Genetic Algorithm Strategy for Optimizing the Structure of the Voltage-Shaping Ring of a High-Voltage DC Divider

作者： GOZ Electric 时间：2024-12-11 09:04:29 阅读：21

Genetic algorithm is a numerical optimization method that describes the mechanism of natural selection and population inheritance. It has a solid biological basis, distinct cognitive significance, achievable parallel computing behavior and wide application value. In the genetic algorithm solution process, candidate solutions will be retained in each iteration and adaptively arranged according to certain characteristic factors such as the environment. Individuals will be selected, crossed and mutated according to their adaptability to the environment to generate new populations, which will then participate in a new round of genetic evolution as input. The essence of genetic algorithm is an iterative algorithm. If the structure of the genetic algorithm is too simple, it will be very sensitive to individuals with high fitness and will continue to reproduce them, resulting in local convergence. Finding the optimal individual in a non-uniform space reduces the real reliability of the genetic algorithm. This paper uses an improved SPRGA to improve the reliability of the genetic algorithm.

There is a nonlinear relationship between the field strength along the surface of the voltage divider and the parameters of the voltage divider. The neural network has multi-dimensional nonlinear mapping capabilities, and can obtain the weights and structures of the network by learning and training the input/output data to obtain the implicit relationship in the input/output data5. The genetic algorithm has a general algorithm framework that does not depend on the type of problem and does not need to know the specific expression of the objective function. Therefore, the optimization technology combining neural networks with genetic algorithms can be used to solve the optimization problem of the voltage divider from a mathematical level.

1 Constraints

The optimal optimized structural parameters of the voltage divider refer to the conditions under which the surface electric field of the voltage divider is less than its corona initiation field strength, and the maximum field strength along the surface of the DC high-voltage voltage divider reaches a minimum of 10.

2 SPRGA genetic algorithm

The basic idea of SPRGA to adjust individual fitness is the idea of niche genetics, which uses shared functions to adjust individual fitness, limit the large increase of individual individuals, and maintain population diversity. The genetic chromosome in this paper consists of 3 gene positions, and the position of each gene corresponds to the parameters of the equalizing loop: (omitted)

The elite retention strategy is adopted to retain the optimal solution in population evolution to the next generation. Suppose the number of elite individuals with the best fitness value retained is N, and the shared fitness of the remaining (M-N) non-elite individuals in the population is introduced, and the N individuals with the lowest shared fitness ranking are eliminated; then, the elite individuals retained in the previous step are input into the population, and another genetic operation is performed to merge the elite individuals with the population to obtain the size is the offspring population of M. If the number of elites N is too large, the number of elites will grow rapidly in the population, resulting in a decrease in search speed; if it is too small, it may lead to the local loss of excellent individuals, resulting in increased genetic inaccuracy. Therefore, N should be reasonably selected according to the size of the population. This paper takes N-10%M.

The generation of new individuals in genetic algorithms is mainly assisted by crossover operations and mutation operations. In order to avoid recent reproduction in SPRGA, when performing crossover operations, individuals with the maximum Euclidean distance are selected to cross the crossover mother, and the traditional uniform mutation algorithm is improved to improve the local search ability of the algorithm. According to the set mutation probability, it is compared with a random number between 0 and 1 to decide whether to jump to subtract or add mutations to the gene.

3 Optimization calculation

3.1 Electric field optimization of equalizing ring

This paper calculates the electric field distribution of ±500kV DC voltage divider with and without equalizing ring. The electric field magnitudes before and after the installation of the equalizing ring are 28.397 and 2.835kV/mm respectively. The position of the maximum field strength of the voltage divider is transferred from the end of the voltage divider cover to the outer surface of the equalizing ring. It can be seen that the installation of the equalizing ring can significantly reduce the maximum field strength near the high-voltage electrode. However, the initial equalizing ring parameters cannot make the maximum field strength on the voltage divider below the corona initiation field strength.

3.2 Optimization of the size and position of the equalizing ring

The change of the electric field strength on the surface of the DC voltage divider with the height h of the equalizing ring from the ground. As shown in the figure, the maximum field strength on the surface of the voltage-equalizing ring does not change with the change of h; while the maximum field strength on the surface of the voltage divider changes with the increase of h. When the maximum field strength appears at the sharp corner of the high-voltage copper rod, the field strength gradually decreases with the increase of h, and when the maximum field strength appears at the upper edge of the voltage divider cover, the field strength increases with the increase of h. Therefore, if the maximum field strength along the surface of the DC high-voltage voltage divider is to be minimized, h should be within the range of 4840~4857mm.

Figure (omitted) shows the change of the electric field strength on the surface of the DC voltage divider with the radius R of the voltage-equalizing ring. As shown in Figure 7, the maximum field strength on the surface of the voltage-equalizing ring decreases very little, but the maximum field strength on the surface of the voltage divider appears on the voltage divider body due to the increase of R, and may even cause corona discharge. Therefore, improving the electric field of the voltage-equalizing ring by changing R cannot achieve significant results. If the maximum field strength along the surface of the DC high-voltage voltage divider is to be minimized, R should be within the range of 430~440mm.

Figure (omitted) shows the variation of the electric field intensity on the surface of the DC voltage divider with the radius r of the voltage-equalizing ring. As shown in Figure 8, the maximum field intensity on the surface of the voltage-equalizing ring shows a decreasing trend as r increases. Considering the actual installation and construction difficulty and manufacturing cost, if the maximum field intensity along the surface of the DC high-voltage voltage divider is to be minimized, r should be in the range of 138~148mm.

Conclusion

In order to rationalize the structure of the voltage divider, this paper combines neural networks with genetic algorithms to optimize the structural parameters of the voltage-equalizing ring. The traditional genetic algorithm and SPRGA are used to optimize the structural parameters, and it is found that the optimization results obtained by the latter are better and the electric field value is lower. The method in this paper improves the use margin of the equipment and provides a new idea for optimizing the structural parameter design of DC power transmission projects.