利用机器学习方法预测资源发电

Yu. N. Zacarinnaya, G. V. Reutin, S. S. Kurilov, O. V. Isaeva, G. S. Kovalev
{"title":"利用机器学习方法预测资源发电","authors":"Yu. N. Zacarinnaya, G. V. Reutin, S. S. Kurilov, O. V. Isaeva, G. S. Kovalev","doi":"10.30724/1998-9903-2023-25-3-81-92","DOIUrl":null,"url":null,"abstract":"RELEVANCE. Today, the degree of integration of renewable energy sources into the energy system is an indicator of the technological and industrial development of the state. Renewable energy is a driver for the development of the economy, science and education. In Russia, the largest technical potential from renewable energy sources in the Sun (in million tons of standard fuel) is 2.3 * 103, the second place is occupied by wind energy - 2 * 103. However, the use of solar energy is associated with great difficulties in predicting the generation of electricity due to its dependence on meteorological conditions, and there is an acute issue of forecasting the generation. In this article, the authors propose a solution to the urgent problem of predicting energy generation from solar power plants using machine learning systems. TARGET. The purpose of this work is to study the performance of modern artificial intelligence methods to create a platform for predicting the power generated from a solar station to an existing network. Develop the architecture of the information and communication system of the distribution network and the model for predicting the photovoltaic power of the power plant based on machine learning methods. METHODS. One approach to solving this problem is to use machine learning algorithms. Such algorithms, with a correctly chosen training model, are capable of predicting the volume of electricity generation a day ahead with a high accuracy of up to 95%. RESULTS. The values of real generation and predicted generation were compared by five machine learning algorithms, such as neural networks, linear regression, decision tree, random forest, adaptive boosting. The random forest algorithm has the smallest mean square error on the test data. The problem of optimization of the radial topology of the network, which minimizes the total loss of active power, is solved. CONCLUSION. An analysis of the construction of a working machine learning model showed that in order to build an optimal model, only the history of the power generation of this plant, compared with the calculated and measured weather data, is needed. The stability of the model was tested by applying the cross-validation method under various training and testing conditions. The results obtained showed that the model works reliably, since the root-mean-square error of the most accurate model is in the region of 600 kWh (4%).","PeriodicalId":222237,"journal":{"name":"Power engineering: research, equipment, technology","volume":"177 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2023-08-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Prediction of electricity generation from res by machine learning methods\",\"authors\":\"Yu. N. Zacarinnaya, G. V. Reutin, S. S. Kurilov, O. V. Isaeva, G. S. Kovalev\",\"doi\":\"10.30724/1998-9903-2023-25-3-81-92\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"RELEVANCE. Today, the degree of integration of renewable energy sources into the energy system is an indicator of the technological and industrial development of the state. Renewable energy is a driver for the development of the economy, science and education. In Russia, the largest technical potential from renewable energy sources in the Sun (in million tons of standard fuel) is 2.3 * 103, the second place is occupied by wind energy - 2 * 103. However, the use of solar energy is associated with great difficulties in predicting the generation of electricity due to its dependence on meteorological conditions, and there is an acute issue of forecasting the generation. In this article, the authors propose a solution to the urgent problem of predicting energy generation from solar power plants using machine learning systems. TARGET. The purpose of this work is to study the performance of modern artificial intelligence methods to create a platform for predicting the power generated from a solar station to an existing network. Develop the architecture of the information and communication system of the distribution network and the model for predicting the photovoltaic power of the power plant based on machine learning methods. METHODS. One approach to solving this problem is to use machine learning algorithms. Such algorithms, with a correctly chosen training model, are capable of predicting the volume of electricity generation a day ahead with a high accuracy of up to 95%. RESULTS. The values of real generation and predicted generation were compared by five machine learning algorithms, such as neural networks, linear regression, decision tree, random forest, adaptive boosting. The random forest algorithm has the smallest mean square error on the test data. The problem of optimization of the radial topology of the network, which minimizes the total loss of active power, is solved. CONCLUSION. An analysis of the construction of a working machine learning model showed that in order to build an optimal model, only the history of the power generation of this plant, compared with the calculated and measured weather data, is needed. The stability of the model was tested by applying the cross-validation method under various training and testing conditions. The results obtained showed that the model works reliably, since the root-mean-square error of the most accurate model is in the region of 600 kWh (4%).\",\"PeriodicalId\":222237,\"journal\":{\"name\":\"Power engineering: research, equipment, technology\",\"volume\":\"177 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2023-08-02\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Power engineering: research, equipment, technology\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.30724/1998-9903-2023-25-3-81-92\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Power engineering: research, equipment, technology","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.30724/1998-9903-2023-25-3-81-92","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

摘要

的相关性。今天,可再生能源融入能源系统的程度是一个国家科技和工业发展的一个指标。可再生能源是经济、科学和教育发展的动力。在俄罗斯,太阳能可再生能源的最大技术潜力(以百万吨标准燃料计算)为2.3 * 103,第二位是风能- 2 * 103。然而,由于太阳能的利用对气象条件的依赖,在预测发电量方面存在很大困难,预测发电量的问题十分突出。在本文中,作者提出了一种利用机器学习系统预测太阳能发电厂发电量的紧迫问题的解决方案。目标。这项工作的目的是研究现代人工智能方法的性能,以创建一个平台,用于预测从太阳能站到现有网络的发电量。开发了配电网信息通信系统的体系结构和基于机器学习方法的电站光伏功率预测模型。方法。解决这个问题的一种方法是使用机器学习算法。这样的算法,在正确选择的训练模型下,能够预测未来一天的发电量,准确率高达95%。结果。通过神经网络、线性回归、决策树、随机森林、自适应增强等5种机器学习算法,比较了实际生成和预测生成的值。随机森林算法对测试数据的均方误差最小。解决了使有功总损耗最小的网络径向拓扑优化问题。结论。对工作机器学习模型构建的分析表明,为了构建最优模型,只需要将该电厂的发电历史与计算和测量的天气数据进行比较。在不同的训练和测试条件下,应用交叉验证方法对模型的稳定性进行了检验。结果表明,该模型工作可靠,最精确模型的均方根误差在600 kWh(4%)左右。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
Prediction of electricity generation from res by machine learning methods
RELEVANCE. Today, the degree of integration of renewable energy sources into the energy system is an indicator of the technological and industrial development of the state. Renewable energy is a driver for the development of the economy, science and education. In Russia, the largest technical potential from renewable energy sources in the Sun (in million tons of standard fuel) is 2.3 * 103, the second place is occupied by wind energy - 2 * 103. However, the use of solar energy is associated with great difficulties in predicting the generation of electricity due to its dependence on meteorological conditions, and there is an acute issue of forecasting the generation. In this article, the authors propose a solution to the urgent problem of predicting energy generation from solar power plants using machine learning systems. TARGET. The purpose of this work is to study the performance of modern artificial intelligence methods to create a platform for predicting the power generated from a solar station to an existing network. Develop the architecture of the information and communication system of the distribution network and the model for predicting the photovoltaic power of the power plant based on machine learning methods. METHODS. One approach to solving this problem is to use machine learning algorithms. Such algorithms, with a correctly chosen training model, are capable of predicting the volume of electricity generation a day ahead with a high accuracy of up to 95%. RESULTS. The values of real generation and predicted generation were compared by five machine learning algorithms, such as neural networks, linear regression, decision tree, random forest, adaptive boosting. The random forest algorithm has the smallest mean square error on the test data. The problem of optimization of the radial topology of the network, which minimizes the total loss of active power, is solved. CONCLUSION. An analysis of the construction of a working machine learning model showed that in order to build an optimal model, only the history of the power generation of this plant, compared with the calculated and measured weather data, is needed. The stability of the model was tested by applying the cross-validation method under various training and testing conditions. The results obtained showed that the model works reliably, since the root-mean-square error of the most accurate model is in the region of 600 kWh (4%).
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信