A new study reveals the best artificial intelligence models to improve the energy network
Authors:
(1) Mingshuo Jia, Department of IT and Electrical Engineering, Eth Zürich, Physikstrasse 3, 8092, Zürich, Switzerland;
(2) Gabriela Hug, Department of Information Technology and Electrical Engineering, ETH Zürich, Physikstrasse 3, 8092, Zürich, Switzerland;
(3) Ning Chang, Department of Electrical Engineering, University of Tsinghua, Shuangzing RD 30, 100084, Beijing, China;
(4) Zhaojian Wang, Automation Department, Shanghai Jiao Tong University, Dongchuan 800 Street, 200240, Shanghai, China;
(5) Yi Wang, Department of Electrical and Electronic Engineering, Hong Kong University, Book Fu Lam, Hong Kong, China;
(6) Chongqing Kang, Department of Electrical Engineering, Tsinghua University, Shuanging RD 30, 100084, Beijing, China.
Links table
Abstract and 1. Introduction
2. Evaluation methods
3. Review the current experiments
4. Circular and application assessments and 4.1. Prediction and response circulation
4.2. Applications for multiple -written situations and 4.3. Zero predict the ability of the application
4.4. Continuous prediction and 4.5. Normalization
5. numerical assessments and 5.1. Experience settings
5.2. Overview of the evaluation
5.3. Failure evaluation
5.4. Accuracy
5.5. Efficiency evaluation
6. Open questions
7. Conclusion
Approach a and references
a summary
Building the theoretical visions of the first part, this paper, as the second part of the tutorial, is deeper into the data -based energy flow line (DPFL), with a focus on comprehensive numerical tests. The necessity of this simulation stems from the restrictions inherent in theoretical analysis, especially the challenge of identifying differences in performance in the real world between DPFL methods with overlapping theoretical capabilities and/or restrictions. The lack of a comprehensive numerical comparison of the DPFL approach to literature stimulates this paper, especially given the fact that more than 95 % of the current DPFL studies have not provided any open source symbols. To bridge the gap, this paper first reviews the current DPFL experiments, study the approved test scenarios, loading fluctuations settings, data sources, data/extreme values, and comparison that has been made so far. Next, this paper evaluates a total of 44 methods, containing more than 30 current DPFL approaches, some innovative DPFL technologies, and many methods of relieving classic energy flow that depend on physics for standards. Various dimensions extending, including generalization, application, accuracy, and mathematical efficiency, using many different test cases that expand from 9 buses to 1354 systems. The numerical analysis is determined in this paper and examines the important trends and the consequences that are consistent through all methods in light of the various tests. At the same time, it provides theoretical visions of phenomena such as lack of performance, failure, excessive calculation times, etc. selecting applications in the real world, providing comprehensive discussions on open questions in DPFL research, indicating ten possible future directions. (Number of words: 9668).
1. Introduction
Linear power flow models are of decisive importance in energy systems accounts, taking into account intensive research and wide application throughout academic circles and industry, opening trillion markets and affecting every global consumer [1, 2, 3, 4]. The accuracy and mathematical efficiency of these linear methods are pivotal for energy and planning systems, especially systems with high inhabitants of renewable energy due to the rapid variable nature of the resulting energy flows. Thus, enhancing the accuracy and efficiency of linear power flow models is not just a nice technical improvement to ascending, but rather a significant progress towards the future of sustainable energy.
Data -based power flow lines (DPFL) appeared as a promising way to acquire high -resolution linear models under comfortable conditions, for example, no need to know the physical model of the power system. Thus it receives wide attention [5]. Although he is in the developing stage, DPFL has already implanted a large knowledge base. This two -part tutorial aims to provide a comprehensive examination of the DPFL approach.
The first part of this tutorial program [6] It offered a comprehensive classification and theoretical analysis of all current DPFL styles, including its mathematical foundations, analytical solutions, critical assessments of the capabilities of each method, restrictions, and application. This work works as a basic guide, which meets the needs of both beginners and experts (S): 0000-2027-5314 (M. Jia) within this field, as well as professionals from other specializations simply looking for reliable written techniques.
Despite the accuracy of theoretical analysis in [6]It has restrictions: When many ways of sin has strengths and/or similar weaknesses, it is almost impossible to predict their differences in terms of practical performance. Thus, with only [6]Determining the most appropriate way to meet the specific needs is difficult. More importantly, the current numerical comparisons in literature do not fully show the full image in terms of the actual performance of DPFL methods. The lack of a clear understanding of the actual differences in performance between the current DPFL methods can hide problems that are not clear from the theoretical analysis of capabilities and restrictions, obscuring the rule of researchers within the DPFL community, and the complexity of the choice of appropriate sin is ways for potential users of other research areas.
In fact, the implementation of a comprehensive comparison requires great efforts, due to the lack of open source symbols for more than 95 % of relevant literature. However, in order to clarify mystery, future research paths determine and benefit from society, this paper, as the second part of the tutorial program, plans to fill this gap. Specifically, this paper performs comprehensive simulations for all DPFL methods, some of the newly introduced DPFL methods to display the normative nature of DPFL, and many methods of energy -based energy flow (PPFL) as standards, which total 44 methods. The main focus of this paper is a comprehensive assessment of these methods in terms of generalization, application, accuracy and mathematical efficiency. The evaluation results also support the determination of possible future trends. Thus, the contributions of this paper are three times:
(I) A comprehensive review of the current DPFL experiments, the study of the approved test scenarios, the loading fluctuation settings, the data sources, and the data of the data/the extremist values. The review also provides an overview of the current comparisons made between DPFL approaches, and determines the capabilities and restrictions of previous experiences, and clarifies the critical need for a comprehensive comparison of all DPFL methods.
(2) A comprehensive digital simulation is made for 44 written methods, including 36 DPFL approaches, four newly developed DPFL methods, and four classic PPFL algorithms. A detailed comparative analysis is provided to these 44 methods, where it discusses the ability to generalize, apply, accurately and mathematical efficiency, and thus clarify the actual performance of all the evaluated methods.
(3) An in -depth discussion is provided on open search questions, which shows ten promising but difficult future trends for DPFL research, which reported the numerical results acquired here and the theoretical conclusions extracted from the first part of the tutorial program [6].
The remaining part of this paper is organized as follows: The second section offers 44 methods. The third section reviews the current experiments in DPFL. The fourth section holds the methods related to its organization and application. Details of the fifth section, numerical assessments in terms of accuracy and mathematical efficiency. The Sixth Section discusses open questions in DPFL, and summarizes possible future trends. The seventh section concludes the paper.
note: We have made every effort to repeat the methods shown in the original research papers accurately. However, due to factors such as the lack of an open source code (with very few exceptions), the details are often incomplete in the literature, we cannot make sure that our applications completely reflect the original authors ’intentions, although the details were especially mysterious, we have done Even by developing multiple versions of the roads, as shown in Table 1 in the next section. However, we acknowledge that it is impossible to create accurate symmetrical copies of methods as its creators imagine. In addition, it is important to note that there is no way without defects. It is not intended to analyze the restrictions in this paper as a criticism, but as part of a comprehensive evaluation in light of certain cases with certain lying devices.
This paper is available on Arxiv under the CC BY-NC-ND 4.0 license (Noncommercial-Noderivs 4.0 International).