Accurate Short-term Forecasting Research Articles

Rasterized Data Image Processing (RDIP) Techniques for Photovoltaic (PV) Data Cleaning and Application in Power Prediction

Photovoltaic (PV) power generation has attracted widespread interest as a clean and sustainable energy source, with increasing global attention given to renewable energy. However, the operation and monitoring of PV power generation systems often result in large amounts of data containing missing values, outliers, and noise, posing challenges for data analysis and application. Therefore, PV data cleaning plays a crucial role in ensuring data quality, enhancing data availability and reliability. This study proposes a PV data cleaning method based on Rasterized Data Image Processing (RDIP) technology, which integrates rasterization and image processing techniques to select optimal contours and extract essential data. To validate the effectiveness of our method, we conducted comparative experiments using three data cleaning methods, including our RDIP algorithm, the Pearson correlation coefficient interpolation method, and cubic spline interpolation method. Subsequently, the cleaned datasets from these methods were utilized for power prediction using two linear regression models and two neural network models. The experimental results demonstrated that data cleaned using the RDIP algorithm improved the short-term forecast accuracy by approximately 1.0% and 3.7%, respectively, compared to the other two methods, indicating the feasibility and effectiveness of the RDIP approach. However, it is worth noting that the RDIP technique has limitations due to its reliance on integer parameters for grid division, potentially leading to coarse grid divisions. Future research efforts could focus on optimizing the selection of binarization thresholds to achieve better cleaning results and exploring other potential applications of RDIP in PV data analysis.

Optimization of Double Exponential Smoothing Using Particle Swarm Optimization Algorithm in Electricity Load

Electricity load forecasting plays a critical role in ensuring the efficient allocation of resources, maintenance optimization, and uninterrupted power supply. The double exponential smoothing (DES) method is widely used in forecasting time series data due to its adaptability and robustness, particularly in handling linear trends without seasonal patterns. However, determining the optimal value of the alpha parameter in DES is crucial for accurate forecasting results. This study proposes the use of the Particle Swarm Optimization (PSO) algorithm to optimize the alpha parameter in DES for electricity load forecasting. PSO is a computational method that iteratively improves candidate solutions by moving particles in the search space based on simple mathematical formulas. By optimizing the alpha parameter using PSO, we aim to enhance the accuracy of short-term electricity load forecasts. Our results demonstrate that the PSO-optimized DES approach achieves a Mean Absolute Percentage Error (MAPE) of 2.89% and an accuracy of 97.11%, indicating significant improvements in forecasting performance. While the PSO algorithm provides promising results, future research may explore the application of other metaheuristic algorithms, such as the whale or orca algorithms, to further enhance the optimization of DES parameters for electricity load forecasting. This study contributes to the advancement of forecasting techniques in the power industry, facilitating more efficient power generation and distribution planning.

Fitness models provide accurate short-term forecasts of SARS-CoV-2 variant frequency.

Genomic surveillance of pathogen evolution is essential for public health response, treatment strategies, and vaccine development. In the context of SARS-COV-2, multiple models have been developed including Multinomial Logistic Regression (MLR) describing variant frequency growth as well as Fixed Growth Advantage (FGA), Growth Advantage Random Walk (GARW) and Piantham parameterizations describing variant R t . These models provide estimates of variant fitness and can be used to forecast changes in variant frequency. We introduce a framework for evaluating real-time forecasts of variant frequencies, and apply this framework to the evolution of SARS-CoV-2 during 2022 in which multiple new viral variants emerged and rapidly spread through the population. We compare models across representative countries with different intensities of genomic surveillance. Retrospective assessment of model accuracy highlights that most models of variant frequency perform well and are able to produce reasonable forecasts. We find that the simple MLR model provides ~0.6% median absolute error and ~6% mean absolute error when forecasting 30 days out for countries with robust genomic surveillance. We investigate impacts of sequence quantity and quality across countries on forecast accuracy and conduct systematic downsampling to identify that 1000 sequences per week is fully sufficient for accurate short-term forecasts. We conclude that fitness models represent a useful prognostic tool for short-term evolutionary forecasting.

Short-Term Electrical Load Forecasting Using an Enhanced Extreme Learning Machine Based on the Improved Dwarf Mongoose Optimization Algorithm

Accurate short-term electrical load forecasting is crucial for the stable operation of power systems. Given the nonlinear, periodic, and rapidly changing characteristics of short-term power load forecasts, this paper introduces a novel forecasting method employing an Extreme Learning Machine (ELM) enhanced by an improved Dwarf Mongoose Optimization Algorithm (Local escape Dwarf Mongoose Optimization Algorithm, LDMOA). This method addresses the significant prediction errors of conventional ELM models and enhances prediction accuracy. The enhancements to the Dwarf Mongoose Optimization Algorithm include three key modifications: initially, a dynamic backward learning strategy is integrated at the early stages of the algorithm to augment its global search capabilities. Subsequently, a cosine algorithm is employed to locate new food sources, thereby expanding the search scope and avoiding local optima. Lastly, a “madness factor” is added when identifying new sleeping burrows to further widen the search area and effectively circumvent local optima. Comparative analyses using benchmark functions demonstrate the improved algorithm’s superior convergence and stability. In this study, the LDMOA algorithm optimizes the weights and thresholds of the ELM to establish the LDMOA-ELM prediction model. Experimental forecasts utilizing data from China’s 2016 “The Electrician Mathematical Contest in Modeling” demonstrate that the LDMOA-ELM model significantly outperforms the original ELM model in terms of prediction error and accuracy.

Short-term power load forecasting for integrated energy system based on a residual and attentive LSTM-TCN hybrid network

In the context of Integrated Energy System (IES), accurate short-term power demand forecasting is crucial for ensuring system reliability, optimizing operational efficiency through resource allocation, and supporting effective real-time decision-making in energy management. However, achieving high forecasting accuracy faces significant challenges due to the inherent complexity and stochastic nature of IES’s short-term load profiles, resulting from diverse consumption patterns among end-users and the intricate coupling within the network of interconnected energy sources. To address this issue, a dedicated Short-Term Power Load Forecasting (STPLF) framework for IES is proposed, which relies on a newly developed hybrid deep learning architecture. The framework seamlessly combines Long Short-Term Memory (LSTM) with Temporal Convolutional Network (TCN), enhanced by an attention mechanism module. By merging these methodologies, the network leverages the parallel processing prowess of TCN alongside LSTM’s ability to retain long-range temporal information, thus enabling it to dynamically concentrate on relevant sections of time series data. This synergy leads to improved prediction accuracy and broader applicability. Furthermore, the integration of residual connections within the network structure serves to deepen its learning capabilities and enhance overall performance. Ultimately, results from a real case study of a user-level IES demonstrate that the Mean Absolute Percentage Error (MAPE) of the proposed framework on the test set is 2.35%. This error rate is lower than the averages of traditional methods (3.43%) and uncombined single submodules (2.80%).

Research on optimization of improved short-term load composite forecasting model based on AM–CNN–Bi–LSTM

Accurate load prediction is a prerequisite for the design, operation, scheduling, and management of energy systems. In the context of the development of smart grids, the extensive integration of highly volatile distributed energy generation into the power system has brought new challenges to the accuracy, reliability, real-time performance, and intelligence of short-term load forecasting. Therefore, this article proposes a novel short-term power load composite prediction model based on AM–CNN–Bi–LSTM. First, CNN is used to extract relevant feature quantities of power load coupling characteristics. Then, AM is used to evaluate the importance of the feature data, highlighting the features that have a greater impact on the prediction results. Finally, the Bi-LSTM network captures bidirectional temporal information from multiple time steps for prediction. Taking one year of measured data as an example, the error comparison of the prediction results of the composite prediction model overlay shows that compared with other models, the composite prediction model has improved prediction accuracy, feature extraction, generalization ability, and other aspects. The research results improve the accuracy of short-term power load forecasting while providing effective model references for decision-making in power system optimization scheduling, safe operation, and reasonable pricing.

XWaveNet: Enabling uncertainty quantification in short-term ocean wave height forecasts and extreme event prediction

To aid accurate short-term forecast of wave heights and extreme events prediction from buoy data, a deep learning approach is proposed. Wave heights are forecasted up to 72 h in advance from multivariate meteorological and oceanographic data, followed by quantification of uncertainty in the forecasted values, and estimation of exceedance probabilities for extreme event conditions. Using Monte Carlo dropout technique, network regularization and forecast uncertainty assessment were performed. To forecast wave heights, a stacked long short-term memory (LSTM) model was trained on historical buoy data augmented with positionally encoded features. Per comparative evaluation of different models, across the 72 h forecast horizon, positionally encoded temporal attributes resulted in a mean prediction error and forecast uncertainty of only 3 % and 13.5 %, respectively. Contrarily, without temporal feature augmentation mean error was 5 % and uncertainty, 15 %. Therefore, the former model was ensembled with a Random Forest model for exceedance probability estimation. The hybrid model yielded a 21.4 % better separability between the extreme and normal observations in the test data from a nearby buoy in the northern Gulf of Mexico than individual models. All the four extreme events in the test set were correctly predicted with 97 % mean accuracy. Further evaluation of the deep ensemble model on 5 other spatially diverse buoy datasets from the Gulf of Mexico, Atlantic and Pacific coasts, suggested excellent generalizability for short-term wave forecasts and extreme event prediction, with an average maximum error of just 5.2 % across all the 6 buoy test datasets and the entire forecast horizon.

Wind-Speed Multi-Step Forecasting Based on Variational Mode Decomposition, Temporal Convolutional Network, and Transformer Model

Reliable and accurate wind-speed forecasts significantly impact the efficiency of wind power utilization and the safety of power systems. In addressing the performance enhancement of transformer models in short-term wind-speed forecasting, a multi-step prediction model based on variational mode decomposition (VMD), temporal convolutional network (TCN), and a transformer is proposed. Initially, the Dung Beetle Optimizer (DBO) is utilized to optimize VMD for decomposing non-stationary wind-speed series data. Subsequently, the TCN is used to extract features from the input sequences. Finally, the processed data are fed into the transformer model for prediction. The effectiveness of this model is validated by comparison with six other prediction models across three datasets, demonstrating its superior accuracy in short-term wind-speed forecasting. Experimental findings from three distinct datasets reveal that the developed model achieves an average improvement of 52.1% for R2. To the best of our knowledge, this places our model at the leading edge of wind-speed prediction for 8 h and 12 h forecasts, demonstrating MSEs of 1.003 and 0.895, MAEs of 0.754 and 0.665, and RMSEs of 1.001 and 0.946, respectively. Therefore, this research offers significant contributions through a new framework and demonstrates the utility of the transformer in effectively predicting short-term wind speed.

Seasonality of influenza-like illness and short-term forecasting model in Chongqing from 2010 to 2022

BackgroundInfluenza-like illness (ILI) imposes a significant burden on patients, employers and society. However, there is no analysis and prediction at the hospital level in Chongqing. We aimed to characterize the seasonality of ILI, examine age heterogeneity in visits, and predict ILI peaks and assess whether they affect hospital operations.MethodsThe multiplicative decomposition model was employed to decompose the trend and seasonality of ILI, and the Seasonal Auto-Regressive Integrated Moving Average with exogenous factors (SARIMAX) model was used for the trend and short-term prediction of ILI. We used Grid Search and Akaike information criterion (AIC) to calibrate and verify the optimal hyperparameters, and verified the residuals of the multiplicative decomposition and SARIMAX model, which are both white noise.ResultsDuring the 12-year study period, ILI showed a continuous upward trend, peaking in winter (Dec. - Jan.) and a small spike in May-June in the 2–4-year-old high-risk group for severe disease. The mean length of stay (LOS) in ILI peaked around summer (about Aug.), and the LOS in the 0–1 and ≥ 65 years old severely high-risk group was more irregular than the others. We found some anomalies in the predictive analysis of the test set, which were basically consistent with the dynamic zero-COVID policy at the time.ConclusionThe ILI patient visits showed a clear cyclical and seasonal pattern. ILI prevention and control activities can be conducted seasonally on an annual basis, and age heterogeneity should be considered in the health resource planning. Targeted immunization policies are essential to mitigate potential pandemic threats. The SARIMAX model has good short-term forecasting ability and accuracy. It can help explore the epidemiological characteristics of ILI and provide an early warning and decision-making basis for the allocation of medical resources related to ILI visits.

A dynamic multi-model transfer based short-term load forecasting

The integration of renewable energy sources in power systems has resulted in increased complexity in dispatch management, necessitating higher accuracy in short-term load forecasting. Although numerous deep learning models have been developed, their training data requirements and difficulties in effectively combining different model structures remain challenges. This study proposes a multi-model transfer-based approach for dynamic short-term load forecasting to address these issues. The method consists of two phases: (1) conducting correlation analysis to determine transferability between datasets and implementing an adaptive parameter transfer policy to optimize data utilization and training efficiency, and (2) constructing a multi-model cooperation platform to fully leverage the strengths of different models and enhance forecasting accuracy. Experimental case studies demonstrate the effectiveness of the proposed method in improving forecasting accuracy when historical load data is limited. This research contributes to advancing load forecasting techniques by leveraging transfer learning and model cooperation, facilitating efficient power system operations in the context of renewable energy integration.

GRU combined model based on multi-objective optimization for short-term residential load forecasting

Reliable and accurate short-term forecasting of residential load plays an important role in DSM. However, the high uncertainty inherent in single-user loads makes them difficult to forecast accurately. Various traditional methods have been used to address the problem of residential load forecasting. A single load forecast model in the traditional method does not allow for comprehensive learning of data characteristics for residential loads, and utilizing RNNs faces the problem of long-term memory with vanishing or exploding gradients in backpropagation. Therefore, a gated GRU combined model based on multi-objective optimization is proposed to improve the short-term residential load forecasting accuracy in this paper. In order to demonstrate the effectiveness, GRUCC-MOP is first experimentally tested with the unimproved model to verify the model performance and forecasting effectiveness. Secondly the method is evaluated experimentally with other excellent forecasting methods: models such as DBN, LSTM, GRU, EMD-DBN and EMD-MODBN. By comparing simulation experiments, the proposed GRU combined model can get better results in terms of MAPE on January, April, July, and November load data, so this proposed method has better performance than other research methods in short-term residential load forecasting.

Combining travel behavior in metro passenger flow prediction: A smart explainable Stacking-Catboost algorithm

Accurately predicting short-term passenger flow is essential to optimize operation resources and improve transportation services in urban metro systems. However, it has become a challenging problem due to spatial-temporal demand fluctuations and heterogeneous passengers’ travel behavior, i.e., the interaction between the departure and arrival passengers. In this paper, we develop an explainable Stacking-Catboost model for passenger flow prediction combining the passenger’s return probability computation. The model explores several basic ensemble learning models and the best stacking strategy. To better characterize the macroscopic spatiotemporal travel patterns other than the micro individual travel behavior, several relevant variables such as train operation characteristics, the nearby bus stations, and points of interest are considered. Ablation studies are conducted to investigate the utility of each component of the proposed model. An explainable analysis is performed to interpret information from “black box” models and quantify the contribution of each feature. We present a real-world case study conducted in the Beijing metro, demonstrating that our proposed method achieves significant improvements over existing techniques in hourly prediction tasks. Specifically, our approach outperforms the CategoricalBoosting (CatBoost) by up to 11.11 % and the Random Forest (RF) by up to 12.90 %, showcasing the effectiveness of incorporating macroscopic spatiotemporal travel patterns and micro individual travel behavior to enhance short-term forecasting accuracy. The global interpretability of models reveals key factors that impact performance, such as returning passenger flow, historical travel demand, and timetable data. Our results offer valuable insights into short-term forecasting challenges and provide a framework for leveraging explainable artificial intelligence to improve transportation services in urban metro systems.

Short-term Electricity Load Forecasting Based on Improved Seagull Algorithm Optimized Gated Recurrent Unit Neural Network

INTRODUCTION: The complexity of the power network, changes in weather conditions, diverse geographical locations, and holiday activities comprehensively affect the normal operation of power loads. Power load changes have characteristics such as non stationarity, randomness, seasonality, and high volatility. Therefore, how to construct accurate short-term power load forecasting models has become the key to the normal operation and maintenance of power.OBJECTIVES: Accurate short-term power load forecasting helps to arrange power consumption planning, optimize power usage and largely reduce power system losses and operating costs.METHODS: A hybrid decomposition-optimization-integration load forecasting method is proposed to address the problems of low accuracy of current short-term power load forecasting methods.RESULTS: The original power load time series is decomposed using the complete ensemble empirical modal decomposition method, while the correlation of power load influencing factors is analysed using Pearson correlation coefficients. The seagull optimisation algorithm is overcome to fall into local optimality by using the random adaptive non-linear adjustment strategy of manipulated variables and the differential variational Levy flight strategy, which improves the search efficiency of the algorithm. Then, the The gated cyclic unit hidden layer parameters are optimised by the improved seagull optimisation algorithm to construct a short-term electricity load forecasting model.The effectiveness of the proposed method is verified by simulation experimental analysis. The results show that the proposed method has improved the accuracy of the forecasting model.CONCLUSION: The CEEMD method is used to decompose the original load time series, which improves the accuracy of the measurement model. The GRU prediction model based on improved SOA optimization not only has better prediction accuracy than other prediction models, but also consumes the least amount of time compared to other prediction models.

Predictive models for short-term load forecasting in the UK's electrical grid.

There are global efforts to deploy Electric Vehicles (EVs) because of the role they promise to play in energy transition. These efforts underscore the e-mobility paradigm, representing an interplay between renewable energy resources, smart technologies, and networked transportation. However, there are concerns that these initiatives could burden the electricity grid due to increased demand. Hence, the need for accurate short-term load forecasting is pivotal for the efficient planning, operation, and control of the grid and associated power systems. This study presents robust models for forecasting half-hourly and hourly loads in the UK's power system. The work leverages machine learning techniques such as Support Vector Regression (SVR), Artificial Neural Networks (ANN), and Gaussian Process Regression (GPR) to develop robust prediction models using the net imports dataset from 2010 to 2020. The models were evaluated based on metrics like Root Mean Square Error (RMSE), Mean Absolute Prediction Error (MAPE), Mean Absolute Deviation (MAD), and the Correlation of Determination (R2). For half-hourly forecasts, SVR performed best with an R-value of 99.85%, followed closely by GPR and ANN. But, for hourly forecasts, ANN led with an R-value of 99.71%. The findings affirm the reliability and precision of machine learning methods in short-term load forecasting, particularly highlighting the superior accuracy of the SVR model for half-hourly forecasts and the ANN model for hourly forecasts.

Short-term Wind Power Forecasts based on VMD-KPCA and CFSBOA-BiLSTM

Accurate prediction of wind power is of great significance to reduce the impact of large-scale wind power connection and improve the security and stability of power grid. A short-term wind power forecasting method based on VMD-KPCA and CFSBOA-BiLSTM is proposed. Firstly, the key environmental factors restricting the volatility of wind power are fully considered, and the variational mode decomposition method is used to decompose the wind power and meteorological factors. Secondly, the kernel principal component analysis method is used to reduce the dimension and construct the sub-sequence characteristic data set. Finally, the multi-strategy improved butterfly algorithm is used to optimize the BiLSTM network hyperparameters and establish the CFSBOA-BiLSTM prediction model. The simulation analysis uses the measured data of a wind farm in northwest China. The experimental results show that the combined forecasting model can fully mine the hidden information of the data and improve the accuracy of short-term wind power forecasting. Compared with the BiLSTM model, RMSE, MAE and MAPE decreased by 60.37%, 66.6% and 67.59%, respectively.

Solar Irradiance Forecasting with Natural Language Processing of Cloud Observations and Interpretation of Results with Modified Shapley Additive Explanations

Forecasting the generation of solar power plants (SPPs) requires taking into account meteorological parameters that influence the difference between the solar irradiance at the top of the atmosphere calculated with high accuracy and the solar irradiance at the tilted plane of the solar panel on the Earth’s surface. One of the key factors is cloudiness, which can be presented not only as a percentage of the sky area covered by clouds but also many additional parameters, such as the type of clouds, the distribution of clouds across atmospheric layers, and their height. The use of machine learning algorithms to forecast the generation of solar power plants requires retrospective data over a long period and formalising the features; however, retrospective data with detailed information about cloudiness are normally recorded in the natural language format. This paper proposes an algorithm for processing such records to convert them into a binary feature vector. Experiments conducted on data from a real solar power plant showed that this algorithm increases the accuracy of short-term solar irradiance forecasts by 5–15%, depending on the quality metric used. At the same time, adding features makes the model less transparent to the user, which is a significant drawback from the point of view of explainable artificial intelligence. Therefore, the paper uses an additive explanation algorithm based on the Shapley vector to interpret the model’s output. It is shown that this approach allows the machine learning model to explain why it generates a particular forecast, which will provide a greater level of trust in intelligent information systems in the power industry.

All sky imaging-based short-term solar irradiance forecasting with Long Short-Term Memory networks

The intermittent nature of solar irradiance, primarily due to cloud movements, leads to rapid short-term fluctuations in the power output of photovoltaic (PV) systems. These fluctuations pose a significant challenge for integrating this renewable energy source into the power grid. Accurate forecasting of solar irradiance is not only crucial but also multi-beneficial. It enables more precise grid management by allowing operators to anticipate power output fluctuations and adjust energy distribution and storage strategies accordingly. This proactive approach reduces the reliance on backup power sources, which are often less sustainable and more expensive. Furthermore, accurate forecasts enhance the overall efficiency and reliability of energy systems by minimizing the impact of power variability on the grid, thereby supporting a more stable and sustainable energy supply.Addressing this need, our study focuses on the development of a forecasting model through innovative feature engineering, systematic design of specific attributes, and optimization of sequence length. The model is tailored to perform efficiently across various weather conditions and offers predictions for a time horizon of 0 to 20 min ahead. Utilizing a Long Short-Term Memory (LSTM) model, we achieve a remarkable ramp Forecast Skill Score of 39% in sunny and 25% in partially cloudy conditions. This work not only contributes to the existing literature but also presents a pioneering methodology for solar energy integration, highlighting the importance and application of accurate short-term solar irradiance forecasting.

Data-Driven Techniques for Short-Term Electricity Price Forecasting through Novel Deep Learning Approaches with Attention Mechanisms

The electricity market is constantly evolving, being driven by factors such as market liberalization, the increasing use of renewable energy sources (RESs), and various economic and political influences. These dynamics make it challenging to predict wholesale electricity prices. Accurate short-term forecasting is crucial to maintaining system balance and addressing anomalies such as negative prices and deviations from predictions. This paper investigates short-term electricity price forecasting using historical time series data and employs advanced deep learning algorithms. First, four deep learning models are implemented and proposed, which are a convolutional neural network (CNN) with an integrated attention mechanism, a hybrid CNN followed by a gated recurrent unit model (CNN-GRU) with an attention mechanism, and two ensemble learning models, which are a soft voting ensemble and a stacking ensemble model. Also, the optimized version of a transformer model, the Multi-Head Attention model, is introduced. Finally, the perceptron model is used as a benchmark for comparison. Our results show excellent prediction accuracy, particularly in the hybrid CNN-GRU model with attention, thereby achieving a mean absolute percentage error (MAPE) of 6.333%. The soft voting ensemble model and the Multi-Head Attention model also performed well, with MAPEs of 6.125% and 6.889%, respectively. These findings are significant, as previous studies have not shown high performance with transformer models and attention mechanisms. The presented results offer promising insights for future research in this field.

Short-term power load forecasting method based on Bagging-stochastic configuration networks.

Accurate short-term load forecasting is of great significance in improving the dispatching efficiency of power grids, ensuring the safe and reliable operation of power grids, and guiding power systems to formulate reasonable production plans and reduce waste of resources. However, the traditional short-term load forecasting method has limited nonlinear mapping ability and weak generalization ability to unknown data, and it is prone to the loss of time series information, further suggesting that its forecasting accuracy can still be improved. This study presents a short-term power load forecasting method based on Bagging-stochastic configuration networks (SCNs). First, the missing values in the original data are filled with the average values. Second, the influencing factors, such as the weather- and week-type data, are coded. Then, combined with the data of influencing factors after coding, the Bagging-SCNs integration algorithm is used to predict the short-term load. Finally, by taking the daily load data of Quanzhou City, Zhejiang Province as an example, the program of the abovementioned method is compiled in Python language and then compared with the long short-term memory neural network algorithm and the single-SCNs algorithm. Simulation results show that the proposed method for medium- and short-term load forecasting has a high forecasting accuracy and a significant effect on improving the accuracy of load forecasting.

Regional short-term load forecasting method based on power load characteristics of different industries

With the gradual increase in the proportion of end-user electricity consumption, the electrical load has become a crucial issue in carbon emission studies. Precise load forecasting will benefit the treatment, calculations, and markets of carbon emission. To explore the characteristics of industrial power load in depth and improve the accuracy of regional short-term load forecasting, this paper proposes a regional short-term load forecasting method based on the power load characteristics of different industries. First, a power load characteristic index system is established to obtain a fine industry classification. Considering the influence of external factors on the power load, the industries are then further classified depending on their sensitivity to external factors using a combination of grey correlation analyses. For the non-external factor-sensitive industries, an integrated model based on the Shapley value method is adopted; for the external factor-sensitive industries, a factor analysis-bidirectional gated recurrent unit-attention mechanism model is proposed. The final combination of the two is the output of the total regional load forecasting results. The combined forecasting method is tested using power load data from a city in Liaoning, China. The results show that the proposed method has apparent advantages over the traditional single forecasting model in terms of forecasting accuracy.