TY - GEN

T1 - Influence of Machine Learning Method Choice on the Accuracy of Power Load Forecast Models and HPP Cascade Mode

T2 - 2023 Belarusian-Ural-Siberian Smart Energy Conference (BUSSEC)

AU - Sidorova, Alena

AU - Haljasmaa, Kristina

N1 - The research was carried out within the state assignment with the financial support of the Ministry of Science and Higher Education of the Russian Federation (subject No. FEUZ-2022-0030 Development of an intelligent multi-agent system for modeling deeply integrated technological.

PY - 2023

Y1 - 2023

N2 - This article proposes a new approach to operational forecasting of power load schedules with the use of machine learning methods. In order to increase the accuracy of models and algorithms for operational forecasting of power load schedules, testing of several machine learning methods was implemented. The developed models were tested in the Python programming language using the sktime and scikit-learn libraries. A calculation of power modes was carried out, which showed high comparability of the results with the actual values of the mode. The residual values of the controlled parameters based on the results of using the calculated and predicted value of the power load with the use of machine learning methods showed a higher comparability of the results with the real power mode. This allows us to talk about obtaining a more accurate result when using the forecast of the power load in nodes with the use of machine learning methods and their possible use in dispatch centers as a basis for planning long-term conditions and coordinating repair schedules for the power grid equipment. © 2023 IEEE.

AB - This article proposes a new approach to operational forecasting of power load schedules with the use of machine learning methods. In order to increase the accuracy of models and algorithms for operational forecasting of power load schedules, testing of several machine learning methods was implemented. The developed models were tested in the Python programming language using the sktime and scikit-learn libraries. A calculation of power modes was carried out, which showed high comparability of the results with the actual values of the mode. The residual values of the controlled parameters based on the results of using the calculated and predicted value of the power load with the use of machine learning methods showed a higher comparability of the results with the real power mode. This allows us to talk about obtaining a more accurate result when using the forecast of the power load in nodes with the use of machine learning methods and their possible use in dispatch centers as a basis for planning long-term conditions and coordinating repair schedules for the power grid equipment. © 2023 IEEE.

UR - http://www.scopus.com/inward/record.url?partnerID=8YFLogxK&scp=85178019976

U2 - 10.1109/BUSSEC59406.2023.10296424

DO - 10.1109/BUSSEC59406.2023.10296424

M3 - Conference contribution

SN - 979-835035807-0

SP - 115

EP - 120

BT - Proceedings of the 2023 Belarusian-Ural-Siberian Smart Energy Conference, BUSSEC 2023

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 25 September 2023 through 29 September 2023

ER -