Intra-day Forecasts Research Articles

Hybrid model for intra-day probabilistic PV power forecast

Renewable generation forecasting has become an essential element of power system management because of the ever-increasing share of renewable power fed into the grid. Solar energy is one of the most common and well-known sources of energy in existing networks. Due to its intermittent and non-linear nature, accurate prediction of solar irradiance is essential for ensuring the reliable integration of photovoltaic (PV) plants with the grid system and for effectively managing supply and demand. The dispatching problem is typically addressed by generating a forecast for a few sample plants and scaling the result of these forecasts by the PV power generation capacity connected to the electric system. This research presents a comprehensive comparison of various methods for predicting PV power generation in fixed and single axis tracking PV systems at two different locations. These methods include time-series statistical regression, parametric physical approaches, machine learning (ML), and ensemble techniques, aimed at intra-day forecasts ranging from 1 to 6 h ahead. The study introduces a novel method for developing a parametric linear model for fixed PV systems and uses model output statistics (MOS) to improve forecast accuracy. Additionally, this study details various optimization algorithms designed to refine model parameters for creating a hybrid model. It also examines key factors influencing PV power forecasts and applies post-processing Kalman filter techniques. The study provides a detailed analysis of deterministic and probabilistic forecasting models, offering not just the most probable power production estimate, but also conveying the associated uncertainty. The results demonstrate a highly skilled ML based hybrid model to forecast PV power for different technologies at two different locations.

Reducing transaction costs using intraday forecasts of limit order book slopes

AbstractMarket participants who need to trade a significant number of securities within a given period can face high transaction costs. In this paper, we document how improvements in intraday liquidity forecasts can help reduce total transaction costs. We compare various approaches for forecasting intraday transaction costs, including autoregressive and machine learning models, using comprehensive ultra‐high‐frequency limit order book data for a sample of NYSE stocks from 2002 to 2012. Our results indicate that improved liquidity forecasts can significantly decrease total transaction costs. Simple models capturing seasonality in market liquidity tend to outperform alternative models.

Optimal Operation Scheduling of Multi-energy Complementary Systems Including Offshore Wind Power

Background: With the rapid development of offshore wind power in China, offshore wind power is accounting for an increasing proportion of the whole installed power generation in China, which put forward higher requirements for operation scheduling of multi-energy complementary systems including offshore wind power. However, existing studies have paid little attention to the complementary output scheduling of multi-energy systems with a high proportion of offshore wind power, this phenomenon is not conducive to the realization of reasonable operation scheduling. Objective: To realize the reasonable operation scheduling of multi-energy systems with a high proportion of offshore wind power. Methods: Firstly, the output characteristics of offshore wind power were analyzed and summarized, and modeling was conducted from the perspectives of time distribution, spatial correlation, volatility and randomness, which would serve as the theoretical basis for subsequent researches. Secondly, the complementary rule among offshore wind power and the multi-energy system was analyzed, and the complementary characteristic index was put forward. Then the safety operation was taken as the constraint, and the two-stage operation model of the integrated energy system was established, which considers the auxiliary service costs. In the first stage, the initial generation scheduling was developed according to the day-ahead forecast values of the offshore wind power; in the second stage, the more accurate intra-day operation scheduling was further achieved by considering the intra-day forecast values of offshore wind power and the day-ahead scheduling results. Costs of rotating reserve capacity could be reduced by correcting the deviation of generating capacity, then the revenue-maximizing scheduling strategies were obtained and the output of the pumped-storage power plants was optimized. Results: The effectiveness of the model proposed in this paper was verified based on the simulation examples. Compared with the traditional scheduling strategy, the scheduling strategy considering auxiliary service cost proposed in this paper can promote the consumption of offshore wind power, reduce the wind curtailment proportion and improve the electricity sales revenues in the end. Conclusion: The research results in this paper provide a reference for the reasonable consumption of multi-energy complementary systems with a high proportion of offshore wind power and improving electricity sales revenues

Intraday probabilistic forecasts of surface solar radiation with cloud scale-dependent autoregressive advection

Solar energy supply is usually highly volatile, limiting its integration into the power grid. Accurate probabilistic intraday forecasts of solar resources are essential to increase the share of photovoltaic (PV) energy in the grid and enable cost-efficient balancing of power demand and supply. Solar PV production mainly depends on downwelling surface solar radiation (SSR). By estimating SSR from geostationary satellites, we can cover large areas with high spatial and temporal resolutions, allowing us to track cloud motion. State-of-the-art probabilistic forecasts of solar resources from satellite imagery account only for the advective motion of clouds. They do not consider the evolution of clouds over time, their growth, and dissipation, even though these are major sources of forecast uncertainty. To address the uncertainty of cloudiness evolution, we present SolarSTEPS, the first optical-flow probabilistic model able to simulate the temporal variability of cloudiness. We demonstrate that forecasting the autocorrelated scale-dependent evolution of cloudiness outperforms state-of-the-art probabilistic advection-based forecasts by 9% in continuous ranked probability score (CRPS). This corresponds to an extension of the forecast lead time by about 45 min at constant CRPS. Our work is motivated by the scale-dependent predictability of cloud growth and decay. We demonstrate that cloudiness is more variable in time at smaller spatial scales than at larger ones. Specifically, we show that the temporal autocorrelation of cloudiness is related to its spatial scale by an inverse power law. We also demonstrate that decomposing cloudiness into multiple spatial scales in the forecasts further improves the forecast skill, reducing the CRPS by 10% and the RMSE by 9%.

Probability prediction of solar irradiance in the tropic using ensemble forecasting

As photovoltaic (PV) power generation systems become more widespread, the instability of electric power grids with PV connection is becoming an issue. For appropriate management of the grids, probability prediction of solar irradiance is proposed. The lagged average forecasting method is used for ensemble forecasting. The 72 h ahead forecasting of solar irradiance is operated in Thailand once a day, and it contains intraday, next-day, and 2-day ahead forecasts. Ensemble forecasting has three ensemble members. The accuracy of intraday forecasting is higher than that of the other members, and it is employed as the most probable value of the forecast. The relation between spreads and forecasting errors is analyzed. From the result, the confidence intervals of the predictions are derived for an arbitrary confidence level. The probability prediction is performed with the most probable value and the confidence intervals. The interval changes its width due to spread changes and captures the observation in it.

Satellite-derived solar radiation for intra-hour and intra-day applications: Biases and uncertainties by season and altitude

Regional to European scale intra-day forecasts of solar resources and photovoltaic (PV) power generation are of key relevance for operators of solar plants and power grids, as well as power traders, who seek to optimize revenues and maintain grid stability by matching power supply and demand. Our study addresses a key prerequisite for such forecasts: the accuracy of intra-hour and intra-day surface solar radiation (SSR) estimates of the HelioMont Meteosat product, a state of the art satellite product particularly suited for the Alpine region. A second satellite product, SARAH-2, as well as longer time scales, are also examined for comparison. The satellite SSR estimates are analyzed based on 136 ground stations across altitudes from 200 m to 3570 m in Switzerland in 2018. In peak daytime periods, the instantaneous satellite SSR deviates from the ground-measured SSR by a mean absolute deviation (MAD) of 110.4 and 99.6 W/m2 for SARAH-2 and HelioMont, respectively. For the daytime SSR, the instantaneous, hourly, and daily-mean MADs amount to 91.7, 81.1, 50.8, and 82.5, 66.7, 42.9 W/m2 for SARAH-2 and HelioMont, respectively. SARAH-2 drastically underestimates instantaneous SSR in winter at altitudes above 1000 m, possibly due to snow cover. HelioMont performs better in this respect, but the remaining substantial, spatially and temporally heterogeneous biases we identify pose a challenge with regard to the use of HelioMont for accurate intra-day PV estimates and forecasts.

Intraday forecast of global horizontal irradiance using optical flow method and long short-term memory model

Short-term global horizontal irradiance (GHI) forecast methodologies are routinely employed to mitigate the instability of photovoltaic power and, more importantly, to secure early participation in the energy auction market. The intra-day forecast model substantially consists of the combination of a cloud motion vector (CMV) derivation and a GHI extraction model. This study utilized optical flow (OF) method to calculate CMV from two subsequent satellite images and predict cloud displacement up to 180 min ahead. Meanwhile, the conventional GHI extraction models (Hammer and the fast all-sky radiation model for solar applications (FARMS) models) were employed to extract future GHI. Moreover, to the best of our knowledge, only a few studies have been dedicated to combining OF method with deep learning model. Hence, this study proposed a novel combination based on OF method and long short-term memory model (LSTM) to enhance temporal horizon and accuracy. After all combinations of OF method and GHI models were simulated and assessed under all-sky conditions, the OF method that combined with LSTM outperformed the OF method combined with conventional models across all time horizons. For root mean square error (RMSE) and forecast skill (FS), the accuracy of OF and LSTM models for all time horizon ranges from 59.64 W/m2 to 120.63 W/m2 and 16.40 % to 54.42 %, respectively. The study findings are expected to encourage the inclusion of satellite data in GHI forecast as well as the optimization of energy management systems that integrate with a photovoltaic system.

Intraday forecasting of OEE through sensor data and machine learning

Key performance indicators are crucial for measuring and managing the performance of industrial processes. They are employed to detect deviations in production, enabling opportunities to improve manufacturing processes within the three dimensions of time, quality, and cost. In this context, the timeliness of information plays a decisive role in the success of measures since delayed information availability can leave decision-makers with no time to react. Usually, key performance indicators are updated on a shift or daily basis and thus only allow a delayed reaction to deviations that have occurred. To tackle this problem, different mixed data sampling and machine learning methods for the intraday monitoring of key performance indicators based on high-frequency sensor and machine data are investigated. By benchmarking the approaches on real-world data from the process industry, the authors show how to obtain models for real-time detection of performance deviations. To enable application for practitioners and academia in the future, a method for implementation and deployment in real-world scenarios is given. The developed methodology offers the construction of high-frequency forecasts for key performance metrics, and therefore accelerates managers’ reactions to associated problems.

Strategic Participation of Active Citizen Energy Communities in Spot Electricity Markets Using Hybrid Forecast Methodologies

The increasing penetrations of distributed renewable generation lead to the need for Citizen Energy Communities. Citizen Energy Communities may be able to be active market players and solve local imbalances. The liberalization of the electricity sector brought wholesale and retail competition as a natural evolution of electricity markets. In retail competition, retailers and communities compete to sign bilateral contracts with consumers. In wholesale competition, producers, retailers and communities can submit bids to spot markets, where the prices are volatile or sign bilateral contracts, to hedge against spot price volatility. To participate in those markets, communities have to rely on risky consumption forecasts, hours ahead of real-time operation. So, as Balance Responsible Parties they may pay penalties for their real-time imbalances. This paper proposes and tests a new strategic bidding process in spot markets for communities of consumers. The strategic bidding process is composed of a forced forecast methodology for day-ahead and short-run trends for intraday forecasts of consumption. This paper also presents a case study where energy communities submit bids to spot markets to satisfy their members using the strategic bidding process. The results show that bidding at short-term markets leads to lower forecast errors than to long and medium-term markets. Better forecast accuracy leads to higher fulfillment of the community programmed dispatch, resulting in lower imbalances and control reserve needs for the power system balance. Furthermore, by being active market players, energy communities may save around 35% in their electrical energy costs when comparing with retail tariffs.

Using time-windowed solar radiation profiles to assess the daily uncertainty of solar thermal electricity production forecasts

This work is aimed at analysing the impact of real-time weather forecasting on the simulation of solar thermal electricity plants. A novel method is presented that allows generating random solar radiation profiles to assess the uncertainty of forecasting solar thermal electricity production. Rather than a weather prediction model, this work proposes a fast and simple method for the generation of synthetic solar radiation data that focuses on the dynamics relevant for electricity production, not requiring long historical data series or meteorological details. Variable window lengths are used to reproduce the growing uncertainty of forecasts with longer time horizons. In addition, a simplified model of solar thermal power plant is described and implemented in MATLAB®. Both the solar profile generator and the plant model are used to simulate massive cases for four representative days with eight forecasting hours based on electricity markets. The results show uncertainty values for day-ahead predictions from 15% to 60% in cloudy days and from 2% to 40% in sunny days, with decreasing ranges for intra-day forecasts.

Interpretable temporal-spatial graph attention network for multi-site PV power forecasting

Accurate forecasting of photovoltaic (PV) and wind production is crucial for the integration of more renewable energy sources into the power grid. To address the limited resolution and costs of methods based on numerical weather predictions (NWP), we take PV production data as main input for forecasting. Since PV power is affected by weather and cloud dynamics, we model spatio-temporal correlations between production data by representing PV systems as nodes of a dynamic graph and embedding production data, geographical information and clear-sky irradiance as signals on that graph. We introduce a new temporal-spatial multi-windows graph attention network (TSM-GAT) for predicting future PV power production. TSM-GAT can adapt to the dynamics of the problem, by learning different graphs over time. It consists of temporal attention with an overlapping-window mechanism that finds the temporal correlations and spatial attention with a multi-window mechanism, which captures different dynamical spatio-temporal correlations for different parts of the forecasting horizon. Thus, it is possible to interpret which PV stations have the most influence when making a prediction for short-, medium- and long-term intra-day forecasts. TSM-GAT outperforms multi-site state-of-the-art models for four to six hours ahead predictions, with average NRMSE 12.4% and 10.5% on a real and synthetic dataset, respectively. Furthermore, it outperforms state-of-the-art models that use NWP as inputs for up to five hours ahead predictions. TSM-GAT yields predicted signals with a closer shape to ground truth than state-of-the-art models, which indicates that it is better at capturing cloud motion and may lead to better generalization capabilities.

Impact of tropical convective conditions on solar irradiance forecasting based on cloud motion vectors

Intra-day forecasts of global horizontal solar irradiance (GHI) are widely produced by displacing existing clouds on a geo-stationary satellite image to their future locations with cloud motion vectors (CMVs) derived from preceding images. The CMV estimation methods assume rigid cloud bodies with advective motion, which performs reasonably well in mid-latitudes but can be strained for tropical and sub-tropical climatic zones during prolonged periods of seasonal convection. We study the impact of the South Asian monsoon time convection on the accuracy of CMV based forecasts by analysing 2 years of forecasts from three commonly used CMV methods—Block-match, Farnebäck (Optical flow) and TV-L1 (Optical flow). Forecasted cloud index (CI) maps of the entire image section are validated against analysis CI maps for the period 2018–2019 for forecast lead times from 0 to 5.5 h. Site-level GHI forecasts are validated against ground measured data from two Baseline Surface Radiation Network stations—Gurgaon (GUR) and Tiruvallur (TIR), located in hot semi-arid and tropical savanna climatic zones respectively. The inter-seasonal variation of forecast accuracy is prominent and a clear link is found between the increase in convection, represented by a decrease in outgoing longwave radiation (OLR), and the decrease in forecast accuracy. The GUR site shows the highest forecast error in the southwest monsoon period and exhibits a steep rise of forecast error with the increase in convection. The highest forecast error occurs in the northeast monsoon period of December in TIR. The impact of convection on the number of erroneous time blocks of predicted photovoltaic production is also studied. Our results provide insights into the contribution of convection to errors in CMV based forecasts and shows that OLR can be used as a feature in future forecasting methods to consider the impact of convection on forecast accuracy.

The reliability of photovoltaic power generation scheduling in seventeen European countries

As photovoltaic (PV) electricity generation is becoming ubiquitous, the ability to forecast solar power becomes crucial for such aspects as economical dispatch, optimal unit commitment and the stability of the grid. Our research investigated the discrepancies between day-ahead and intraday country-specific PV power generation forecasts and the real generation data in the member states of the European Network of Transmission System Operators for Electricity for the year 2021. The goal of the research was, firstly, to offer an insight into the schedule accuracy achieved in the examined countries, secondly, to present the potentials for improvement in each of them (plausible even by the practical application of the methods available in 2021), and thirdly, to show the amount of regulation need resulting from PV electricity production forecasting error, the knowledge of which can help with the estimation of the size of the necessary storage capacities. This is the first work that analyzes the schedule accuracy from a practical perspective by examining the accuracy of the day-ahead and intraday PV forecasts of seventeen European countries, following the latest recommendations for verification. The results found significant differences between the forecast accuracy of different countries, which indicates a huge potential for imbalance reduction in most countries by improving their own algorithms or creating a uniform European PV power generation forecasting system. Overall, the intraday forecasts are less skillful than the day-ahead forecasts in all but one of the countries, which highlights the significance of further application-related studies in the intraday horizon. The aggregate regulatory need resulting from the discrepancies between the forecasted and actually produced amounts of energy of all studied countries is 4 TWh in the positive and 3 TWh in the negative direction, which calls for a large amount of energy storage capacity to provide carbon-free balancing power and reduce the regulation challenges of the increasing PV penetration.

A Forecast Model of City Natural Gas Daily Load Based on Data Mining

Data mining technology is more and more widely used in the daily load forecasting of natural gas systems. It is still difficult to carry out high-precision, timely intraday load forecasting and intraday load dynamic characteristics clustering for natural gas systems. Based on data mining technology, this paper proposes a stable intraday load forecasting method for the natural gas flow state-space model. The load sensitivity under the current operating conditions of the system is obtained by calculation; the sample space of the state space is established through data processing; the partitions under different clustering radii are calculated; and the best intraday load flow is obtained through the state space effectiveness evaluation method. The experimental results show that the model load forecasting accuracy and relative error reached 98.5% and 0.026, respectively, which solved the problem of processing the long-term accumulated historical data of gas intra-day load. At the same time, the amount of data calculation was reduced by 33.6%, which effectively promoted the quantification of intraday load influencing factors and qualitative analysis.

Resilient Predictive Control Coupled with a Worst-Case Scenario Approach for a Distributed-Generation-Rich Power Distribution Grid

In a context of accelerating deployment of distributed generation in power distribution grid, this work proposes an answer to an important and urgent need for better management tools in order to ‘intelligently’ operate these grids and maintain quality of service. To this aim, a model-based predictive control (MPC) strategy is proposed, allowing efficient re-routing of power flows using flexible assets, while respecting operational constraints as well as the voltage constraints prescribed by ENEDIS, the French distribution grid operator. The flexible assets used in the case study—a low-voltage power distribution grid in southern France—are a biogas plant and a water tower. Non-parametric machine-learning-based models, i.e., Gaussian process regression (GPR) models, are developed for intraday forecasting of global horizontal irradiance (GHI), grid load, and water demand, to better anticipate emerging constraints. The forecasts’ quality decreases as the forecast horizon grows longer, but quickly stabilizes around a constant error value. Then, the impact of forecasting errors on the performance of the control strategy is evaluated, revealing a resilient behaviour where little degradation is observed in terms of performance and computation cost. To enhance the strategy’s resilience and minimise voltage overflow, a worst-case scenario approach is proposed for the next time step and its contribution is examined. This is the main contribution of the paper. The purpose of the min–max problem added upstream of the main optimisation problem is to both anticipate and minimise the voltage overshooting resulting from forecasting errors. In this min–max problem, the feasible space defined by the confidence intervals of the forecasts is searched, in order to determine the worst-case scenario in terms of constraint violation, over the next time step. Then, such information is incorporated into the decision-making process of the main optimisation problem. Results show that these incidents are indeed reduced thanks to the min–max problem, both in terms of frequency of their occurrence and the total surface area of overshooting.

Intraday wind power forecasting employing feedback mechanism

The increasing thrust of the world to reduce the dependency on the fossil fuel-based sources has given rise to the high penetration of Renewable Energy Sources (RES). The major RES that are increasing on the large scale throughout the world are solar and wind energy. However, at the grid level, these technologies present a major drawback due to variable energy on account of intermittency and unpredictability. This study presents an intraday forecasting of wind power using the Autoregressive Integrated Moving Average model and wavelet decomposition. The forecasting interval of 15 min has been used for forecasting employing two data sets: first the actual wind power and second the error wind power, generated form NIWE forecasting model. The results were computed for 106 Wavelet-ARIMA cases to obtain the best wavelet model for the wind forecasting problem. The interval of past data blocks for forecasting has also been optimally worked alongside the discussion of the best wavelet for the forecasting problem. The study develops the concept of wavelet decomposition in-depth along with the feedback mechanism that improves the ability of the existing forecasting models with respect to following of actual wind power curve.

Grid balancing challenges illustrated by two European examples: Interactions of electric grids, photovoltaic power generation, energy storage and power generation forecasting

The past decade has not only witnessed a great shift in the energy mix, with electricity gaining a more and more prominent role, but has also seen electricity coming increasingly from renewable energy sources, whose proportion is forecasted to reach 50% by 2030. As the integration of photovoltaic energy cannot be deemed successful without the electricity supply being both sustainable and secure, such far-reaching developments prompt legislations and policy makers, including those of the European Union, to make changes to accommodate not only ever-changing technologies, including energy storage solutions, but also new players in the market. Currently, battery energy storage systems are not used for enhancing the precision of photovoltaic power generation schedules, so actors in the market find it difficult to make well-grounded decisions on the viability of utilizing batteries for such a purpose. The innovative novelty of the procedure presented in this paper is that it is suitable for the planning, analysis and evaluation of the schedule accuracy enhancing impacts of sodium-sulfur and lithium-ion-based battery energy storage systems of various capacities. By the examples of two European Union countries, this article studied the deviations of day-ahead and intraday photovoltaic power generation forecasts from the actual electricity generation of 1000 MWp photovoltaic systems. The two main objectives of the project were to investigate the amount of the photovoltaic balancing needs, based on actual data compared to the day-ahead and intraday forecasts, and to establish to what extent sodium-sulfur and lithium-ion-based battery storage systems can decrease those requirements. Furthermore, by using a model for simulation, the potentials of a number of different energy storage capacities (nominal net storage capacities from 10 MWh to 10,000 MWh) for reducing grid balancing needs were assessed. The practical use of the model is that its application makes it possible to create a dynamic design, analysis and assessment system mechanism that can allow one to establish the multilevel schedule accuracy enhancing impacts of sodium-sulfur and lithium-ion energy storage systems linked to solar power plant systems. Real photovoltaic data from Belgium and Hungary were used to find out how the accuracy of PV power generation forecasts influence the level of the annual utilization of energy storage systems.

Day‐ahead and intra‐day wind power forecasting based on feedback error correction

The major hindrance in the development of large-scale grid integration of wind energy into the power system is the production of intermittent and variable power. A largescale integration requires a forecasting mechanism to support the power system operators while operating the grids. This study forecasts the day-ahead and intra-day wind power using the wavelet decomposition, followed by the autoregressive integrated moving average. The forecasting has been attempted on two datasets, actual and error wind power. The forecasting using the wavelet decomposition employs the discrete wavelets consisting of 16 wavelets, grouped into 5 families. The optimum length of the past data used for the model formulation has also been examined for the various scenarios. The forecasting results have been compared based on various performance metrics. The results of the forecasted values have been compared with the reference ARIMA model to see the effectiveness of the proposed wavelet-ARIMA model. The results indicate that the feedback mechanism used in the error dataset have improved the forecasting efficiency over the use of actual data in both the day-ahead and intra-day forecasting. Also, it has been observed that some of the wavelet families outperformed the other families in terms of accuracy and speed.

A deep learning model for intra-day forecasting of solar irradiance using satellite-based estimations in the vicinity of a PV power plant

This work proposes an intra-day forecasting model, which does not require to be trained or fed with real-time data measurements, for global horizontal irradiance (GHI) at a given location. The proposed model uses a series of time-dependant irradiance estimates near the target location as the main input. These estimates are derived from satellite images and are combined with other secondary inputs in an advanced neural network, which features convolutional and dense layers and is trained using a deep learning approach. For the various input combinations, the performance of the model is validated with a quantitative analysis on the forecast accuracy using different error metrics. Accuracies are compared with a commercial solution for irradiance forecasting made by the European Centre for Medium-Range Weather Forecasts (ECMWF) and publications with similar approaches and forecasting horizons, showing state-of-the-art performance even without irradiance measurements.

Assessing the performance of deep learning models for multivariate probabilistic energy forecasting

Deep learning models have the potential to advance the short-term decision-making of electricity market participants and system operators by capturing the complex dependences and uncertainties of power system operation. Currently, however, the adoption of global deep learning models for multivariate energy forecasting in power systems is far behind the developments in the deep learning research field. In this context, the objectives of this study are to review recent developments in the field of probabilistic, multivariate, and multihorizon time series forecasting and empirically evaluate the performance of novel global deep learning models for forecasting wind and solar generation, electricity load, and wholesale electricity price for intraday and day-ahead time horizons. Two forecast types, deterministic and probabilistic forecasts, are studied. The evaluation data consist of real-world datasets with hourly resolution at the levels of an individual customer and regional and national electricity market bidding zones. The model evaluation criteria include achievable levels of forecasting accuracy and uncertainty risks, hyperparameter sensitivity, the effect of exogenous variables and fieldwise dataset split, and run-time efficiency factors, such as memory utilization, simulation time, electricity consumption, and convergence rate. We conclude that the performance of the global models is more beneficial for intraday forecasts of heterogeneous datasets with nonuniform patterns of time series, but can be affected by the hyperparameter sensitivity and hardware limitations with the growth of dataset dimensionality. The results can serve as a reference point for the quantitative evaluation of deep learning models for probabilistic multivariate energy forecasting in power systems.