Keywords
SARIMAX grid
Time series modeling
Mean absolute error (MAE)
Root mean squared error (RMSE)
How to Cite
Abstract
The SARIMAX (Seasonal Autoregressive Integrated Moving Average with eXogenous regressors) model is a time series model that can be used to forecast future values of a time series, given its past values. It is beneficial for modeling time series data that exhibits seasonality and incorporating additional exogenous variables (variables that are not part of the time series itself but may affect it).
One way to optimize the performance of a SARIMAX model is to use a grid search approach to find the best combination of hyperparameters for the model. A grid search involves specifying a set of possible values for each hyperparameter and then training and evaluating the model using all possible combinations of these values. The combination of hyperparameters that results in the best model performance can then be chosen as the final model. To perform a grid search for a SARIMAX model, you must define the grid of hyperparameters you want to search over. This will typically include the values of the autoregressive (AR) and moving average (MA) terms and the values of any exogenous variables you want to include in the model. We will also need to define a metric to evaluate the model's performance, such as mean absolute or root mean squared error.
Once we have defined the grid of hyperparameters and the evaluation metric, you can use a grid search algorithm (such as a brute force search or a more efficient method such as random search or Bayesian optimization) to evaluate the performance of the model using all possible combinations of hyperparameters. The combination of hyperparameters that results in the best model performance can then be chosen as the final model.
In this article, we will explore the potential of SARIMAX for PV time series modeling. The objective is to find the optimal set of hyperparameters. Grid Search passes all hyperparameter combinations through the model individually and checks the results. Overall, it returns the collection of hyperparameters that yield the most outstanding results after running the model. One of the most optimal SARIMAX (p,d,q) x (P, D, Q,s) combinations is SARIMAX (0,0,1) x (0,0,0,4).
References
Huang X, Li Q, Tai Y, Chen Z, Liu J, Shi J, et al. Time series forecasting for hourly photovoltaic power using conditional generative adversarial network and Bi-LSTM. Energy. 2022; 246: 123403. https://doi.org/10.1016/j.energy.2022.123403
Vogt S, Schreiber J, Sick B. Synthetic photovoltaic and wind power forecasting data. ArXiv 2022 ; arXiv:2204.00411v1 [cs.LG].
Markovics D, Mayer MJ. Comparison of machine learning methods for photovoltaic power forecasting based on numerical weather prediction. Renew Sustain Energy Rev. 2022; 161: 112364. https://doi.org/10.1016/j.rser.2022.112364
Carneiro TC, de Carvalho PCM, Alves dos Santos H, Lima MAFB, Braga AP de S. Review on photovoltaic power and solar resource forecasting: current status and trends. J Sol Energy Eng. 2022; 144: 010801. https://doi.org/10.1115/1.4051652
Zhou Y, Wang J, Li Z, Lu H. Short-term photovoltaic power forecasting based on signal decomposition and machine learning optimization. Energy Convers Manag. 2022; 267: 115944. https://doi.org/10.1016/j.enconman.2022.115944
Fan GF, Wei HZ, Chen MY, Hong WC. Photovoltaic power generation forecasting based on the ARIMA-BPNN-SVR model. Glob J Energy Technol Res Updat. 2022; 9: 18-38. https://doi.org/10.15377/2409-5818.2022.09.2
Sah S, Surendiran B, Dhanalakshmi R, Yamin M. Covid‐19 cases prediction using SARIMAX Model by tuning hyperparameter through grid search cross‐validation approach. Expert Syst. 2022; e13086. https://doi.org/10.1111/exsy.13086
Zhao Z, Zhai M, Li G, Gao X, Song W, Wang X, et al. Study on the prediction effect of a combined model of SARIMA and LSTM based on SSA for influenza in Shanxi Province, China. Res Sq. 2023; 23: Article number 71. https://doi.org/10.21203/RS.3.RS-2081379/V1
Ismail BI. Maximizing power generation of a solar PV system for a potential application at musselwhite gold mine site in northwestern ontario, canada. Glob J Energy Technol Resh Updat. 2020; 7: 1-11. https://doi.org/10.15377/2409-5818.2020.07.1
Harat ZA, Asadi Zarch MA. Comparison of SARIMA and SARIMAX for long-term drought prediction. Desert Management 2022; 10: 1-16.
Kumar SS, Kumar A, Agarwal S, Syafrullah M, Adiyarta K. Forecasting indoor temperature for smart buildings with ARIMA, SARIMAX, and LSTM: A fusion approach. 2022 9th International Conference on Electrical Engineering, Computer Science and Informatics (EECSI), IEEE; 2022, p. 186-92. https://doi.org/10.23919/EECSI56542.2022.9946498
Sharadga H, Hajimirza S, Balog RS. Time series forecasting of solar power generation for large-scale photovoltaic plants. Renew Energy. 2020; 150: 797-807. https://doi.org/10.1016/j.renene.2019.12.131.
Zheng J, Zhang H, Dai Y, Wang B, Zheng T, Liao Q, et al. Time series prediction for output of multi-region solar power plants. Appl Energy. 2020; 257: 114001. https://doi.org/10.1016/j.apenergy.2019.114001
Zhou H, Zhang Y, Yang L, Liu Q, Yan K, Du Y. Short-term photovoltaic power forecasting based on long short term memory neural network and attention mechanism. IEEE Access 2019; 7: 78063-74. https://doi.org/10.1109/ACCESS.2019.2923006
Khazaei H, Moslemi R, Sharma R. Stochastic decision-making model for aggregation of residential units with PV-systems and storages. 2020 IEEE Power & Energy Society General Meeting (PESGM), IEEE; 2020, p. 1-5. https://doi.org/10.1109/PESGM41954.2020.9281448
Rafi A, Lee T, Wu W. A comparative study of methods to forecast domestic energy consumption aggregated with photovoltaic and heat pumps system. 2021 26th International Conference on Automation and Computing (ICAC), IEEE; 2021, p. 1-6. https://doi.org/10.23919/ICAC50006.2021.9594149
Steenbergen W. Adjusted grid search to find hyper-parameters in SARIMAX models: efficiently filling the shelves in kruidvat stores. Student Res Conf. 2018; 4: 1-4.
Omar MS, Kawamukai H. Comparison between the holt-winters and SARIMA models in the prediction of NDVI in an arid region in Kenya using pixel-wise NDVI time series. Acad J Res Sci Publish. 2021; 2: 1-15. https://doi.org/10.52132/Ajrsp/en.2231
Singh RK, Rani M, Bhagavathula AS, Sah R, Rodriguez-Morales AJ, Kalita H, et al. Prediction of the COVID-19 pandemic for the top 15 affected countries: Advanced autoregressive integrated moving average (ARIMA) model », JMIR public health and surveillance. JMIR Public Health Surveill. 2020; 6(2): e19115. https://doi.org/10.2196/19115
Chicco D, Warrens MJ, Jurman G. The coefficient of determination R-squared is more informative than SMAPE, MAE, MAPE, MSE and RMSE in regression analysis evaluation. PeerJ Comput Sci. 2021; 7. https://doi.org/10.7717/peerj-cs.623
Wang W, Lu Y. Analysis of the mean absolute error (MAE) and the root mean square error (RMSE) in assessing rounding model. IOP Conf Ser Mater Sci Eng 2018; 324: 012049. https://doi.org/10.1088/1757-899X/324/1/012049
Karunasingha DSK. Root mean square error or mean absolute error? Use their ratio as well. Inf Sci (N Y). 2022; 585: 609-29. https://doi.org/10.1016/j.ins.2021.11.036
Priyadarshini I, Mohanty P, Kumar R, Taniar D. Monkeypox outbreak analysis: an extensive study using machine learning models and time series analysis. Computers. 2023;12: 36. https://doi.org/10.3390/computers12020036
Barajas MA, Murphy MP, Lasseter LC, Sunny GI, Mazumdar H, Gohel HA, et al. Seasonal trend assessment for groundwater contamination detection and monitoring using ARIMA model. 2023 IEEE 2nd International Conference on AI in Cybersecurity (ICAIC), févr; 2023, p. 1-7. https://doi.org/10.1109/ICAIC57335.2023.10044126
Alhakeem ZM, Jebur YM, Henedy SN, Imran H, Bernardo LFA, Hussein HM. Prediction of ecofriendly concrete compressive strength using gradient boosting regression tree combined with GridSearchCV hyperparameter-optimization techniques. Materials. 2022; 15: 7432. https://doi.org/10.3390/ma15217432
Priya SM. Hyper tuning using gridsearchcv on machine learning models for prognosticating dementia. Research Square. 2022; p. 1-15.
B. Jovičić. Mecasolar supplies 1 MW in 2-Axis trackers. Balkan Green Energy News, May 27, 2015.
Kim E, Akhtar MS, Yang O-B. Designing solar power generation output forecasting methods using time series algorithms. Electr Power Syst Res. 2023; 216: 109073. https://doi.org/10.1016/j.epsr.2022.109073
Idman E, Idman E, Yildirim O. Estimating solar power plant data using time series analysis methods. 2020 International Congress on Human-Computer Interaction, Optimization and Robotic Applications (HORA), Ankara, Turkey: IEEE; 2020, p. 1-6. https://doi.org/10.1109/HORA49412.2020.9152839
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
Copyright (c) 2022 Todizara Andrianajaina, David Tsivalalaina Razafimahefa, Raonirivo Rakotoarijaina, Cristian Goyozo Haba