MULTIOBJECTIVE OPTIMIZATION IN REACTIVE POWER COMPENSATION
Relevance. Currently, Russia is creating an intelligent power system with an actively-adaptive network – IES AAS (abroad – Smart Grid). The basic Smart Grid group architecture is FACTSdevices; the complexity is the multi-criterion nature of the problem. Reactive power optimization is a secondary p...
| Format: | Article |
|---|---|
| Language: | Russian |
| Published: |
Tomsk Polytechnic University
2018-12-01
|
| Series: | Известия Томского политехнического университета: Инжиниринг георесурсов |
| Subjects: | |
| Online Access: | http://izvestiya.tpu.ru/archive/article/view/25 |
| _version_ | 1797811155605913600 |
|---|---|
| collection | DOAJ |
| description |
Relevance. Currently, Russia is creating an intelligent power system with an actively-adaptive network – IES AAS (abroad – Smart Grid). The basic Smart Grid group architecture is FACTSdevices; the complexity is the multi-criterion nature of the problem. Reactive power optimization is a secondary problem of the optimal power flow, when the setting of the correct reactive power variables, such as values of voltage, transformer position stages and reactive power characteristics of compensation devices, is determined. The solution for the problems to reactive power optimization, which are not linear and discrete, using traditional optimization methods is accompanied by certain difficulties associated with the processing of data that has different nature. Therefore, at the present, an adequate method of multi-object data processing is being searched, for example, using the evolutionary optimization algorithm. The aim of the research is to develop a mathematical method to find an optimal solution from the whole set of possible ones, which would be better than others for at least one objective. In this case, the model must perform a calculation of the power flux at the fundamental and harmonic frequency for a particular mode, with a large number of restrictions. Methods. The simulation modeling of the FACTS device implementation was carried out in a program called DYCSE. We used the random search algorithm, which is a variable integration method modification and allows solving convergence problems when it is applied to a very large data set. The calculation method and the results of the study of Arzola Ruiz Jose As were taken as an example and the basis for developing our method. The Chebyshev method was used in the objective function. This methods allows reducing the weighted distance from the calculated value to the desired one of each indicator included in the objective function. It is obvious that a population with a high level of initial data represents the best solution to the problem and, under certain conditions, it can present even the only optimal solution. From the initial generation of potential solutions for the process that is repetitive, the new generations of solutions were derived, each time with better characteristics approaching the optimal solution of the problem. The criteria for stopping the calculation were a mixed condition – the difference between the worst and the best decisions. Each experiment was performed with the initial population that has a random character, obtaining acceptable solutions about 7 % according to the estimated value. Results. The use of evolutionary methods in optimization allows simultaneous consideration of several independent solutions, creating a set of so-called optimally effective solution or Pareto solutions that satisfy the research objectives. In all the experiments, the effective solutions were obtained to estimate the population size about 10 % of all possible solutions. The obtained solutions can be considered effective in comparison with the calculations that could be performed with absolutely all required initial data and full-scale calculations performed. Conclusions. To achieve energy efficiency in industrial networks, new optimization methods that improve the technical and economic performance of networks are required. The use of the Chebyshev method, which makes it possible to reduce the weighted distance from the calculated value to the desired one of each indicator included in the objective function, was tested obtaining satisfactory results in theoretical and practical studies. It is proved that for values close to 10 % of the spectrum of all possible solutions we can obtain solutions that satisfy the search for effective solutions, and that coincide with the recommendations proposed by Arzola. The developed algorithm significantly reduces the computation time, with results convergence guarantee.
|
| first_indexed | 2024-03-13T07:19:32Z |
| format | Article |
| id | doaj.art-c5abd9025929477d94545722c2cb2156 |
| institution | Directory Open Access Journal |
| issn | 2500-1019 2413-1830 |
| language | Russian |
| last_indexed | 2024-03-13T07:19:32Z |
| publishDate | 2018-12-01 |
| publisher | Tomsk Polytechnic University |
| record_format | Article |
| series | Известия Томского политехнического университета: Инжиниринг георесурсов |
| spelling | doaj.art-c5abd9025929477d94545722c2cb21562023-06-04T21:06:44ZrusTomsk Polytechnic UniversityИзвестия Томского политехнического университета: Инжиниринг георесурсов2500-10192413-18302018-12-013291210.18799/24131830/2018/12/25MULTIOBJECTIVE OPTIMIZATION IN REACTIVE POWER COMPENSATION Relevance. Currently, Russia is creating an intelligent power system with an actively-adaptive network – IES AAS (abroad – Smart Grid). The basic Smart Grid group architecture is FACTSdevices; the complexity is the multi-criterion nature of the problem. Reactive power optimization is a secondary problem of the optimal power flow, when the setting of the correct reactive power variables, such as values of voltage, transformer position stages and reactive power characteristics of compensation devices, is determined. The solution for the problems to reactive power optimization, which are not linear and discrete, using traditional optimization methods is accompanied by certain difficulties associated with the processing of data that has different nature. Therefore, at the present, an adequate method of multi-object data processing is being searched, for example, using the evolutionary optimization algorithm. The aim of the research is to develop a mathematical method to find an optimal solution from the whole set of possible ones, which would be better than others for at least one objective. In this case, the model must perform a calculation of the power flux at the fundamental and harmonic frequency for a particular mode, with a large number of restrictions. Methods. The simulation modeling of the FACTS device implementation was carried out in a program called DYCSE. We used the random search algorithm, which is a variable integration method modification and allows solving convergence problems when it is applied to a very large data set. The calculation method and the results of the study of Arzola Ruiz Jose As were taken as an example and the basis for developing our method. The Chebyshev method was used in the objective function. This methods allows reducing the weighted distance from the calculated value to the desired one of each indicator included in the objective function. It is obvious that a population with a high level of initial data represents the best solution to the problem and, under certain conditions, it can present even the only optimal solution. From the initial generation of potential solutions for the process that is repetitive, the new generations of solutions were derived, each time with better characteristics approaching the optimal solution of the problem. The criteria for stopping the calculation were a mixed condition – the difference between the worst and the best decisions. Each experiment was performed with the initial population that has a random character, obtaining acceptable solutions about 7 % according to the estimated value. Results. The use of evolutionary methods in optimization allows simultaneous consideration of several independent solutions, creating a set of so-called optimally effective solution or Pareto solutions that satisfy the research objectives. In all the experiments, the effective solutions were obtained to estimate the population size about 10 % of all possible solutions. The obtained solutions can be considered effective in comparison with the calculations that could be performed with absolutely all required initial data and full-scale calculations performed. Conclusions. To achieve energy efficiency in industrial networks, new optimization methods that improve the technical and economic performance of networks are required. The use of the Chebyshev method, which makes it possible to reduce the weighted distance from the calculated value to the desired one of each indicator included in the objective function, was tested obtaining satisfactory results in theoretical and practical studies. It is proved that for values close to 10 % of the spectrum of all possible solutions we can obtain solutions that satisfy the search for effective solutions, and that coincide with the recommendations proposed by Arzola. The developed algorithm significantly reduces the computation time, with results convergence guarantee. http://izvestiya.tpu.ru/archive/article/view/25Multi-objective optimizationreactive power compensationFACTS-devicesSmart Gridsevolutionary algorithmsgenetic algorithms |
| spellingShingle | MULTIOBJECTIVE OPTIMIZATION IN REACTIVE POWER COMPENSATION Известия Томского политехнического университета: Инжиниринг георесурсов Multi-objective optimization reactive power compensation FACTS-devices Smart Grids evolutionary algorithms genetic algorithms |
| title | MULTIOBJECTIVE OPTIMIZATION IN REACTIVE POWER COMPENSATION |
| title_full | MULTIOBJECTIVE OPTIMIZATION IN REACTIVE POWER COMPENSATION |
| title_fullStr | MULTIOBJECTIVE OPTIMIZATION IN REACTIVE POWER COMPENSATION |
| title_full_unstemmed | MULTIOBJECTIVE OPTIMIZATION IN REACTIVE POWER COMPENSATION |
| title_short | MULTIOBJECTIVE OPTIMIZATION IN REACTIVE POWER COMPENSATION |
| title_sort | multiobjective optimization in reactive power compensation |
| topic | Multi-objective optimization reactive power compensation FACTS-devices Smart Grids evolutionary algorithms genetic algorithms |
| url | http://izvestiya.tpu.ru/archive/article/view/25 |