear autoregressive neural network with exogenous inputs (NARX) using historical power demand with the correlated climate variables and a nonlinear autoregressive neural network (NAR) using only historical power demand dataset was made. The various benchmarked models were evaluated and compared for their performances using statistical indices such as normalized root-mean-square error (NMSE) and coefficient of determination (R2). The results showed the NARX model could perform better than the NAR model for predicting electricity demand time-series.

Modeling and Forecasting Electricity Demand and Prices: A Comparison of Alternative Approaches

Electricity demand and price forecasting are key components for the market participants and system operators as precise forecasts are necessary to manage power systems effectively. However, forecasting electricity demand and prices are challenging due to their specific features, such as high frequency, volatility, long trend, nonconstant mean and variance, mean reversion, multiple seasonalities, calendar effects, and spikes/jumps. Thus, the main aim of this study is to propose models that can efficiently forecast electricity demand and prices. To this end, the time series (demand/price) is divided into two components. The first component is considered a deterministic component that includes a trend, yearly, seasonal, and weekly periodicities, calendar effects, and lagged exogenous information and is modeled by parametric and nonparametric approaches. The second component is known as a stochastic (residual) component that is estimated using univariate autoregressive (AR) and multivariate vector autoregressive (VAR) models. The estimation of these models is carried out by four different estimation methods, including ordinary least squares (O), Lasso (L), Ridge (R), and Elastic‐net (E). The proposed modeling scheme is applied to Nordic electricity demand and price time series, and one‐day‐ahead out‐of‐sample forecasts are obtained for a whole year. Besides descriptive statistics, a statistical significance test is also used to evaluate the models’ forecasting accuracy. The results suggest that the proposed methodology effectively forecasts the price and demand for electricity. In addition, the choice of the estimation procedure used for both deterministic and stochastic components has a significant effect on the forecasting results. Furthermore, multivariate vector autoregressive gives superior performance compared to univariate autoregressive models.

Short-Term Load Forecasting in a microgrid environment: Investigating the series-specific and cross-learning forecasting methods

A reliable and accurate load forecasting method is key to successful energy management of smart grids. Due to the non-linear relations in data generating process and data availability issues, load forecasting remains a challenging task. Here, we investigate the application of feed forward artificial neural networks, recurrent neural networks and crosslearning methods for day-ahead and three days-ahead load forecasting. The effectiveness of the proposed methods is evaluated against a statistical benchmark, using multiple accuracy metrics. The test data sets are high resolution multi-seasonal time series of electricity demand of buildings in Belgium, Canada and the UK from private measurements and open access sources. Both FFNN and RNN methods show competitive results on benchmarking datasets. Best method varies depending on the accuracy metric selected. The use of cross-learning in fitting a global RNN model has an improvement on the final accuracy.

An open tool for creating battery-electric vehicle time series from empirical data, emobpy

There is substantial research interest in how future fleets of battery-electric vehicles will interact with the power sector. Various types of energy models are used for respective analyses. They depend on meaningful input parameters, in particular time series of vehicle mobility, driving electricity consumption, grid availability, or grid electricity demand. As the availability of such data is highly limited, we introduce the open-source tool emobpy. Based on mobility statistics, physical properties of battery-electric vehicles, and other customizable assumptions, it derives time series data that can readily be used in a wide range of model applications. For an illustration, we create and characterize 200 vehicle profiles for Germany. Depending on the hour of the day, a fleet of one million vehicles has a median grid availability between 5 and 7 gigawatts, as vehicles are parking most of the time. Four exemplary grid electricity demand time series illustrate the smoothing effect of balanced charging strategies.

N-BEATS neural network for mid-term electricity load forecasting

This paper addresses the mid-term electricity load forecasting problem. Solving this problem is necessary for power system operation and planning as well as for negotiating forward contracts in deregulated energy markets. We show that our proposed deep neural network modeling approach based on the deep neural architecture is effective at solving the mid-term electricity load forecasting problem. Proposed neural network has high expressive power to solve non-linear stochastic forecasting problems with time series including trends, seasonality and significant random fluctuations. At the same time, it is simple to implement and train, it does not require signal preprocessing, and it is equipped with a forecast bias reduction mechanism. We compare our approach against ten baseline methods, including classical statistical methods, machine learning and hybrid approaches, on 35 monthly electricity demand time series for European countries. The empirical study shows that proposed neural network clearly outperforms all competitors in terms of both accuracy and forecast bias. Code is available here: https://github.com/boreshkinai/nbeats-midterm.

Impact of Covid-19 pandemic on electricity demand in the UK based on multivariate time series forecasting with Bidirectional Long Short Term Memory

Due to lockdown measures taken by the UK government during the Coronavirus disease 2019 pandemic, the national electricity demand profile presented a notably different performance. The Coronavirus disease 2019 crisis has provided a unique opportunity to investigate how such a landscape-scale lockdown can influence the national electricity system. However, the impacts of social and economic restrictions on daily electricity demands are still poorly understood. This paper investigated how the UK-wide electricity demand was influenced during the Coronavirus disease 2019 crisis based on multivariate time series forecasting with Bidirectional Long Short Term Memory, to comprehend its correlations with containment measures, weather conditions, and renewable energy supplies. A deep-learning-based predictive model was established for daily electricity demand time series forecasting, which was trained by multiple features, including the number of coronavirus tests (smoothed), wind speed, ambient temperature, biomass, solar & wind power supplies, and historical electricity demand. Besides, the effects of Coronavirus disease 2019 pandemic on the Net-Zero target of 2050 were also studied through an interlinked approach.

Electricity Demand Time Series Forecasting Based on Empirical Mode Decomposition and Long Short-Term Memory

Load forecasting is critical for a variety of applications in modern energy systems. Nonetheless, forecasting is a difficult task because electricity load profiles are tied with uncertain, non-linear, and non-stationary signals. To address these issues, long short-term memory (LSTM), a machine learning algorithm capable of learning temporal dependencies, has been extensively integrated into load forecasting in recent years. To further increase the effectiveness of using LSTM for demand forecasting, this paper proposes a hybrid prediction model that incorporates LSTM with empirical mode decomposition (EMD). EMD algorithm breaks down a load time-series data into several sub-series called intrinsic mode functions (IMFs). For each of the derived IMFs, a different LSTM model is trained. Finally, the outputs of all the individual LSTM learners are fed to a meta-learner to provide an aggregated output for the energy demand prediction. The suggested methodology is applied to the California ISO dataset to demonstrate its applicability. Additionally, we compare the output of the proposed algorithm to a single LSTM and two state-of-the-art data-driven models, specifically XGBoost, and logistic regression (LR). The proposed hybrid model outperforms single LSTM, LR, and XGBoost by, 35.19%, 54%, and 49.25% for short-term, and 36.3%, 34.04%, 32% for long-term prediction in mean absolute percentage error, respectively.

Classification of daily electric load profiles of non-residential buildings

We investigated clustering techniques on time series of daily electric load profiles of fourteen higher education buildings on the same campus. A k-means algorithm is implemented, and three different methods are compared: time-series features extraction with Manhattan distance and raw time series with Euclidian distance and Dynamic Time Warping. The impact of data characteristics with data collection time-steps and timeframes is studied using a database of more than 6,500 daily electric load profiles. We show that Euclidian distance applied to electric demand time series with three-month timeframes and ten-minute time-step provides the most consistent clustering results. In addition, useful insights are highlighted for non-residential buildings electric demand modeling and forecasting. Two groups of buildings can be distinguished regarding electric load profile patterns. On one hand, teaching, research, libraries, and gymnasium buildings show similar patterns distributed in two clusters corresponding to business days and closing days load profiles. On the other hand, campus office buildings present a larger number of clusters inconsistent with day-type dependent load profiles. A seasonal effect is also observed using six-month and one-year timeframes. Finally, a two-cluster distribution is obtained when aggregating all buildings load profiles.

A bottom-up weather-sensitive residential demand model for developing countries. A case study of Abuja, Nigeria

The frequency of power outages being experienced in Sub-Saharan Africa mean that traditional methods of electricity demand forecasting which rely on directly observed demand data are inadequate for use in projections. Nevertheless, accurate forecasting methods are urgently required to ensure efficient power system operations and expansion planning. To address this gap, we develop a novel method to estimate unsuppressed electricity demand for developing countries. This follows a bottom-up approach based on socioeconomic data and a time-use database developed from a householder survey, which are used to generate household profiles using a Markov Chain approach. These profiles are then converted into electrical load time series by a series of appliance models, using reanalysis weather data to accurately represent ambient conditions for the generation of cooling demand profiles. We apply our method to a Nigerian case study, obtaining the first time series of unsuppressed residential electricity demand for the country using the first Time-of-Use Survey (TUS) for Nigerian households. We validate our model outputs using the results of a small-scale residential metering trial, which yielded a correlation coefficient of 0.97, RMSE of 0.04, and percentage error of 6% between measured and model data. This evidences that our method is a credible and practical tool for electrical demand studies in developing countries. Using the model, the forecasted domestic demand for Abuja Electricity Distribution Company ranges between 345 and 575 MW, while that of Nigeria ranges between 3829 and 6605 MW.

A novel hybrid forecasting scheme for electricity demand time series

Electricity demand/load forecasting always plays a vital role in the management and operation of power systems, since it can help develop an optimal action program for power producers, end-consumers and government entities. Inaccurate prediction may cause an additional production or waste of resources due to high operational costs. This paper investigated the benefit of combining data features to produce short-term electricity demand forecast. The nature of the electricity usually presents the complex characteristic and obvious seasonal tendency. In this paper, the advantage of adaptive Fourier decomposition is firstly used to extract the fluctuation characteristics. Then, the condition of the linear and stationary sequence is satisfied and the sub-series are performed to measure and eliminate the seasonal pattern. In the process of seasonal adjustment, the average periodicity length is identified quantitatively. In addition, to realize the generalization performance on real electricity demand data, the sine cosine optimization algorithm is applied to select the penalty and kernel parameters of support vector machine. The empirical study showed that the superior property of the proposed hybrid method profits from the effect of data pretreatment and the findings prove that this hybrid modeling scheme can yield promising prediction results within acceptable computational complexity.