Review on Load Forecasting in Modern Power Systems: Comparative Analysis of Techniques, Challenges, Research Gap and Emerging Trends

Abstract

Accurate prediction of load is a fundamental aspect of planning and operation of power systems. The complexity and uncertainty of load forecasting have been significantly increased due to the evolution of renewable energy sources, the integration of electric vehicles and other innovations associated with the smart grid concept. This work presents a detailed overview of load forecasting techniques that include traditional statistical models, various AI, machine learning, and deep learning models. Load forecasting methods based on statistical models, AI-based methods, deep learning methods,machine learning methods, and Transformer-based architecture is reviewed and analyzed. A comparative analysis of these methods is performed on the basis of category, complexity, strengths, and limitations. The factors affecting electrical load are also studied such as the temperature, humidity, holidays, and economic activity to understand their effect on the accuracy of forecasting. In additionchallenges,research gaps and future direction are discussed. The review aims to help researchers select the most appropriate forecasting technique for their research and to gain a better understanding of the recent developments in the area. It will also help utilities to select effective forecasting methods for better grid management and energy planning.

Introduction

Electrical load forecasting predicts future electricity demand to support efficient power system operation, including generation scheduling, energy trading, and infrastructure planning. The increasing adoption of smart grids, distributed energy resources (DERs), electric vehicles (EVs), demand response programs, and prosumers has made electricity consumption patterns more complex and nonlinear, reducing the effectiveness of traditional forecasting methods. While statistical techniques such as regression and ARIMA remain useful for simple forecasting tasks, modern machine learning and deep learning models—including Artificial Neural Networks (ANNs), Support Vector Regression (SVR), Random Forest, XGBoost, Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Convolutional Neural Networks (CNNs), and Transformer-based architectures—have significantly improved forecasting accuracy, often achieving Mean Absolute Percentage Error (MAPE) below 2%.

Load forecasting is categorized into short-term (hours to one week), medium-term (one week to one year), and long-term (several years) forecasting, each serving different operational and planning purposes. Forecast accuracy is influenced by multiple factors, including weather conditions (especially temperature), economic and demographic variables, calendar effects, and social or behavioral patterns such as smart home adoption and consumer energy usage. Hybrid forecasting models that combine statistical methods with artificial intelligence effectively capture both linear and nonlinear relationships, improving prediction performance. This review comprehensively examines load forecasting techniques from 2000 to 2025, highlighting the evolution from traditional statistical models to advanced AI, machine learning, deep learning, ensemble, and hybrid approaches, while emphasizing their importance in supporting reliable, scalable, and intelligent smart grid operations.

Conclusion

Thepaperoffersanoverviewofthecurrentstateofartinloadforecastingmethods.StatisticalmodelssuchasARIMA and regression have been compared to more modern architectures such as LSTM, Transformer, CNN–LSTM, and CNN. The modern methods reflects the influence of data availability and computational resources. Fromtheresultsofthecomparativeanalysis,itcanbeconcludedthatdeeplearningarchitecturesarethemostaccuratemodels forshort-termforecasts,whilegradientboostingmodelscanbeconsideredacompromisebetweenaccuracy,explainability,andcomputationalcosts.Thus,thechoiceofaforecastingalgorithmdependsonspecificforecastingtasks. This review also emphasizes that weather conditions, seasonal variations, holidays and socio-economic parameters have a significantimpactontheelectricalloadandshouldbetakenintoaccountwhileselectinganappropriateforecastingmodel for accurate predictions. Despite the impressive results achieved, there are still unresolved issues related to data quality, explainability of models, integration of renewable sources, and data protection. Future forecasting models will critically need novel approaches like federated learning, transfer learning, and probabilistic forecasting.

References

[1] Q.Ma,D.Wu,andX.Luo,“Areviewofdeeplearning-basedpowerloadforecastingmethods,”InternationalJournalofNetworkDynamicsandIntelligence, vol. 4, no. 4, pp. 100027, 2025. [2] H.-Y.Su,T.-Y.Chen,andY.-H.Lin,“Towardsimprovedloadforecastinginsmartgrids:Arobustdeepensemblelearningframework,”IEEETransactionson Smart Grid, 2024, Early Access. [3] F.Yang,X.Fu,Q.Yang,andZ.Chu,“Decompositionstrategyandattention-basedlongshort-termmemorynetworkformulti-stepultra-short-termagricultural power load forecasting,” Expert Systems with Applications, vol. 238, pp. 122226, 2024. [4] B. Zhong, “Deep learning integration optimization of electric energy load forecasting and market price based on the ANN–LSTM–Transformer method,”Frontiers in Energy Research, vol. 11, pp. 1292204, 2023. [5] Z. Xu, H. Wang, Y. Zhang, T. Liu, and K. Chen, “PhaCIA-TCNs: Short-term load forecasting using temporal convolutional networks with parallel hybridactivatedconvolutionandinputattention,”IEEETransactionsonNetworkScienceandEngineering,vol.11,no.1,pp.427–438,2024. [6] B.Peng,J.Zuo,X.Gong,Y.Li,W.Wu,andY.Li,“Short-termelectricityloadforecastingforcommunityresidentsbasedonfederatedlearningconsidering privacy protection,” EAI Endorsed Transactions on Energy Web, vol. 11, 2024. [7] Y. Wang, N. Gao, and G. Hug, “Personalized federated learning for individual consumer load forecasting,” CSEE Journal of Power and Energy Systems,vol. 9, no. 1, pp. 326–330, 2023. [8] F.Dewangan,A.Y.Abdelaziz,andM.Biswal,“Loadforecastingmodelsinsmartgridusingsmartmeterinformation:Areview,”Energies,vol.16,no. 3, pp. 1338, 2023. [9] S.RyuandY.Yu,“Quantile-Mixer:Anoveldeeplearningapproachforprobabilisticshort-termloadforecasting,”IEEETransactionsonSmartGrid,vol. 15, no. 2, pp. 2237–2250, 2024. [10] A. Wan, Q. Chang, A. L. B. Khalil, and J. He, “Short-term power load forecasting for combined heat and power using CNN-LSTM enhanced by attentionmechanism,” Energy, vol. 282, pp. 128274, 2023. [11] B.S.Vardhan,M.Khedkar,I.Srivastava,andS.K.Patro,“Impactofintegratedclassifier-regressionmappedshort-termloadforecastingonpowersystem management in a grid-connected multi-energy system,” Electric Power Systems Research, vol. 230, pp. 110222, 2024. [12] G.Dudek,P.Piotrowski,andD.Baczyn´ski,“Forecastinginmodernpowersystems:Challenges,techniques,andemergingtrends,”Energies,vol.18,no. 14, p. 3589, 2025, doi: 10.3390/en18143589. [13] J. Liu, Z. Hou, and T. Yin, “Short-term power load forecast using OOA optimized bidirectional long short-term memory network with spectral attentionfor the frequency domain,” Energy Reports, vol. 12, pp. 4891–4908, 2024. [14] Y. Zhang and W. Chen, “Empowering data-driven load forecasting by leveraging long short-term memory recurrent neural networks,” Heliyon, vol. 10,no. 24, pp. e40934, 2024. [15] International Energy Agency, “Global EV Outlook 2024,” IEA, Paris, France, Tech. Rep., 2024. [Online]. Available: https://www.iea.org/reports/global-ev-outlook-2024 [16] S.Rallabandi,“AIdrivencapacityplanninginlargescaleinfrastructure:AcomparativeanalysisofLSTMnetworksandtraditionalforecastingmethods,”InternationalJournalofScientificResearchinComputerScience,EngineeringandInformationTechnology,vol.10,no.6,pp.950–962,2024. [17] M.Turkog?lu,B.Erkmen,Y.Eren,O.Erdinc¸,andI.Ku¨c¸u¨kdemiral,“IntegratedapproachesinresilienthierarchicalloadforecastingviaTCNandoptimalvalley filling based demand response application,” Applied Energy, vol. 360, pp. 122722, 2024. [18] B.Biswal,S.Deb,S.Datta,T.S.Ustun,andU.Cali,“Reviewonsmartgridloadforecastingforsmartenergymanagementusingmachinelearningand deep learning techniques,” Energy Reports, vol. 12, pp. 3654–3670, 2024, doi: 10.1016/j.egyr.2024.09.056. [19] G. K. Ediga, P. S. Arya, P. A. Tomar, S. Tyagi, A. Mittal, and M. I. Habelalmateen, “Smart energy management: Real-time prediction and optimizationfor IoT-enabled smart homes,” Cogent Engineering, vol. 11, no. 1, p. 2390674, 2024, doi: 10.1080/23311916.2024.2390674. [20] J. Moon, M. Maqsood, D. So, S. W. Baik, S. Rho, and Y. Nam, “Advancing ensemble learning techniques for residential building electricity consumptionforecasting: Insight from explainable artificial intelligence,” PLOS ONE, vol. 19, no. 11, p. e0307654, Nov. 2024, doi: 10.1371/journal.pone.0307654. [21] S.Khan,“Short-termelectricityloadforecastingusinganewintelligence-basedapplication,”Sustainability,vol.15,no.16,p.12311,2023,doi: 10.3390/su15161231. [22] Y. Wang, J. Wang, G. Zhao, and Y. Dong, “Application of residual modification approach in seasonal ARIMA for electricity demand forecasting: A casestudy of China,” Energy Policy, vol. 48, pp. 284–294, 2012. [23] R.Hachache,M.Labrahmi,A.Grilo,A.Chaoub,R.Bennani,A.Tamtaoui,andB.Lakssir,“Energyloadforecastingtechniquesinsmartgrids:Across-country comparative analysis,” Energies, vol. 17, no. 10, Art. no. 2251, 2024. [24] L.D.X.Ribeiro,“AdaptiveKalmanBasedForecastingforElectricLoadandDistributedGeneration,”Ph.D.dissertation,UniversityofBras´?lia,Bras´?lia,Brazil, 2017. [25] A.I.Arvanitidis,D.Bargiotas,A.Daskalopulu,V.M.Laitsos,andL.H.Tsoukalas,“Enhancedshort-termloadforecastingusingartificialneuralnetworks,” Energies, vol. 14, no. 22, p. 7788, 2021, doi: 10.3390/en14227788. [26] R.K.Shiwakoti,C.Charoenlarpnopparut,andK.Chapagain,“Adeeplearningapproachforshort-termelectricitydemandforecasting:AnalysisofThailand data,” Applied Sciences, vol. 14, no. 10, Art. no. 3971, 2024. [27] A.TehlanandV.Kumar,“Fuzzylogicbasedloadforecasting,”InternationalJournalofAppliedEngineeringResearch,vol.11,pp.6625–6626,2016. [28] J. Chakravorty, S. Shah, and H. N. Nagraja, “ANN and ANFIS for short term load forecasting,” Engineering, Technology & Applied Science Research,vol. 8, no. 2, pp. 2818–2820, Apr. 2018. [29] L.Hu,L.Zhang,T.Wang,andK.Li,“Short-termloadforecastingbasedonsupportvectorregressionconsideringcoolingloadinsummer,”inProc.Chinese Control and Decision Conference (CCDC), pp. 5495–5498, Aug. 2020, doi: 10.1109/CCDC49329.2020.9164387. [30] “Financialforecastingwithmultivariateadaptiveregressionsplinesandqueengeneticalgorithm-supportvectorregression,”IEEEAccess,vol.7, [31] pp.91862–91873,2019,doi:10.1109/ACCESS.2019.2927277. [32] O.Ezugwu,C.Opata,A.Dimeze,C.Ogbuka,andC.Nwosu,“DevelopmentofahybridARIMA-ANNmodelforenhancedelectricloadforecastingaccuracy,” in Proc. 1st International Conference/Exhibitions of the Department of Mechatronic Engineering, University of Nigeria, Nsukka, Feb. 2025. [33] M.Asiri,G.Aldehim,F.Alotaibi,andM.M.Alnfiai,“Short-termloadforecastinginsmartgridsusinghybriddeeplearning,”IEEEAccess,pp.1–1,Jan. 2024, doi: 10.1109/ACCESS.2024.3358182. [34] M. Abumohsen, A. Y. Owda, and M. Owda, “Electrical load forecasting using LSTM, GRU, and RNN algorithms,” Energies, vol. 16, no. 5, pp. 2283,2023. [35] B. Singh, “Seasonal electrical load forecasting using machine learning techniques and meteorological variables,” International Journal of Robotics andControl Systems, vol. 4, no. 3, pp. 1092–1108, 2024. [36] T.Osgonbaatar,P.Matrenin,M.Safaraliev,I.Zicmane,A.Rusina,andS.Kokin,“Short-termelectricalloadforecastingbasedonXGBoostmodel,”Energies, vol. 18, no. 19, pp. 5144, 2025. [37] U. Singh and S. Vadhera, “Random forest and XGBoost technique for short-term load forecasting,” in Proc. IEEE Int. Conf. Sustainable Technology forPower and Energy Systems (STPES), 2023, pp. 1–6. [38] X.Chen,Y.Li,andZ.Wang,“Short-termloadforecastingbasedonCEEMDAN-FE-ISSA-LightGBMmodel,”FrontiersinEnergyResearch,vol.11,pp.1111786,2023. [39] D.Yang,J.-E.Guo,Y.Li,S.Sun,andS.Wang,“Short-termloadforecastingwithanimproveddynamicdecomposition-reconstruction-ensembleapproach,” Energy, vol. 263, pp. 125609, 2023. [40] K.Ullah,M.Ahsan,S.M.Hasanat,M.Haris,H.Yousaf,andS.F.Raza,“Short-termloadforecasting:Acomprehensivereviewandsimulationstudywith CNN-LSTM hybrid approach,” IEEE Access, vol. 12, pp. 111858–111881, 2024. [41] M.Waheedandothers,“Datadrivenlongshort-termloadprediction:LSTM-RNN,XG-Boostandconventionalmodelsincomparativeanalysis,”Computational Intelligence, vol. 41, no. 3, pp. e70084, 2025. [42] Y.Bao,“Deeplearning-drivenhybridmodelforshort-termloadforecastingandsmartgridinformationmanagement,”ScientificReports,vol.14,no.1,pp.13720,2024.doi:10.1038/s41598-024-63262-x [43] S.Jangandothers,“Comparativeanalysisofdeeplearningtechniquesforloadforecastinginpowersystemsusingsingle-layerandhybridmodels,” [44] InternationalTransactionsonElectricalEnergySystems,vol.2024,pp.5587728,2024.doi:10.1155/2024/5587728 [45] S. M. Hasanat, K. Ullah, H. Yousaf, and Z. Zahid, “Enhancing load forecasting accuracy in smart grids: A novel parallel multichannel network approachusing1DCNNandBi-LSTMmodels,”InternationalJournalofEnergyResearch,vol.2024,pp.2403847,2024.doi:10.1155/2024/2403847 [46] S.Li,M.Chen,D.Zhang,andY.Wang,“PowergridloadforecastingusingaCNN-LSTMnetworkbasedonamulti-modalattentionmechanism,” [47] AppliedSciences,vol.15,no.5,pp.2435,2025.doi:10.3390/app15052435 [48] J. Gong, Z. Qu, Z. Zhu, H. Xu, and Q. Yang, “Ensemble models of TCN-LSTM-LightGBM based on ensemble learning methods for short-term electricalload forecasting,” Energy, vol. 318, pp. 134757, 2025. [49] H. Zhou, S. Zhang, J. Peng, S. Zhang, J. Li, H. Xiong, and W. Zhang, “Informer: Beyond efficient transformer for long sequence time-series forecasting,”Proceedings of the AAAI Conference on Artificial Intelligence, vol. 35, pp. 11106–11115, 2023. [50] Y. Nie, N. H. Nguyen, P. Sinthong, and J. Kalagnanam, “A time series is worth 64 words: Long-term forecasting with transformers,” in Proc. Int. Conf.Learning Representations (ICLR), 2023. [51] B. Lim, S. O. Ar?k, N. Loeff, and T. Pfister, “Temporal fusion transformers for interpretable multi-horizon time series forecasting,” International Journalof Forecasting, vol. 37, no. 4, pp. 1748–1764, 2021. [52] G.N.,“Short-termloadforecastingusingARIMAmodelforKarnatakastateelectricalload,”InternationalJournalofEngineeringResearchandDevelopment, vol. 13, no. 7, pp. 75–79, Jul. 2017. [53] A.Alqaysi,A.Bozkurt,andY.Ates¸,“ElectricalloadforecastingbasedonartificialneuralnetworkinIraq,”inProc.10thInternationalCongressonEngineering, Architecture and Design, Apr. 2023. [54] Z.Hu,Y.Bao,andT.Xiong,“Electricityloadforecastingusingsupportvectorregressionwithmemeticalgorithms,”TheScientificWorldJournal,vol. 2013, Art. no. 292575, 2013, doi: 10.1155/2013/292575. [55] C.-H.Wu,Y.-T.Tseng,W.-F.Lo,andY.-H.Haung,“Forecastingrenewableenergygenerationusingrandomforestanalysis,”inProc.InternationalAssociation for Management of Technology 2025 (IAMOT 2025), Montreal, Canada, Jun. 2025. [56] H. I. Beloev, S. R. Saitov, A. A. Filimonova, N. D. Chichirova, O. E. Babikov, and I. K. Iliev, “Short-term electrical load forecasting based on XGBoostmodel,” Energies, vol. 18, no. 19, Art. no. 5144, 2025. [57] S.Bouktif,A.Fiaz,A.Ouni,andM.A.Serhani,“OptimaldeeplearningLSTMmodelforelectricloadforecastingusingfeatureselectionandgeneticalgorithm: Comparison with machine learning approaches,” Energies, vol. 11, no. 7, Art. no. 1636, 2018. [58] M.Alhussein,K.Khursheed,andS.Haider,“HybridCNN-LSTMmodelforshort-termindividualhouseholdloadforecasting,”IEEEAccess,vol.8,pp.180544–180557,2020,doi:10.1109/ACCESS.2020.3028281. [59] S.Huang,C.Yan,andY.Qu,“Deeplearningmodel-transformerbasedwindpowerforecastingapproach,”FrontiersinEnergyResearch,vol.10,Art. no. 1055683, 2023, doi: 10.3389/fenrg.2022.1055683. [60] T. Hong, P. Pinson, and S. Fan, “Global energy forecasting competition 2012,” International Journal of Forecasting, vol. 30, no. 2, pp. 357–363, 2014.doi: 10.1016/j.ijforecast.2013.07.001 [61] T. Hong, P. Pinson, S. Fan, H. Zareipour, A. Troccoli, and R. J. Hyndman, “Probabilistic energy forecasting: Global energy forecasting competition 2014and beyond,” International Journal of Forecasting, vol. 32, no. 3, pp. 896–913, 2016. doi: 10.1016/j.ijforecast.2016.02.001 [62] R.Bryce,G.Buster,K.Doubleday,C.Feng,R.Ring-Jarvi,M.Rossol,F.Zhang,andB.-M.Hodge,“SolarPV,windgeneration,andloadforecastingdataset for ERCOT 2018: P.E.R.F.O.R.M.,” National Renewable Energy Laboratory, Golden, CO, USA, Tech. Rep., 2023. doi: 10.2172/1972698 [63] PJMInterconnection,“2024Long-termloadforecastreport,”PJM,Audubon,PA,USA,Tech.Rep.,2024.[Online].Available:https://www.pjm.com/library/reports-notices/load-forecast [64] G.He´brailandA.Be´rard,“Individualhouseholdelectricpowerconsumption[Dataset],”UCIMachineLearningRepository,2006.doi:10.24432/C58K54 [65] A.Trindade,“Electricityloaddiagrams2011–2014[Dataset],”UCIMachineLearningRepository,2015.[Online].Available:https://archive.ics.uci.edu/dataset/321 [66] IndependentElectricitySystemOperator(IESO),“Ontariohourlyelectricitydemanddata,”IESO,Toronto,ON,Canada,2024.[Online].Available:https://www.ieso.ca/en/Power-Data/Data-Directory [67] C.Termritthikunandothers,“Electricityloadforecastingusinghybriddatasetswithlinearinterpolationandsyntheticdata,”Eng.,Technol.&Appl.Sci.Res.,vol.14,no.6,pp.17931–17938,2024.doi:10.48084/etasr.8577 [68] X.Li,Y.Chen,D.Zhao,andZ.Wang,“Identificationandcorrectionofabnormal,incompletepowerloaddatainelectricityspotmarketdatabases,”Energies,vol.18,no.1,pp.176,2025.doi:10.3390/en18010176 [69] L. Baur, K. Ditschuneit, M. Schambach, C. Kaymakci, T. Wollmann, and A. Sauer, “Explainability and interpretability in electric load forecasting usingmachine learning techniques – A review,” Energy and AI, vol. 16, pp. 100358, 2024. doi: 10.1016/j.egyai.2024.100358 [70] V.ArabzadehandR.Frank,“Afour-dimensionalanalysisofexplainableAIinenergyforecasting:Adomain-specificsystematicreview,”MachineLearning and Knowledge Extraction, vol. 7, no. 4, p. 153, 2025, doi: 10.3390/make7040153. [71] M.A.A.Sarker,B.Shanmugam,S.Azam,andS.Thennadil,“Enhancingsmartgridloadforecasting:Attention-baseddeeplearningwithfederatedlearning and XAI,” Intell. Syst. Appl., vol. 23, pp. 200422, 2024. doi: 10.1016/j.iswa.2024.200422 [72] C.Feng,Y.Wang,Q.Chen,Y.Ding,G.Strbac,andC.Kang,“Smartgridencountersedgecomputing:Opportunitiesandapplications,”Adv.Appl.Energy,vol.1,pp.100006,2021.doi:10.1016/j.adapen.2020.100006 [73] A.X.WangandJ.J.Li,“Anovelcloudedgecollaborationbasedshorttermloadforecastingmethodforsmartgrid,”Front.EnergyRes.,vol.10,pp.977026,2022.doi:10.3389/fenrg.2022.977026 [74] B.Peng,J.Zuo,X.Gong,Y.Li,W.Wu,andY.Li,“Short-termelectricityloadforecastingbasedonfederatedlearningconsideringprivacyprotection,”EAIEndorsedTrans.EnergyWeb,vol.11,2024.doi:10.4108/ew.6007 [75] Y.Wang,N.Gao,andG.Hug,“Personalizedfederatedlearningforindividualconsumerloadforecasting,”CSEEJ.PowerEnergySyst.,vol.9,no.1,pp.326–330,2023.doi:10.17775/CSEEJPES.2021.08300

Copyright

Copyright © 2026 Pratima Patel, Prof. Seema Pal. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.