Power grid planning and optimal dispatch based on intelligent optimization and genetic algorithm

Power grid planning is a prerequisite for optimal design, operation, and planning of the power grid. Based on the characteristics of the power system, this article explored a genetic algorithm-based power network optimization method. In genetic algorithms, simulated annealing algorithms were used to optimize candidate individuals, resulting in the optimal ratio of the number of selected individuals to the number of all mutated individuals. At the same time, a simulated annealing algorithm was added to the optimization process to improve the shortcomings of GA (Genetic Algorithm). When the genetic algorithm selected individuals from the parent population, a simulated annealing algorithm was used to optimize candidate individuals, making the ratio of the number of selected individuals to the number of all mutated individuals optimal. In order to overcome the shortcomings of GA, the tabu search algorithm was introduced into the optimization process, making it have better optimization ability. Through optimization experiments on 15 nodes, it was found that this method had good convergence speed. The final experiment also proved that the power planning based on the simulated annealing genetic algorithm was superior to other models (the load balance, power supply capacity matching, and coordination index scores of the substation based on the algorithm model in this article were 0.5111, 2.2693, and 120.215, respectively, which were higher than other schemes).


Introduction
To achieve optimal power supply for multi-source power systems, Younes et al. [1] proposed a memory-based gravity search MBGSA (Modified Binary Grey Wolf Optimization Algorithm) method.They applied it to microgrid economic load dispatch.The existing Metaheuristic algorithm and GSA (Gravitational Search Algorithm) both face problems such as slow acquisition speed, small storage space, inability to store the optimal proxy location for the optimal solution, and poor effectiveness in solving complex optimization problems.This algorithm is based on Newton's law of gravity and calculates a new substitution factor by retaining the best result from the previous iteration process.Younes et al. [1] optimized the system using multi-mode distributed power sources, targeting diverse power sources such as photovoltaic power generation, cogeneration, and diesel generator sets.His research results were compared with classic quadratic programming and other metaheuristic algorithms such as GSA, artificial bee colony, genetic algorithm, particle swarm optimization, etc.Compared with other algorithms, his proposed algorithm had better performance in solving the optimal generation problem [1].Guo et al. [2] proposed a new approach to the economic dispatch of smart grids.In recent years, a large number of scholars have conducted extensive research on this issue and developed various efficient solving algorithms.However, most existing algorithms currently face the problem of high computational complexity.Therefore, Guo et al. [2] adopted the optimal method based on machine learning to study the problem.The core idea of this method is to treat the optimal solution as an unknown mapping relationship.On this basis, based on the classic ED (economic dispatch) optimization method, they used DNN (deep neural network) to learn the optimization strategy of ED and applied it to online ED.The innovation was that they accurately approximated a class of classical stochastic differential formulas using a well-developed DNN model with a certain scale.Finally, they also conducted case analysis on three-unit power grids and IEEE-30 bus power grids to prove the correctness of the proposed algorithm [2].
The rapid development of computer technology has made large-scale computing possible.Therefore, how to achieve global optimal design while ensuring reliability and cost-effectiveness is an urgent problem to be solved.This article introduced a multi-objective optimization algorithm based on a genetic algorithm.Heuristic algorithms are based on intuition, but they have their own shortcomings and are difficult to provide globally optimal results.In nonlinear and discontinuous situations, genetic algorithms can find the optimal solution within the globally permissible solution space, thereby avoiding falling into local optima.This method can effectively overcome the "dimensional chaos" and "optimization" problems existing in traditional methods [3].

Mathematical model for power grid planning
The application of genetic algorithms in the constrained economic dispatch of power systems requires obtaining an initial feasible solution that satisfies both unit constraints and power balance constraints [4].To effectively improve the initial search interval of GA, it is necessary to determine a reasonable unit combination while meeting the needs of peak shaving and frequency regulation in actual power grid operation.The consumption characteristic of thermal power units is a quadratic function characteristic curve.When low load is met, after the optimal load distribution of each generator unit in the grid, the optimal operating output of each generator unit should not be lower than the optimal distribution load point of the previous period as the grid load increases, that is, when the grid load rises from low valley Pp to Pp3 (PDx>Por), any unit Pu+1≥P, thus effectively adapting to the growth of grid load.The initial feasible region's lower limit of GA search is adaptively increased, reducing the lower limit value of GA initial search interval and improving GA search efficiency.Similarly, the same strategy is also adopted for power grid load reduction [5].
The distribution network planning problem is mainly a large-scale nonlinear planning problem.The nonlinear programming problem mainly includes two parts: objective function and constraint conditions.In terms of objective function and constraint conditions, the selection of mathematical models greatly affects the determination of the optimal solution.This article takes the minimum annual planning cost of the power grid as the objective function, and the expression of the mathematical model is: where t Z cos is the annual planning cost; 1 f is the annual operating cost of the network; 2 f is the cost of network loss; n is the total number of installed lines; Among them, i a represents the depreciation and maintenance cost rate of equipment; i T is the investment cost for the new branch road i; i L is the erected line.When branch i is newly created, it is taken as 1, otherwise, it is taken as 0. (1) Node voltage inequality constraint: It is required that the voltage drop must meet the given requirements and meet the standard of safe power supply.
where u K is the node voltage penalty factor, which serves as a penalty for deviating from the operating limit; i U is the voltage of node i; (2) Inequality constraint of wire current: It is required that the long-term current carrying capacity of the wire should be less than the rated current carrying capacity of the wire.
where i K is the penalty factor for wire current; i I is the current of branch i; hi I is the current carrying capacity of branch i wire.
(3) Substation capacity to load ratio constraint: The constraint formula is:

Power grid scheduling experiment based on simulated annealing genetic algorithm
This experiment aimed to select the power grid planning of a certain province as the experimental object, optimize the power grid nodes and lines based on the actual geographical distribution of the power grid, and validate the proposed method.This railway has 11 nodes, 9 nodes, and will be expanded to 15 nodes in the future, as shown in Figure 1.Under these conditions, a total of 18 new pathways were obtained, namely 18 chromosome segments, n=100 chromosomal regions, p=0.5 crossover factors, pm=0.05,threshold C=1.3xl0'.On this basis, a new manufacturing process plan was proposed, assuming that the construction costs of each production line are the same, which can be used to calculate the construction costs of a new production line [6].
The genetic algorithm was compared with the simulated annealing genetic algorithm (Simulated Annealing GA).In Figures 1 and 2, the average moderate values and maximum fitness curves of each chromosome segment were shown as a function of the number of iterations.Through the analysis of "fitness", a new "adaptive" method was proposed, which can effectively achieve "adaptation" and "optimization" [7].As shown in Figure 2, from the variation curve of the average fitness value, it can be seen that the model based on the simulated annealing genetic algorithm did not produce oscillations and had fast convergence, indicating that this method can accelerate convergence, eliminate oscillations, and accelerate approximation to the global optimum.From the above results, it can be seen that during the iteration process, as the degree of evolution increases, the increase in the average value directly affects the maintenance of the optimal ranking [8].
From the convergence curve in Figure 2, it can be seen that before the improvement of the genetic algorithm, there were significant fluctuations, resulting in a longer solution time and a significant deviation from the true value.The model based on a simulated annealing genetic algorithm has the advantages of fast termination of fluctuations, fast convergence speed, and approaching the true value; that is, the improved GA can quickly converge to the optimal solution, while the traditional GA is prone to oscillations and falls into local optima during the solution process [9].From Figure 3, it can be seen that the experiment compared the construction cost, coal consumption cost, power shortage cost, and economic indicators of different methods for power grid dispatch planning.The three costs of the power grid scheduling model based on a simulated annealing genetic algorithm were much lower than traditional power grid planning schemes.This is mainly due to the traditional design mode based on economy and reliability.In contrast, the integrated design mode considers the balance of the power grid, which limits the alternative design schemes during optimization, thereby increasing the cost of power grid construction.Introducing the equilibrium index into power system planning can transform the optimization of the power system from a single economic optimal to an integrated economic and collaborative approach.On this basis, the simulated annealing genetic algorithm and GA algorithm introduce an equilibrium index, transforming the optimization objective from simple economic optimization to comprehensive economic and collaborative optimization while improving the balance and overall performance of the power grid, making its overall performance better.Compared to traditional power grid planning, the three costs of the power grid scheduling model based on the GA algorithm have also been reduced.However, the degree of cost reduction is not as high as that of the simulated annealing genetic algorithm.Therefore, in the comparative analysis of economic indicators, the power grid scheduling model based on the simulated annealing genetic algorithm was 76.46, which was the highest.The electricity shortage cost used to characterize the reliability of the power grid has decreased from the traditional 20934600 yuan to 1345300 yuan using a simulated annealing genetic algorithm, a decrease of 35.74%, indicating that the introduction of equilibrium indicators significantly improves the overall balance ability of the power grid and the reliability of the system.When the load balance of the substation is high, it means that the load distribution of each power equipment is relatively uniform, and the load difference between equipment is small.At the same time, high load balancing can also fully utilize the capacity of devices, improving their efficiency and utilization.The load rate balance of the line reflects whether the load distribution of each line is reasonable and whether the capacity and performance of the line can be fully utilized.The matching degree of power supply capacity refers to the reasonable configuration and management of power supply capacity.In the power system, the coordination between various devices refers to their ability to cooperate and work together during operation.The purpose of evaluating coordination is to improve the reliability and stability of the power system and reduce the occurrence of faults and power outages.
From Table 1, it can be seen that the power grid scheduling model based on the algorithm proposed in this paper had an index of 0.0811 in terms of line load rate balance, which was lower than traditional power grid planning and GA algorithm models.This indicates that the power grid planning can cause unreasonable distribution of power loads on various lines, which may lead to adverse phenomena such as wire heating.Fortunately, the difference is not significant.The performance of power grid planning based on the algorithm proposed in this paper is superior to other schemes in other indicators [10].

Conclusions
With the progress of human society, the energy demand continues to increase, and fossil fuels are an important component of China's energy industry.They not only bring environmental pollution but also seriously constrain China's economic development.With the continuous increase in energy demand, reasonable allocation of power resources can improve the management efficiency of power enterprises.Based on the above issues, this article explored how to use the GA algorithm to construct a power grid dispatch model.This article believed that using the GA algorithm and simulated annealing algorithm can avoid the shortcomings of both and improve the algorithm's ability to optimize power allocation.The experiment has proven the effectiveness of the algorithm based on this paper, which is a relatively significant research topic.

i C 2
is the unit electricity price; i max  is the maximum load utilization hours; i S  is the active power loss of the branch.

K
is the penalty factor for the capacity-to-load ratio of the substation; G P is the total load capacity of the substation; max G P and min G P are the upper and lower limits of the total load capacity of the substation, respectively; L P is the current substation load capacity.

Figure 3 .
Figure 3. Economic part of improved algorithm planning results.

Table 1 .
Coordination of improved algorithm planning results.