Accurate Short-term Forecasting Research Articles

UniLF: A novel short-term load forecasting model uniformly considering various features from multivariate load data.

Accurate short-term load forecasting (STLF) provides important support for the economic and stable operation of the power system. Although various deep learning methods have achieved good results in STLF, they usually model load features only from a limited perspective, i.e., they do not uniformly utilize the three features of multivariate load data: the influence of covariates, multiscale features and local-global variations. The insufficient mining of these three features limits the improvement of prediction accuracy. To address the above problems, we design a novel STLF model called UniLF based on Transformer framework, which contains the proposed convolutional enhancement-fusion embedding method to capture the correlations between load and covariates for embedding, the proposed feature reconstruction-decomposition block to distill multiscale features as well as more detailed local-global variations from 2D space and the core mask-guided multiscale interactive self-attention mechanism to further realize the enhanced interactions of scale features and temporal features. Experiments conducted on three load datasets from Australia, Panama and Austria show that UniLF achieves superior forecasting accuracy with competitive practical efficiency under different prediction lengths, providing a new solution for STLF.

Explorative Short-Term Predictive Models for the Belgian (Energy) Renovation Market Incorporating Macroeconomic and Sector-Specific Variables

Retrofitting existing buildings is a cornerstone of Europe’s strategy for a sustainable built environment. Therefore, accurate short-term forecasts to evaluate policy impacts and inform future strategies are needed. This study investigates the short-term predictive modelling of renovation activity in Belgium, focusing on overall renovation activity (RA) and energy-specific renovation activity (EA). Using data from 2012 to 2023, linear modelling was employed to analyze the relationships between RA/EA and macroeconomic indicators, market confidence, building permits, and loan data, with model performance evaluated using Mean Absolute Percentage Error (MAPE). Monthly data and time lags of up to 24 months were considered. The three best-performing models for RA achieved MAPE values between 2.9% and 3.1%, with validated errors ranging from 0.1% to 4.1%. For EA, the best models yielded MAPE values between 4.4% and 4.6% and validated errors between 8.9% and 14%. Renovation loans and building permits emerged as strong predictors for RA, while building material prices and loans were more relevant for EA. The time lag analysis highlighted that renovation processes typically span 15–24 months following loan approval. However, the low accuracy observed for EA underscores the need for further refinement. This explorative effort forms a solid base, inviting additional research to enhance our predictive capabilities and improve short-term modelling of the (green) residential renovation market.

Sub-hourly forecasting of fire potential using machine learning on time series of surface weather variables

Background Rapidly developing pre-fire weather conditions contributing to sudden fire outbreaks can have devastating consequences. Accurate short-term forecasting is important for timely evacuations and effective fire suppression measures. Aims This study aims to introduce a novel machine learning-based approach for forecasting fire potential and to test its performance in the Sunshine Coast region of Queensland, Australia, over a period of 15 years from 2002 to 2017. Methods By analysing real-time data from local weather stations at a sub-hourly temporal resolution, we aimed to identify distinct weather patterns occurring hours to days before fires. We trained random forest machine learning models to classify pre-fire conditions. Key results The models achieved high out-of-sample accuracy, with a 47% higher accuracy than the standard fire danger index for the region. When simulating real forecasting conditions, the model anticipated 75% of the fires (11 out of 15). Conclusions This method provides objective, quantifiable information, enhancing the precision and effectiveness of fire warning systems. Implications The proposed forecasting approach supports decision-makers in implementing timely evacuations and effective fire suppression measures, ultimately reducing the impact of fires.

Modeling and Evaluation of Forecasting Models for Energy Production in Wind and Photovoltaic Systems

The comprehensive change from known, classical energy production methods to the increased use of renewable energy requires new methods in the field of efficient application and use of renewable energy. The urban energy supply presents complex challenges in improving efficiency; therefore, the prediction of the dynamical availability of energy is required. Several approaches have been explored, including statistical models and machine learning using historical data and numerical weather prediction models using mathematical models of the atmosphere and weather conditions. Accurately forecasting renewable energy production involves analyzing factors such as related weather conditions, conversion systems, and their locations, which influence both energy availability and yield. This study focuses on the short-term forecasting of wind and photovoltaic (PV) energy using historical data and machine learning approaches, aiming for accurate 8 h predictions. The goal is to develop models capable of producing accurate short-term forecasts of energy production from both resources (solar and wind), suitable for later use in a model predictive control scheme where generation and demand, as well as storage, must be considered together. Methods include regression trees, support vector regression, and regression neural networks. The main idea in this work is to use past and future information in the model. Inputs for the PV model are past PV generation and future solar irradiance, while the wind model uses past wind generation and future wind speed data. The performance of the model is evaluated over the entire year. Two scenarios are tested: one with perfect future predictions of wind speed and solar irradiance, and another considered realistic situation where perfect future prediction is not possible, and uncertain prediction is accounted for by incorporating noise models. The results of the second scenario were further improved using the output filtering method. This study shows the advantages and disadvantages of different methods, as well as the accuracy that can be expected in principle. The results show that the regression neural network has the best performance in predicting PV and wind generation compared to other methods, with an RMSE of 0.1809 for PV and 5.3154 for wind, and a Pearson coefficient of 0.9455 for PV and 0.9632 for wind.

An optimized method for short-term load forecasting based on feature fusion and ConvLSTM-3D neural network

As renewable energy continues to penetrate modern power systems, accurate short-term load forecasting is crucial for optimizing power generation resource allocation and reducing operational costs. Traditional forecasting methods often overlook key factors such as holiday load variations and differences in user electricity consumption behavior, resulting in reduced accuracy. To address this, we propose an optimized short-term load forecasting method based on time and weather-fused features using a ConvLSTM-3D neural network. The Prophet algorithm is first employed to decompose historical electricity load data, extracting feature components related to time variables. Simultaneously, the SHAP algorithm filters weather variables to identify highly correlated weather features. A time attention mechanism is then applied to fuse these features based on their correlation weights, enhancing their impact within the time series. Finally, the ConvLSTM-3D model is trained on the fused features to generate short-term load forecasts. A case study using real-world data validates the proposed method, demonstrating significant improvements in forecasting accuracy.

Hybrid CNN-Multivariate LSTM Model for Accurate Short-Term Electricity Price Forecasting and Energy System Optimization

Hybrid CNN-Multivariate LSTM Model for Accurate Short-Term Electricity Price Forecasting and Energy System Optimization

Short-Term Load Forecasting Model Based on Time Series Clustering and Transformer in Smart Grid

Accurate Short-Term Load Forecasting (STLF) is a critical task in managing and operating smart grids. Existing STLF methods primarily rely on mathematical modeling or neural networks, often struggle to effectively capture the correlations between influencing factors and load data, and frequently lack interpretability. To address these challenges, this paper proposes an intelligent framework for STLF that combines a pattern extraction and attention mechanism, which leverages the characteristics of electricity consumption data. The proposed framework facilitates the integration of prior knowledge, identifies intrinsic data patterns, and more accurately maps the relationships between influencing factors and load patterns. Finally, we conduct experiments on real-world and publicly available datasets to evaluate the performance of the proposed model. Specifically, the proposed model improves the accuracy of STLF compared to that of existing methods and reduces the mean absolute percentage error by 2% to 5%. The model performs superiorly on the real datasets, with root mean squared error and mean absolute percentage error values of 0.810 MWh and 7.09%.

Research on short-term precipitation forecasting method based on CEEMDAN-GRU algorithm

Precipitation forecasting is vital for managing disasters, urban traffic, and agriculture. This study develops an improved model for short-term precipitation forecasting by combining Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and Gated Recurrent Unit (GRU). Using precipitation data from January 1, 2019, to December 31, 2022, as a sample, the model capitalizes on CEEMDAN’s superior signal decomposition capabilities and GRU’s ability to capture nonlinear dynamic patterns in time series. To assess the model’s effectiveness, comparisons were conducted with 12 benchmark models, including CEEMDAN-LSTM, EMD-GRU, EMD-LSTM, BI-LSTM, GRU, LSTM, and TCN. The results demonstrate that the CEEMDAN-GRU model achieves higher accuracy and stability in short-term precipitation forecasting. Leveraging an Adam optimizer with adaptive learning rate reduction enhances convergence and ensures reliable predictions, achieving an R²of 0.7915, MAE of 0.05382, and MSE of 0.09081.

Electric Vehicle Charging Load Forecasting Based on K-Means++-GRU-KSVR

The accurate short-term forecasting of an electric vehicle (EV) load is crucial for the reliable operation of a power grid and for effectively reducing energy consumption. Due to the fluctuations in EV charging loads, particularly the significant load variation between commercial and non-commercial areas, global models often suffer from prediction errors when forecasting loads. To address this issue, this paper proposes a regional forecasting method based on K-means++ clustering and deep learning algorithms. First, the K-means++ algorithm was used to partition the data into different regions, and an independent load-forecasting model was established for each region. Then, a combination of kernel support vector regression (KSVR) and gated recurrent unit (GRU) models was used to handle nonlinear features and time-dependent data, where particle swarm optimization (PSO) further optimized the model parameters to improve the forecasting accuracy. Finally, a weighted summation method was used to integrate the forecast results from each region, resulting in a more accurate overall load forecast. The experimental results show that the proposed model provided better prediction performance by capturing the spatiotemporal characteristics of the EV charging load, effectively addressing the challenges posed by regional differences, and outperforming the single-model forecasts.

Sky-Image-Based Sun-Blocking Index and PredRNN++ for Accurate Short-Term Solar Irradiance Forecasting

Sky-Image-Based Sun-Blocking Index and PredRNN++ for Accurate Short-Term Solar Irradiance Forecasting

A hybrid K-means and KNN approach for enhanced short-term load forecasting incorporating holiday effects

Accurate short-term load forecasting (STLF) is essential for efficient power system operations, particularly during holidays when demand patterns vary significantly. Traditional methods often struggle with these non-linear changes, leading to increased operational costs. This research introduces a hybrid model combining K-means clustering and K-nearest neighbors (KNN) classification to enhance forecast accuracy by accounting for holiday effects. Using historical electricity consumption data, the model applies K-means to classify load patterns and KNN to predict load clusters based on the day of the week, month, and holiday status. Advanced forecasting algorithms for each cluster, including linear regression, neural networks with Bayesian optimization, Support Vector Regression with Bayesian Optimization, and long short-term memory with Bayesian optimization, were evaluated using mean absolute percentage error (MAPE), ANOVA, and Tukey's HSD test. Results show significant improvements in forecast precision, especially during holidays, with the LSTM-BO model achieving the lowest MAPE values. The proposed hybrid model achieved an average 56.1 % improvement in forecasting accuracy across various scenarios, providing a reliable tool for power system planners and operators to optimize performance and reduce costs.

Photovoltaic output prediction based on VMD disturbance feature extraction and WaveNet

Traditional photovoltaic (PV) forecasting methods often overlook the impact of the correlation between different power fluctuations and weather factors on short-term forecasting accuracy. To address this, this paper proposes a PV output forecasting method based on Variational Mode Decomposition (VMD) disturbance feature extraction and the WaveNet model. First, to extract different feature variations of the output and enhance the model’s ability to capture PV power fluctuation details, VMD is used to decompose the PV output time series, obtaining IMFs modes representing output disturbances and quasi-clear sky IMF modes. Then, to reveal power changes, especially the underlying patterns of disturbances and their relationship with weather factors, K-means clustering is applied to the IMF modes representing output disturbances, clustering the disturbance IMFs into different power change feature clusters. This is combined with Spearman correlation analysis of weather factors and the construction of an experimental dataset. Finally, to enhance the model’s learning ability and improve short-term output forecasting accuracy, the WaveNet model is employed during the forecasting phase. Separate WaveNet models are constructed and trained with the corresponding datasets, and the total PV output forecast is obtained by superimposing the predictions of different IMF modes. Experimental results are compared with traditional methods, demonstrating a significant improvement in forecasting accuracy, with a Mean Absolute Percentage Error (MAPE) error of 6.94%, highlighting the effectiveness of our method and providing strong technical support for the refined management and intelligent forecasting of PV energy.

Impact of WRF Model Parameterization Settings on the Quality of Short-Term Weather Forecasts over Poland

This research examines the impact of various parameterization settings within the Weather Research and Forecasting (WRF) model on the accuracy of short-term weather forecasts for Poland. The study focuses on the sensitivity of key meteorological variables—namely, air temperature, wind speed, relative humidity, and atmospheric pressure—to different combinations of physical parameterization schemes. Utilizing data from the Global Forecast System (GFS) spanning 2019 to 2022, a series of model simulations were conducted with support from the Poznań Supercomputing and Networking Center (PCSS). To assess the model’s performance across different weather stations, statistical metrics such as the mean absolute error (MAE) and root mean square error (RMSE) were employed. The findings indicate that the configuration labeled “p2” produced the most accurate forecasts for temperature, wind speed, and atmospheric pressure, achieving MAE values of 1.5 °C, 1.6 m/s, and 2 hPa, respectively. However, forecast inaccuracies were notably higher in mountainous regions, particularly regarding wind speed. These results underscore the importance of selecting appropriate parameterization settings tailored to regional characteristics, as different configurations can significantly impact the forecast accuracy, especially in complex terrains. This study contributes to the understanding of short-term weather forecasting models for Central Europe, offering potential pathways for improving localized forecast accuracy.

Short-term wave forecasting for offshore wind energy in the Baltic Sea

The increasing interest in ocean-based renewable energy sources and the consequent need for advanced marine infrastructure have heightened the demand for accurate short-term forecasting of significant wave height (SWH). Given the impending investments in offshore infrastructure within the Baltic Sea, it becomes imperative to devise artificial intelligence strategies that can accurately predict wave parameters under the unique conditions of a shallow, tideless sea. This study presents an extensive evaluation of established machine learning techniques tailored to site specific wave spectra. We explore the effectiveness of artificial neural networks (ANNs) in the short-term prediction of SWH. Rigorous optimization of neural network hyperparameters is applied to improve forecast accuracy, with models validated on real-world datasets, proving their ability to predict SWH accurately across various forecast horizons. In addition, the performance of ANNs models is compared to optimized traditional forecasting methods, noting significant improvements in both accuracy and reliability. The results underscore the critical role of careful optimization in achieving precise short-term forecasts at targeted locations. They also suggest that even relatively simple models, when fine-tuned, can surpass the performance of more complex approaches commonly found in the literature. Building upon the foundational advancements in ANNs from the early 2000s, this research adopts a direct forecasting methodology utilizing data from marine observational platforms. Significantly, the networks are trained and validated with data directly obtained from a measurement buoy positioned in a deep water zone, devoid of seabed wave interactions. This aspect is particularly crucial for the strategic placement of future floating offshore farms, where these algorithms will be directly applied for short-term SWH predictions. Our findings underscore the importance of selecting optimal methods to ensure the utmost accuracy of short-term forecasts at targeted locations.

DRN-DSA: A hybrid deep learning network model for precipitation nowcasting using time series data

Precipitation nowcasting involves short-term weather forecasting, predicting rain or snow within the next two hours. By analyzing current atmospheric conditions, it aids meteorologists in identifying weather patterns and preparing for severe events such as flooding. These nowcasts are typically displayed on geographical maps by weather services. However, the rapidly changing climate conditions make precipitation nowcasting a formidable challenge, as accurate short-term forecasts are hindered by immediate weather fluctuations. Traditional nowcasting methods, like numerical models and radar extrapolation, have limitations in delivering highly detailed and timely precipitation nowcasts. To overcome this issue, an effective solution is framed for precipitation nowcasting using a hybrid network approach named Deep Residual Network-Deep Stacked Autoencoder (DRN-DSA). Initially, the input time series data is acquired from the dataset. Thereafter, the effective technical indicators are extracted at the feature extraction stage. Later on, precipitation-type nowcasting is carried out using the proposed hybrid DRN-DSA, which is developed by incorporating a Deep Stacked Autoencoder (DSA) and Deep Residual Network (DRN). Finally, Weather nowcasting is carried out using the same proposed hybrid DSA-DRN. Moreover, when compared to other traditional models, the proposed DRN-DSA has gained superior results with a Relative Absolute Error (RAE) of 0.295, Root Mean Square Error (RMSE) of 0.154, low Mean Square Error (MSE) of 0.0236, Mean Absolute Percentage Error (MAPE) of 0.295, and False Acceptance Rate (FAR) of 0.0118.

Forecasting age distribution of life-table death counts via α-transformation

We introduce a compositional power transformation, known as an α-transformation, to model and forecast a time series of life-table death counts, possibly with zero counts observed at older ages. As a generalisation of the isometric log-ratio transformation (i.e. α = 0 ), the α transformation relies on the tuning parameter α, which can be determined in a data-driven manner. Using the Australian age-specific period life-table death counts from 1921 to 2020, the α transformation can produce more accurate short-term point and interval forecasts than the log-ratio transformation. The improved forecast accuracy of life-table death counts is of great importance to demographers and government planners for estimating survival probabilities and life expectancy and actuaries for determining annuity prices and reserves for various initial ages and maturity terms.

Short-term power load forecast using OOA optimized bidirectional long short-term memory network with spectral attention for the frequency domain

Accurate short-term power load forecasting is crucial for maintaining the stability and efficiency of modern power systems, especially in the face of increasing volatility and complexity. This paper introduces a novel approach for short-term power load forecasting by integrating the Osprey Optimization Algorithm (OOA) with a Bidirectional Long Short-Term Memory (BiLSTM) network enhanced by a spectral attention mechanism. The OOA optimizes the hyperparameters of the BiLSTM, improving the model's global search capability and avoiding local optima. The spectral attention mechanism further refines the model by focusing on key frequency components in the time series data. Experimental results demonstrate the superior performance of the proposed model, achieving an RMSE of 0.1382, MAE of 0.0659, and MAPE of 1.14 %, significantly outperforming traditional methods. The model's effectiveness is particularly notable in capturing complex patterns during critical peak and trough periods, making it a valuable tool for enhancing grid operation and management.

A novel short-term multi-energy load forecasting method for integrated energy system based on two-layer joint modal decomposition and dynamic optimal ensemble learning

Accurate short-term multi-energy load forecasting is the cornerstone for optimal dispatch and stable operation of integrated energy system (IES). However, due to the complexity and coupling inside IES, multi-energy load forecasting faces serious challenges with data nonlinearity and instability, leading to reduced prediction accuracy. To this end, a novel short-term multi-energy load forecasting method for IES based on two-layer joint modal decomposition (TLJMD) and dynamic optimal ensemble (DOE) learning is developed in this paper. Firstly, the TLJMD method is proposed to decompose the nonlinear and nonstationary multi-energy load into several intrinsic mode functions (IMFs) to capture the periodicity and regularity within the multi-energy load. Secondly, the uniform information coefficient method is employed to select calendar, meteorological, and coupling feature that exhibit strong correlation with the multi-energy load. Eventually, the DOE model consisting of four base learners and the ensemble weight forecasting model is constructed, the IMFs and selected features are input into the DOE model to achieve the final forecasting results. The proposed method is tested on the publicly available data set from real-world scenario and compared with various forecasting methods to assess its effectiveness and accuracy. The simulation results indicate that the proposed method outperforms other forecasting methods in short-term multi-energy load forecasting for IES, with mean absolute percentage error values of 1.7025 %, 2.2244 %, and 2.3808 % for electric, heating, and cooling load forecasting, respectively.

Optimizing Models and Data Denoising Algorithms for Power Load Forecasting

To handle the data imbalance and inaccurate prediction in power load forecasting, an integrated data denoising power load forecasting method is designed. This method divides data into administrative regions, industries, and load characteristics using a four-step method, extracts periodic features using Fourier transform, and uses Kmeans++ for clustering processing. On this basis, a Transformer model based on an adversarial adaptive mechanism is designed, which aligns the data distribution of the source domain and target domain through a domain discriminator and feature extractor, thereby reducing the impact of domain offset on prediction accuracy. The mean square error of the Fourier transform clustering method used in this study was 0.154, which was lower than other methods and had a better data denoising effect. In load forecasting, the mean square errors of the model in predicting long-term load, short-term load, and real-time load were 0.026, 0.107, and 0.107, respectively, all lower than the values of other comparative models. Therefore, the load forecasting model designed for research has accuracy and stability, and it can provide a foundation for the precise control of urban power systems. The contributions of this study include improving the accuracy and stability of the load forecasting model, which provides the basis for the precise control of urban power systems. The model tracks periodicity, short-term load stochasticity, and high-frequency fluctuations in long-term loads well, and possesses high accuracy in short-term, long-term, and real-time load forecasting.

A short-term load forecasting method based on QRLSTM considering multiple factors analysis

Abstract Complex characteristics of power loads terribly affect the safety of the power grid. Thus, accurate short-term load forecasting is regarded as a solution to mitigate the effect of load complications. In this regard, this work is putting forward a Long Short-Term Memory Network Quantile Regression (LSTMQR) for power load, obtaining the probability density function (PDF). Firstly, the quantile of future power load was obtained using the LSTMQR model. The conditional quantile obtained by the LSTMQR model is used as the input of the Kernel Density Estimation model (KDE) to predict the future probability density distribution of short-term power load at different prediction times. Furthermore, an assessment method is proposed to quantify the effect of different factors on power load to improve prediction accuracy. Finally, a numerical simulation based on actual data in China is established to validate the effectiveness of the proposed forecasting algorithm, indicating the proposed forecasting algorithm can effectively evaluate the relationship between different factors and power load, and probability density prediction can accurately quantify the uncertainty of load information. The proposed algorithm results in a reduction of 37.6% in forecast error compared to the traditional forecasting algorithm.