Power grid remote monitoring control method and system
By combining particle swarm algorithm and genetic algorithm, optimize parameters and parallel processing, and dynamically adjust algorithm parameters, the problems of low efficiency and high power consumption in power grid fault reconstruction are solved, and fast and accurate fault recovery and low power loss are achieved.
Patent Information
- Application Number
- CN202411007465.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-25
- Publication Date
- 2024-12-06
AI Technical Summary
The existing grid fault reconstruction algorithm is inefficient and has high power consumption, so it cannot quickly and accurately restore power supply. Especially in complex power systems, traditional methods rely on manual inspection and have problems such as local interference and difficulty in parameter adjustment.
Using a combination of particle swarm algorithm and genetic algorithm, through optimized parameter adjustment and parallel processing, the minimum power loss is used as the objective function, combined with linear or nonlinear decreasing inertial weights and learning factor updates, dynamically adjust the probability of crossing and mutation to achieve rapid Fault location and recovery.
It improves the efficiency and accuracy of fault reconstruction, reduces the power loss of the distribution network, achieves fast and reliable fault recovery, and reduces dependence on human resources.
Smart Images

Figure CN119093580A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a method and system for remote monitoring and control of a power grid, and belongs to the technical field of power system monitoring. Background Art
[0002] With the continuous progress of my country's socialist industrial modernization process, the demand for electricity in the industrial field and people's daily life is also increasing, and the complexity of the power system structure is becoming more and more complex. However, due to various reasons such as equipment aging, weather disasters, and human operation errors, the circuit system may fail, resulting in power outages and unstable power consumption. Therefore, the fault safety of the power system has become an issue that relevant institutions and staff must consider. These issues include the detection and filling of grid loopholes, remote monitoring of the power system, the location judgment of remote tripping faults in the power structure, the timely repair of power outages due to faults and the efficiency of restoring power supply, as well as the maintenance of the power grid in severe weather. Harmful factors such as device errors and quality losses will make its monitoring and maintenance work extremely difficult. In addition, there will be problems such as failures caused by human errors and line damage. The traditional fault handling method mainly relies on manual inspections, troubleshooting and manual operation to restore power supply. However, this method has many limitations and challenges, such as long fault location time, high operational risks, and heavy reliance on human resources. With the expansion of the scale of the power system and the increase in network complexity, the traditional fault handling method can no longer meet the needs of fast, accurate and reliable fault recovery. Therefore, researchers began to focus on the study of distribution network fault reconstruction algorithms for remote tripping of control loops.
[0003] The distribution network fault reconstruction algorithm aims to achieve rapid and reliable recovery of the distribution network by using modern information and communication technologies to monitor, identify and locate distribution network faults in real time and automatically perform fault recovery. By taking effective optimization measures, the faulty distribution network can be reconstructed, which can not only quickly restore it to normal, but also effectively reduce its power loss. The core idea of this technology is to ensure that specific constraints are met by taking effective optimization schemes, so as to quickly obtain the optimal solution.
[0004] At present, the algorithms used to solve the fault reconstruction of distribution networks include branch exchange method, optimal power flow model method, simulated annealing algorithm, particle swarm algorithm, ant colony algorithm, genetic algorithm, expert system method, neural network method, fuzzy computing, greedy computing, etc. Among them, the particle swarm algorithm (PSO) is characterized by its simple operation, rapid convergence, strong local search ability, and fewer parameters. It can solve complex discrete optimization problems, but there are difficulties in parameter adjustment. If the adjustment is improper, the particle swarm algorithm will be locally disturbed; in contrast, the genetic algorithm is more flexible and robust in solving complex problems, and can more accurately predict future development trends. By integrating the advantages of the particle swarm algorithm and the genetic algorithm, and adaptively adjusting the parameters such as inertia weight, crossover probability, and mutation probability, and processing the data and iterative tasks in parallel, the convergence efficiency and solution efficiency of the algorithm are accelerated.
[0005] For example, the Chinese patent application document with application publication number CN115906914A discloses an improved particle swarm hybrid algorithm for fault location in distributed power distribution networks. The method integrates the particle swarm algorithm and the genetic algorithm. The particle swarm algorithm is first used to initialize the population and calculate the fitness. If the random number is greater than the set threshold, the genetic algorithm is entered for iteration. The iteration speed, accuracy and fault tolerance of fault location are provided by the fusion of the two algorithms. Although this scheme improves the efficiency and accuracy to a certain extent, since it enters the genetic algorithm only when the random number is greater than a set value, it means that the update of the genetic algorithm is not performed in each iteration cycle, but only starts after meeting specific conditions. This approach may cause the algorithm to fail to use the global search capability of the genetic algorithm to optimize the population in a timely manner, thereby affecting the calculation efficiency of the optimal solution. In other words, this update mechanism based on random trigger conditions may delay the evolution of the population, making the algorithm not flexible and timely enough in the process of searching for the optimal solution. In addition, the design of its fitness function does not consider the power loss problem. In the study of power system fault reconstruction, power loss is a very important indicator because it is directly related to the efficiency and cost of power grid operation. The fitness function is a key factor in guiding the algorithm's search process. If the fitness function does not correctly reflect the actual needs of the problem, then even if the algorithm can find solutions, these solutions may not be optimal or most needed in practical applications. Summary of the invention
[0006] The purpose of the present invention is to provide a method and system for remote monitoring and control of a power grid, so as to solve the problems of low efficiency and high power consumption in the current power grid fault reconstruction process.
[0007] In order to solve the above technical problems, the present invention provides a remote monitoring and control method for a power grid, which comprises the following steps:
[0008] 1) Taking the minimum power loss in the distribution network as the objective function;
[0009] 2) generating an initial population according to the topological structure of the distribution network, iteratively updating the population using a particle swarm algorithm, and calculating the fitness value of each individual and the global optimal fitness value of each generation during the iterative updating process; the fitness function used by the fitness value is the objective function;
[0010] 3) When the fitness value of an individual is better than the global optimal fitness value, the genetic algorithm is used to update the population to generate a new generation of population, and the particle swarm algorithm is used to iteratively update the generated new generation of population until the termination condition is met; the population with the best fitness value at the end of the iteration is taken as the optimal solution to control the switches in the distribution network.
[0011] Furthermore, in step 2), when the particle swarm algorithm is used to iteratively update the population, the inertia weight and the learning factor are updated in a linear decreasing or nonlinear decreasing manner to achieve the update of the individual speed and position.
[0012] Furthermore, the update formula of inertia weight and learning factor in nonlinear decreasing mode is:
[0013]
[0014] ω is the inertia weight factor, c 1 ,c 2 is the learning factor, t is the current number of iterations, and t max is the total number of iterations, ω max and ω min are the maximum and minimum values of the inertia weight factor, c max and c min are the maximum and minimum values of the inertia weight factor, respectively.
[0015] Furthermore, when the particle swarm algorithm is used to iteratively update the population, the population is divided into several sub-groups, each sub-group performs calculations on different processors, updates its particle positions and velocities in parallel, and information is exchanged between the sub-groups.
[0016] Furthermore, the objective function is:
[0017]
[0018] Where P loss is the power loss of the distribution network, N b is the total number of branches in the distribution network, R i For branch b i The branch resistance; P i For flow through branch b i The active power, fi b is the current flowing through the branch i Reactive power; U i For branch b i The end node voltage.
[0019] Furthermore, when performing crossover and / or mutation operations, the crossover probability and / or mutation probability are adjusted according to the fitness value or the evolution speed. When the fitness value is high or the evolution speed is fast, the crossover probability and / or mutation probability are reduced to strengthen local search; when the fitness value is low or the evolution speed is slow, the crossover probability and / or mutation probability are increased.
[0020] Furthermore, the method also includes adjusting the population size according to the fitness value or the evolution speed. When the fitness value is high or the evolution speed is fast, the population size is reduced; when the fitness value is low or the evolution speed is slow, the population size is increased.
[0021] Furthermore, the termination condition is any one of the following conditions: reaching the maximum number of iterations, the fitness no longer improves, the quality of the current solution reaches a preset accuracy, and the computing resources are exhausted.
[0022] The present invention also provides a power grid remote monitoring and control system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned power grid remote monitoring and control method when executing the computer program.
[0023] The beneficial effects of the present invention are as follows: the present invention firstly utilizes the strong local search capability of the particle swarm algorithm to obtain the local optimal solution, and when the fitness value of an individual is better than the global optimal fitness value, the genetic algorithm is used to update the population, a new generation of population is generated, and the particle swarm algorithm is used to iteratively update the generated new generation of population, making full use of the search capability of the genetic algorithm and further improving the search efficiency; at the same time, the present invention uses the minimum power loss in the distribution network as the fitness. Therefore, the present invention can quickly perform fault reconstruction, restore power supply to the fault area, and minimize the power loss of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] Figure 1 is a flow chart of the power grid remote monitoring and control method of the present invention;
[0025] Figure 2 is a schematic diagram of the IEEE33 distribution network system model targeted by the embodiments of the present invention;
[0026] Figure 3 It is a comparison chart of simulation results of the relationship between the number of iterations and power loss of the present invention and the existing algorithm when the population size is 30;
[0027] Figure 4 It is a comparison chart of simulation results of the relationship between the number of iterations and power loss of the present invention and the existing algorithm when the population size is 300. DETAILED DESCRIPTION
[0028] The specific implementation of the present invention will be further described below in conjunction with the accompanying drawings.
[0029] Grid remote monitoring and control method embodiment
[0030] The present invention first uses the strong local search capability of the particle swarm algorithm to find the local optimal solution. When the fitness value of an individual is better than the global optimal fitness value, the genetic algorithm is used to update the population to generate a new generation of population. The particle swarm algorithm is used to iteratively update the generated new generation of population, making full use of the search capability of the genetic algorithm and further improving the search efficiency. At the same time, the present invention uses the minimum power loss in the distribution network as the fitness. Therefore, the present invention can quickly perform fault reconstruction, restore power supply to the faulty area, and minimize the power loss of the distribution network. The implementation process of this method is as follows: Figure 1 As shown, the detailed description is given below.
[0031] 1. Establish the objective function and use it as the fitness function.
[0032] The present invention aims to reduce the power loss in the distribution network and realizes the reconstruction of the distribution network after the failure. The power loss of the distribution network mainly includes the copper loss and iron loss of the transformer and the power loss of the wire. The present invention only considers the power loss of the wire, and the mathematical model of reconstruction is:
[0033]
[0034] Where P loss is the power loss of the distribution network, N b is the total number of branches in the distribution network, R i For branch b i The branch resistance; P i For flow through branch b i The active power, f i b is the current flowing through the branch i Reactive power; U i For branch b i The end node voltage.
[0035] The above reconstructed mathematical model is used as the objective function, and the objective function is used as the subsequent fitness function. The objective function mainly has the following two functions:
[0036] 1) Evaluate the quality of the solution: The objective function can be used to evaluate the quality of each solution to determine whether it meets the goal of reducing the distribution network. By calculating the objective function value, the advantages and disadvantages of different solutions can be compared, so as to select the best reconstruction scheme.
[0037] 2) Search process guidance: The objective function can guide the direction of the algorithm during the search process, and can guide the algorithm to search towards a better solution space, thereby accelerating the convergence speed and solution efficiency of the algorithm.
[0038] In addition, the objective function must meet the following specific conditions: there must be no loops or isolated parts in the circuit to avoid islands; many wrong circuit topologies may be generated when using a completely random method for initialization and iteration. If the topological constraints can be taken into account in the process of particle swarm initialization and update, all particles can be guaranteed to be in a valid state, while also narrowing the search range of the particle swarm and reducing the possibility of local convergence. If all switches are turned on, a series of loops will be formed, thus avoiding the appearance of a loop structure; only one switch needs to be turned on in different loops to break the loop structure, and around a node with n loops, only n-1 switches need to be disconnected at most, thus avoiding the appearance of islands.
[0039] According to this embodiment Figure 2 The network structure uses loop-based decimal coding to encode the nodes of the system. When solving the objective function, the following conditions must also be met:
[0040] |I bi |<|I bi | max
[0041] S i <S imax
[0042] U imin <U i <U imax
[0043] Where |I bi | max For branch b i The maximum current amplitude allowed to pass through, U i is the voltage amplitude at node i, U imin , U imax are the minimum and maximum voltage amplitudes allowed at node i when the power grid is operating normally, S i is the load of the transformer in the distribution network, with a maximum capacity of S imax .
[0044] The calculation of current and voltage for distribution network reconstruction requires flow calculation, which means solving a set of multivariate nonlinear equations and requires iterative solution. ij have:
[0045] U j =U i -I ij C ij
[0046] Among them U j It is branch b ij The node voltage at the end, U i For branch b ij The first node voltage, I ij is flowing through branch b ij The branch current, C ij For branch b ij branch impedance.
[0047] If node j is branch b ij The load current I of node j is j for:
[0048]
[0049] Where P j is the active power of node j, f j is the reactive power of node j, and the conjugate of the voltage of node j is
[0050]
[0051] If node j is not branch b ij The end point of ij is the load current I at node j j The sum of the currents of all other branches with node j as the head end node:
[0052]
[0053] Where n is the set of another node in the branch connected to node j, and q is a node in the set n.
[0054] The structure of the distribution network can be divided into three stages through power flow calculation: forward push, back substitution and convergence. Forward push refers to the derivation from the root node to the terminal node to obtain the circuit information of each branch in the distribution network, while back substitution refers to the recursion from the root node to the terminal node to obtain the voltage information of each node in the distribution network.
[0055] 2. Generate the initial population according to the topological structure of the distribution network, use the particle swarm algorithm to iteratively update the population, and calculate the fitness value of each individual and the global optimal fitness value of each generation during the iterative update process.
[0056] This embodiment uses the particle swarm algorithm (PSO) to solve the above objective function. The algorithm is inspired by the foraging of birds and has the advantages of simple method, fast convergence speed, strong local search ability and few parameters. The basic idea of the particle swarm algorithm is to update its own speed v and position x in each iteration according to the individual extreme value and the group extreme value to track the optimal example and obtain the optimal solution. When the algorithm is applied to the reconstruction problem of distribution network faults, the particle speed introduces the concepts of inertia weight and learning factor, and the speed and position update formulas are as follows:
[0057]
[0058] where i = 1, 2, ..., M; j = 1, 2, ..., m; t = 1, 2, ..., T max , v ij is the velocity of the jth dimension of the ith particle; γ is the velocity authority factor; ω is the inertia weight factor, c 1 ,c 2 is the learning factor; r 1 ,r 2 is a random number between [0,1], p ij is the jth dimension of the current optimal position of the ith particle; x ij is the jth dimension of the current position of the ith particle; M is the total number of particles in the particle swarm, m is the dimension of each particle; t is the current number of iterations, T max is the total number of iterations.
[0059] The initial population is randomly generated based on the topological structure of the distribution network. The topological structure of the distribution network includes grid nodes, connection relationships between nodes, and locations of switches. This embodiment uses the random state of the switch to generate the initial population.
[0060] In each iteration, the speed and position of each individual are updated. Since the speed update is affected by the inertia weight and the learning factor, and the difference between the individual optimal solution and the global optimal solution is taken into account, the inertia weight and the learning factor should be updated using a linear decreasing method or a nonlinear decreasing method.
[0061]
[0062] ω is the inertia weight factor, c 1 ,c 2 is the learning factor, t is the current number of iterations, and tmax is the total number of iterations, ω max and ω min are the maximum and minimum values of the inertia weight factor, c max and c min are the maximum and minimum values of the inertia weight factor, respectively.
[0063] For inertia weight and learning factor, the closer the value is to 1, the more conducive it is to global search, and the closer it is to 0, the more conducive it is to local search. Since the particle swarm algorithm has strong local search ability but weak globality, the linear parameter adjustment method does not meet the global requirements of the algorithm. When x∈[0,1], The range of is higher than that of the direct proportional function, and the derivative of this function is:
[0064]
[0065] The first-order derivative in x∈[0,1] is less than 0, and the increasing speed becomes slower and slower until the complementarity is strong. The adaptive mechanism of parameters such as inertia weight and learning factor can make the particle swarm genetic algorithm more robust and adaptable, and can automatically adjust the algorithm's parameter values to adapt to different search environments and objective functions.
[0066] After the position is updated, boundary processing is performed to ensure that the independent variable is within the value range. At the same time, the fitness value of each individual is calculated.
[0067] 3. When the fitness value of an individual is better than the optimal fitness value, the genetic algorithm population is updated until the termination condition is met.
[0068] Update the fitness function value to update the individual extreme value and the group extreme value, update the optimal solution of each individual and the global optimal solution according to the new fitness value, and record the global optimal fitness value: record the global optimal fitness value of each generation.
[0069] When it is found that the fitness value of an individual is better than the global optimal solution, selection, crossover and mutation operations are performed.
[0070] Selection operation: Randomly select N individuals, perform selection operations according to the index and save them as population 1.
[0071] Crossover operation: Perform weighted averaging on population 1 and the global optimal position to obtain a new population 2. The specific calculation formula used is:
[0072]
[0073] Mutation operation: randomly select M individuals and perform mutation operation on them (randomly generate new individuals within the value range) to obtain a new population 3. The specific calculation formula used is:
[0074]
[0075] The population after crossover and the population after mutation are integrated into a new population as population 4, and population 4 is used as the population for iterative update of the particle swarm.
[0076] In order to improve the computational efficiency of the algorithm, the present invention divides the entire population into several subgroups, each of which performs calculations on different processors and updates its particle positions and velocities in parallel. This approach requires ensuring information exchange and coordination between the subgroups, which is usually achieved by message passing or shared memory. Each subgroup performs iterative updates at the same time and merges after each iteration.
[0077] In each generation of the population, the effect of the current search is judged by monitoring the average fitness value of the population or the fitness value of the best individual. According to the change of the fitness value, some strategies can be used to dynamically adjust the crossover probability and mutation probability. For example, when the fitness value is low or the evolution speed is slow, the crossover probability and mutation probability can be increased to promote exploration; when the fitness value is high or the evolution speed is fast, the crossover probability adjustment and mutation probability can be reduced to strengthen local search.
[0078] The purpose of the mutation probability is to generate new individuals to maintain the diversity of the population and prevent premature maturation. m The probability of replacing individuals is 0.1, so as to obtain offspring individuals.
[0079]
[0080] where w i ,w m They are random numbers used to perform crossover and mutation operations, uniformly distributed in the interval [0,1]. The random number in the crossover operation is used to determine the location of the crossover point and whether to crossover a specific individual when performing the crossover operation; the random number in the mutation operation is used to determine whether to mutate a specific gene position of an individual when performing the mutation operation. Random numbers can generally be generated using a pseudo-random number generator, which can generate a series of seemingly random numbers based on a specific seed value. In the implementation of genetic algorithms, the quality and characteristics of the random number generator have a great impact on the performance of the algorithm. In order to ensure the fairness and repeatability of the algorithm, it is usually necessary to set the seed of the random number generator so that the results of the algorithm can be reproduced when needed.
[0081] The population size has a certain influence on the final result of the genetic algorithm. The initial population size of the present invention is 300. When the number of iterations of the algorithm reaches 300, the algorithm terminates. The crossover probability determines the probability of the parent individual performing a crossover operation, and the value is usually a minimum of 0.6 and a maximum of 0.9.
[0082] The termination condition is to meet any of the following conditions: reaching the maximum number of iterations, the fitness is no longer improved, the quality of the current solution reaches the preset accuracy, and the computing resources are exhausted. The maximum number of iterations means that the algorithm stops after a certain number of iterations, which is pre-set according to the nature of the problem and the accuracy requirements of the solution. The fitness is no longer improved means that if the global best fitness value or the average fitness value of the population has not increased significantly after multiple consecutive iterations, it can be considered that the algorithm has converged to a stable state, and the iteration can be terminated at this time. Meeting specific accuracy requirements means that if the quality of the current solution has reached or exceeded the preset accuracy threshold, it is considered that a good enough solution has been found and the iteration can be terminated. Computational resource exhaustion means that under limited computing resources (such as time or memory), when the resources are exhausted, the algorithm needs to terminate. When it is found that the fitness value of an individual is better than the global optimal solution, the choice to enter the genetic algorithm for operation is mainly based on the following reasons:
[0083] Retention of dominant individuals: The core idea of genetic algorithms is to simulate the process of natural selection and retain individuals with higher fitness in the hope that these individuals can pass on excellent characteristics to the next generation.
[0084] Maintaining diversity: Through selection operations, we can ensure that excellent genes will not be lost during the iteration process. At the same time, crossover and mutation operations are used to increase the diversity of the population and prevent the algorithm from falling into local optimality.
[0085] Global search capability: Genetic algorithms have good global search capabilities. When the particle swarm algorithm finds a local optimal solution, the genetic algorithm can help the algorithm jump out of the local optimal solution and explore a broader solution space.
[0086] Optimization of fitness function: Genetic algorithms can generate new solutions through crossover and mutation operations. These solutions may have better fitness values, thereby optimizing the fitness function.
[0087] Advantages of algorithm fusion: Particle swarm optimization (PSO) is good at fast convergence, but may fall into local optimality in some cases, while genetic algorithm (GA) performs better in global search. Combining the two can take advantage of the fast convergence characteristics of PSO and the global search ability of GA to increase the probability of finding the global optimal solution.
[0088] When the individual fitness value in the particle swarm algorithm is better than the global optimal solution, the genetic algorithm is used to update it, which can fully utilize the advantages of the genetic algorithm and improve the algorithm's search ability and solution quality.
[0089] In addition, the population size is also adjusted. At the beginning of the algorithm, the initial population size is set according to the scale and complexity of the problem, and the quality of the current population and the progress of the search are evaluated by monitoring the fitness of the population. The population size can be adjusted dynamically according to the changes in fitness. For example, when the fitness value is low or the evolution speed is slow, the population size can be increased to increase the breadth of the search; when the fitness value is high or the evolution speed is fast, the population size can be reduced to save computing resources.
[0090] Implementation of a remote monitoring and control system for a power grid
[0091] In another embodiment of the present invention, a power grid fault reconstruction system is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the above-mentioned power grid remote monitoring and control method when executing the computer program.
[0092] In order to further illustrate the effect of the present invention, a simulation experiment is now conducted on the fault reconstruction method of the present invention. This simulation experiment is aimed at the reconstruction process of a network fault in an IEEE 33-node distribution network.
[0093] In this simulation experiment, when the fault point is known, the present invention is compared with the particle swarm algorithm, the particle swarm algorithm with linear parameter adjustment, and the particle swarm algorithm with nonlinear parameter adjustment. The hybrid algorithm adopted by the present invention, the hybrid algorithm with linear parameter adjustment, and the hybrid algorithm with nonlinear parameter adjustment are also compared. The system targeted by this simulation experiment contains 33 nodes, 32 segment switches, 37 branches, and 5 tie switches, with a rated voltage of 12.66 kV. Assume that the fault point is at Figure 2 On the branch 15-16 shown, before the distribution network after the fault is reconstructed, the branches 9-15, 8-21, 25-29, and 12-22 are interconnecting branches in a disconnected state.
[0094] The maximum number of iterations for each algorithm is 100, and the design of other parameters is as follows: the population size (number of particles) M of each algorithm is 30 and 300 respectively, the speed limit factor is 1, the inertia weight value is 1, the two learning factors are 1.49445 and 1 respectively, the maximum value of the inertia weight factor in the process of linear and nonlinear parameter adjustment is 0.9, the minimum value is 0.4, the maximum value of the learning factor is 1.5, and the minimum value is 0.5. The initial contact switch is closed when the algorithm is running. The simulation results of various algorithms are shown in Table 1. Figure 3 and Figure 4,It can be seen from Table 1 that the tie switch sets of ,various algorithms all contain the sectional switches on the branch where the fault point is ,located.
[0095] Table 1
[0096]
[0097] Depend on Figure 3 It can be seen that when the population size is 30, according to Darwin's theory of natural selection, the degree of variation is also affected by the population size, resulting in a power loss of 154.9566kW under this condition. The particle swarm algorithm needs 59 iterations to find the optimal solution, which is too many iterations. The effect of nonlinear parameter adjustment is better than that of linear parameter adjustment, which effectively reduces the number of iterations. Similarly, the hybrid algorithm converges to the optimal solution after 6 iterations, proving that the hybrid algorithm can obtain the global optimal solution at a faster speed, which shows that it is a feasible and effective method.
[0098] Combination Figure 4 It can be seen that when the population size is 300, the particle swarm algorithm needs 34 iterations to reach the optimal value, and the power loss of the optimal switch combination found is the largest, which is 155.3071kW; the particle swarm algorithm with linearly changing the decreasing inertia weight and learning factor converges after 9 iterations, and the convergence speed is improved. The power loss corresponding to the optimal switch combination found is 154.8510kW. The particle swarm algorithm with nonlinearly changing the decreasing inertia weight and learning factor converges after 2 iterations, and the convergence speed is the fastest. The power loss corresponding to the optimal switch combination found is also 154.8510kW. However, because the distribution network reconstruction belongs to a discrete optimization problem, the network loss value found by the algorithm is not the minimum, and it is also possible that the population diversity is destroyed and falls into local interference. The convergence speed of the hybrid algorithm is faster than that of the particle swarm algorithm, and it has converged after about 4 iterations compared with the particle swarm algorithm with linearly changing the decreasing inertia weight, and the power loss is the smallest, which is 154.4841kW. The particle swarm algorithm with linearly changing decreasing inertia weight and learning factor converged after 4 iterations, and the convergence speed was improved. The particle swarm algorithm with nonlinearly changing decreasing inertia weight and learning factor converged after 2 iterations, and the convergence speed was the fastest. The power loss corresponding to the optimal switch combination found was also 154.8510kW. It can be seen that the hybrid algorithm adopted by the present invention combines the advantages of particle swarm algorithm and genetic algorithm, avoids the shortcomings of both, and adds parallel steps and adaptive parameter adjustment, which can effectively reduce the loss in the distribution network and improve the power supply quality, proving the feasibility, effectiveness and high efficiency of the algorithm.
Claims
1. A remote monitoring and control method for a power grid, characterized in that: The method comprises the following steps: 1) Taking the minimum power loss in the distribution network as the objective function; 2) generating an initial population according to the topological structure of the distribution network, iteratively updating the population using a particle swarm algorithm, and calculating the fitness value of each individual and the global optimal fitness value of each generation during the iterative updating process; the fitness function used by the fitness value is the objective function; 3) When the fitness value of an individual is better than the global optimal fitness value, the genetic algorithm is used to update the population to generate a new generation of population, and the particle swarm algorithm is used to iteratively update the generated new generation of population until the termination condition is met; the population with the best fitness value at the end of the iteration is taken as the optimal solution to control the switches in the distribution network.
2. The remote monitoring and control method for a power grid according to claim 1, characterized in that: In the step 2), when the particle swarm algorithm is used to iteratively update the population, the inertia weight and the learning factor are updated in a linear decreasing or nonlinear decreasing manner to achieve the update of the individual speed and position.
3. The power grid remote monitoring and control method according to claim 2, characterized in that: The update formula of inertia weight and learning factor in nonlinear decreasing mode is: ω is the inertia weight factor, c1, c2 are learning factors, t is the current number of iterations, t max is the total number of iterations, ω max and ω min are the maximum and minimum values of the inertia weight factor, c max and c min are the maximum and minimum values of the inertia weight factor, respectively.
4. The remote monitoring and control method for a power grid according to any one of claims 1 to 3, characterized in that: When the particle swarm algorithm is used to iteratively update the population, the population is divided into several sub-groups. Each sub-group performs calculations on different processors, updates its particle positions and velocities in parallel, and information is exchanged between the sub-groups.
5. The remote monitoring and control method for a power grid according to claim 1, characterized in that: The objective function is: Where P loss is the power loss of the distribution network, N b is the total number of branches in the distribution network, R i For branch b i The branch resistance; P i For flow through branch b i The active power, f i b is the current flowing through the branch i Reactive power; U i For branch b i The end node voltage.
6. The power grid remote monitoring and control method according to claim 1, characterized in that: When performing crossover and / or mutation operations, the crossover probability and / or mutation probability are adjusted according to the fitness value or the evolution speed. When the fitness value is high or the evolution speed is fast, the crossover probability and / or mutation probability are reduced to strengthen local search; when the fitness value is low or the evolution speed is slow, the crossover probability and / or mutation probability are increased.
7. The power grid remote monitoring and control method according to claim 1, characterized in that: The method also includes adjusting the population size according to the fitness value or the evolution speed. When the fitness value is high or the evolution speed is fast, the population size is reduced; when the fitness value is low or the evolution speed is slow, the population size is increased.
8. The power grid remote monitoring and control method according to claim 1, characterized in that: The termination condition is any one of the following conditions: reaching the maximum number of iterations, the fitness no longer improving, the quality of the current solution reaching the preset accuracy, or the computing resources being exhausted.
9. A power grid remote monitoring and control system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the remote monitoring and control method for a power grid as described in any one of claims 1 to 8 is implemented.
Citation Information
Patent Citations
Improved particle swarm hybrid algorithm for fault location of distributed power distribution network
CN115906914A
Cited By
Road construction equipment safety monitoring system based on Internet of Things
CN119783013A
