Line backup constant value tuning method based on improved particle swarm optimization
By improving the particle swarm algorithm, dynamically adjusting the number of fixed value groups and inertia weight factors, the problem of poor compatibility of range fixed value and time fixed value tuning in the power grid is solved, and fast and accurate global fixed value optimization of the power grid is achieved, avoiding the power grid The accident occurred.
Patent Information
- Application Number
- CN202410983642.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-22
- Publication Date
- 2024-11-12
AI Technical Summary
When the existing particle swarm algorithms handle the tuning of range and time values in the power grid, they are not compatible with continuous and discrete tuning accuracy, resulting in reduced tuning accuracy and may cause grid accidents.
Using the improved particle swarm algorithm, the number of fixed value groups and inertia weight factors are dynamically adjusted, combined with the rapidity, sensitivity and selectivity of relay protection, multiple iterations are carried out to find the global optimal fixed value group, which is compatible with continuous Search and iterate between type and discrete fixed values.
It realizes the rapid and accurate finding of the global optimal value of the power grid, avoids errors in the matching process after the range and time value are respectively tuned, and improves the overall tuning efficiency.
Smart Images

Figure BDA0004956775450000021 
Figure BDA0004956775450000031 
Figure BDA0004956775450000032
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of power grid relay protection, and relates to a line backup fixed value tuning method, in particular to a method for finding a fixed value optimal solution after improving a particle swarm algorithm. Background Art
[0002] In order to obtain the global optimal value of the power grid, it is necessary to use certain algorithms to find the optimal solution or a solution close to the optimal solution through search space, evaluation function and optimization strategy. In order to solve the optimization problem, various heuristic algorithms can be used. The basic idea of the heuristic algorithm is to guide the search process through a series of rules, strategies or evaluation functions to find the optimal solution or a solution close to the optimal solution. It searches in an approximate and iterative way, and does not guarantee the global optimal solution, but it can usually find a better solution within a reasonable time. Common heuristic algorithms include genetic algorithms, simulated annealing algorithms, particle swarm optimization algorithms, ant colony optimization algorithms, etc. Each algorithm has its own characteristics and scope of application. You can choose a suitable heuristic algorithm to solve the problem according to the nature of the problem.
[0003] The particle swarm algorithm is an intelligent optimization algorithm that simulates the collective cooperation of bird flocks to forage to achieve the best result. The particle swarm algorithm is between the genetic algorithm and the evolutionary programming. Its idea is intuitive, its implementation is simple, and it has a high execution efficiency. It can effectively optimize various functions and is suitable for both scientific research and engineering applications. The basic idea of the particle swarm algorithm is to regard the problem to be optimized as a search problem in a multidimensional space. A group of particles are randomly generated in the solution space. Each particle corresponds to a solution and is iteratively updated according to its current position and speed to find a better solution. The speed and position update of the particles are affected by the individual historical experience and the group's shared information. In this scheme, the particles in the particle swarm algorithm are equivalent to a set of fixed values, the number of particles is equivalent to the number of fixed value groups, and the movement of particles is equivalent to the generation of another set of fixed values after the fixed value group is iterated once.
[0004] As we all know, constants include range constants and time constants, where range constants can be continuous, while time constants are generally not continuous and are discrete, such as 0.15 and 0.5, which are discontinuous. However, in the control of inertia weight factors, the existing particle swarm algorithm generally uses a continuous method to assign values, which is not compatible with the iteration and search of time constants, and the accuracy of time constant tuning will be greatly reduced. The prior art uses a discrete particle swarm algorithm to tune the time constant. However, the problem that arises is that in the process of tuning the range constant and the time constant in two ways and then matching them, errors are caused in the later matching, which in turn causes power grid accidents, and the overall tuning efficiency is relatively low. Summary of the invention
[0005] In order to solve the technical problem of being able to simultaneously optimize the range setting and the time setting, the present invention designs a line backup setting optimization method based on an improved particle swarm algorithm. By improving the particle swarm algorithm, it is compatible with simultaneous search and iteration of continuous setting and discrete setting, thereby enabling rapid setting optimization.
[0006] The technical solution adopted by the present invention is a line backup setting optimization method based on an improved particle swarm algorithm. On the basis of the preset values of the switchgear, the speed, sensitivity and selectivity of the relay protection are taken as the principles, and the setting with the best relay protection performance is found through multiple iterations. The optimization method uses the improved particle swarm algorithm and iteratively calculates all the setting values, and then compares them with the relay protection performance objective function to finally find the optimal setting group.
[0007] The basic formula of particle swarm algorithm is:
[0008]
[0009] In formula 1, V i =(V i1 ,V i2 ,···,V iN ), represents a set of fixed values, k represents the current number of iterations, x i =(x i1 ,x i2 ,···,x iN ) represents the N-dimensional position vector in the i-th set of constants, i = (1, 2, ..., m) represents a set of constants, P i =(P i1 ,P i2 ,···,P iN ) represents the performance corresponding to the currently searched optimal fixed value, P g =(P g1 ,P g2 ,···,P gN ) represents the performance corresponding to the optimal fixed value searched so far, c1 and c2 are learning factors, r1 and r2 are random numbers between 0 and 1, w represents the inertia weight factor, V i k Represents the value of the constant group before the current iteration, V i k+1 Represents the value of the constant group after the current iteration,
[0010] Among them, the improved particle swarm algorithm includes:
[0011] a. Control the number m of the constant value groups currently being searched and adjust them dynamically during the iteration process. The formula is:
[0012]
[0013] In formula 2, m min is the minimum population size set; m max is the maximum population size set; n is the number of protected species required to calculate the constant value; n max The total number of protections for calculating the fixed value;
[0014] b. Control the inertia weight factor w to perform global and local searches and make dynamic adjustments. The formula is:
[0015]
[0016] In formula 3, w max represents the maximum inertia weight factor, w min represents the minimum inertia weight factor, k max is the maximum number of iterations, k is the current number of iterations,
[0017] c. The optimal set R obtained by the improved particle swarm algorithm of step a and step b is input into the relay protection setting calculation and tuning system to obtain the performance parameters of average action time, number of mismatch points, average sensitivity, and number of set values to be adjusted. The objective function of optimizing relay protection performance is:
[0018]
[0019] In formula 4, T mean is the average action time, Num sp is the number of mismatch points, K lm-mean is the average sensitivity, Num tz To adjust the number,
[0020] When comparing the performance of relay protection, after finding all the optimal constant value group sets R through the improved particle swarm algorithm, the first target f1 is optimized to obtain the constant value group set R1 with the shortest average action time, and then the second target f2 is optimized to obtain the constant value group set R2 with the largest number of mismatch points, and then the third target f3 is optimized to obtain the constant value group set R3 with the lowest average sensitivity, and then the fourth target f4 is optimized to obtain the constant value group x with the least number of adjustments, and finally the optimized constant value group is obtained.
[0021] In step b, the maximum inertia weight factor w max Obtained through the following methods,
[0022]
[0023] In formula 5, is the lower limit of the maximum inertia weight factor, is the upper limit of the maximum inertia weight factor; is the lower limit of the minimum inertia weight factor, is the upper limit of the minimum inertia weight factor.
[0024] In step b, the minimum inertia weight factor w min Obtained through the following methods,
[0025]
[0026] In formula 6, is the lower limit of the maximum inertia weight factor, is the upper limit of the maximum inertia weight factor; is the lower limit of the minimum inertia weight factor, is the upper limit of the minimum inertia weight factor.
[0027] The set value includes a range set value and a time set value, and the value of N is 2.
[0028] If there are multiple fixed value groups after optimization, just randomly select one group.
[0029] The beneficial effect of the present invention is that a multi-dimensional and multi-objective model for constant value optimization of the global power grid is constructed in combination with the constant value data characteristics, the continuity of the range constant value and the discreteness of the time constant value are combined, and by setting appropriate inertia weight factors, the number of constant value groups and the number of iterations, an improved particle swarm algorithm is proposed, which can find the optimal solution for the constant value coordination of the global power grid at one time and more quickly, avoiding the tedious process of respectively optimizing the range constant value and the time constant value and then matching them, as well as the easy occurrence of errors in the matching process. DETAILED DESCRIPTION
[0030] In this scheme, the principles for the optimal solution are defined, namely, speed, sensitivity, and selectivity, while also considering the practicality of the tuning results. The performance objective function is defined as
[0031]
[0032] In formula 4, T mean is the average action time, Num sp is the number of mismatch points, K lm-mean is the average sensitivity, Num tzis the number that needs to be adjusted. At the same time, the above four principles are used in turn to find the best layer by layer. Among the initial many sets of fixed value solutions R, first find the set R1 with the least number of mismatch points, then find the set R2 with the largest average sensitivity within the range of set R1, then find the set R3 with the shortest average action time within the range of set R2, and finally find the result with the least number of adjustments within the range of set R3. If there is only one final result, then select the set of fixed values corresponding to this performance, which is the fixed value that should be selected after tuning. If there are 2 or more final results, then they are all optimal in terms of performance, so you can randomly select one of the sets of fixed values.
[0033] The above is the comparison process of the performance objective function in this scheme. Let's see how the set R of many sets of fixed value schemes is obtained before the performance comparison. The basic formula of the traditional particle swarm algorithm is as follows:
[0034]
[0035] In formula 1, V i =(V i1 ,V i2 ,···,V iN ), represents a set of fixed values, which is equivalent to a particle in the particle swarm algorithm, k represents the current iteration number, which is equivalent to the number of movements in the particle swarm algorithm, and x i =(x i1 ,x i2 ,···,x iN ) represents the N-dimensional position vector in the i-th set of constants, which is equivalent to the velocity vector of the particle in the particle swarm algorithm. i = (1, 2, ..., m) represents a set of constants, which is equivalent to the number of particles in the particle swarm algorithm. i =(P i1 ,P i2 ,···,P iN ) represents the performance corresponding to the currently searched optimal fixed value, P g =(P g1 ,P g2 ,···,P gN ) represents the performance corresponding to the optimal fixed value searched so far, c1 and c2 are learning factors, r1 and r2 are random numbers between 0 and 1, w represents the inertia weight factor, V i k Represents the value of the constant group before the current iteration, V i k+1Represents the value of the constant group after the current iteration. In the above formula, the inertia weight factor w and the number of particles m are critical parameter indicators, because these values determine how the particles change. In the field of relay protection, the inertia weight factor is an important parameter to balance the global optimal ability and local optimal ability of the particle swarm optimization algorithm. When the inertia weight factor is small, such as w<0.8, the particle swarm optimization algorithm tends to be locally optimal, its search time is short, and all particles will quickly gather together; when the inertia weight factor is large, such as w>1.2, the particle swarm optimization algorithm tends to be globally optimal, and each iteration always explores new areas, and the required calculation time is longer, but it is more likely that the global optimal solution will not be found.
[0036] Since the power grid model is relatively complex, if this method is used for calculation in each iteration in actual operation, it will take up a lot of computer memory, take a long time to calculate, and affect the convergence speed of the algorithm. Therefore, this method is needed to improve the traditional particle swarm algorithm.
[0037] Taking the typical backup protection distance II section of 220kV power grid as an example, there is a protection position on each side of each line, and each protection distance II section needs to set two fixed values, range and time. The setting value of each protection position needs to meet the sensitivity of the line end fault and the setting principle of coordination with the adjacent I section, longitudinal connection or II section.
[0038] The variation range and constraint conditions of particles should meet the requirements of actual relay protection setting procedures according to the characteristics of the fixed value items. Taking the optimization of the phase-to-phase distance protection fixed value of the whole network as an example, the global fixed value automatic tuning technology using the improved particle swarm algorithm is explained. The phase-to-phase distance protection fixed value can be divided into 5 fixed value items (phase-to-phase distance I fixed value, phase-to-phase distance II fixed value, phase-to-phase distance II time fixed value, phase-to-phase distance III fixed value, phase-to-phase distance III time fixed value), and each set of fixed values is regarded as a particle.
[0039] First, the particles are initialized. Each particle represents a set of potential solutions. In the relay protection constant value tuning, each particle represents a set of constant values, including two dimensions: constant value and time. In order to obtain effective solutions faster and expand the search space of particles, when initializing the position and speed of the particle population, the positions of some particles take theoretical extreme values, the positions of some particles are randomly generated, and the positions of some particles are selected from initial values. This not only ensures that a set of solutions that are not greater than the initial state fitness value of the system can be obtained in the first iteration, but also expands the activity range of particles and increases the diversity of particles, thereby effectively preventing the population from falling into local optimality. One of the particles is the current operating constant value of the constant value. When generating other particles, the requirements of the relay protection setting calculation procedure must be met.
[0040] Secondly, encode each particle for identification and convenient recording and comparison. In this scheme, since the protection setting value and time setting value are continuous and discrete respectively, they are combined and encoded for unified dynamic tuning. This step is only for identity recognition for the scheme itself and has nothing to do with the overall tuning process.
[0041] Again, regarding the setting of the number of particle populations, the entire particle swarm algorithm is performed on the group of solutions, and has the parallelism and globality of the search process. The larger the population size, the higher the diversity of individuals in the group, and the smaller the risk of the algorithm falling into a local optimal solution, but the amount of calculation will also increase significantly; on the contrary, if the population size is too small, the search space will be limited, the group lacks diversity, and premature phenomenon is prone to occur. The search space size of the optimization problem is different, so it is necessary to dynamically adjust the population size according to different partition situations. When the present invention dynamically adjusts the number of particles, it is controlled by the following formula:
[0042]
[0043] In formula 2, m min is the minimum population size set; m max is the maximum population size set; n is the number of protected species required to calculate the constant value; n max To calculate the total number of protected constants; the population number determined by this method changes linearly with the number of control variables in the partition. When the number of control variables in the partition is small, the number of particles in the group also decreases. If the number of control variables in the partition is large, the number of particles in the group also increases.
[0044] Next is the setting of the inertia weight factor w.
[0045] When the number of particles is the same, when the target search space is small, the particle flight process is more like a local optimization. A small inertia weight factor can make the particle swarm algorithm adjust in the local area and shorten the search time. However, when the target search space is large, the particle search process is more like a global optimization. A large inertia weight factor can make particles explore new areas during the iteration process and expand the search range. The setting of the inertia weight factor is improved in combination with the characteristics of the dynamic change of the number of control variables. The maximum inertia weight factor w required for each optimization is set to max and the minimum inertia weight factor w min It dynamically performs self-correction as the number of control variables in the partition changes, namely:
[0046]
[0047] In formula 3, w max represents the maximum inertia weight factor, wmin represents the minimum inertia weight factor, k max is the maximum number of iterations, and k is the current number of iterations; this scheme is based on power grid partitioning, which can reduce the number of control variables. Although the number of initial partition control variables is small, they are all control variables closely related to the objective function, so it is like looking for a local optimum. This function makes the inertia factor of the improved particle swarm algorithm change not only with the number of iterations in the optimization process, but also with the number of control variables, thereby improving the performance of the particle swarm algorithm.
[0048] In formula 3, the maximum inertia weight factor w max and the minimum inertia weight factor w min Obtained through the following methods,
[0049]
[0050]
[0051] In formula 5 and formula 6, is the lower limit of the maximum inertia weight factor, is the upper limit of the maximum inertia weight factor; is the lower limit of the minimum inertia weight factor, is the upper limit of the minimum inertia weight factor.
[0052] The above is the method and process of optimizing the line backup constant value of this solution. Of course, in the specific implementation process, when encountering some special situations, special treatment is required, such as:
[0053] 1) The set value of distance section III shall not be less than the set value of distance section II. If the requirement is not met, the set value of section III shall be the set value of section II.
[0054] 2) If the initial constants of the double-circuit line are the same, the constants after iteration will also remain consistent.
[0055] 3) The time constant is set according to a fixed value. If it is the load line distance section II, there are only two cases: 0.15 and 0.5. If it is the connecting line distance section II, the minimum time is 0.5.
[0056] In summary, this method combines the characteristics of constant value data to construct a multi-dimensional and multi-objective model for constant value optimization of the global power grid. The improved particle swarm algorithm proposed in the constant value optimization process can find the optimal solution for the constant value coordination of the global power grid more quickly by setting appropriate inertia weights, number of particles and number of iterations.
Claims
1. A line backup setting optimization method based on an improved particle swarm algorithm, based on the preset value of the switchgear, based on the principle of the speed, sensitivity and selectivity of the relay protection, finds the optimal setting of the relay protection performance through multiple iterations, characterized in that: The tuning method uses the improved particle swarm algorithm and iterative calculation to obtain all the fixed values, and then compares them with the relay protection performance objective function to finally find the optimal fixed value group. The basic formula of particle swarm algorithm is: In formula 1, V i =(V i1 ,V i2 ,···,V iN ), represents a set of fixed values, k represents the current number of iterations, x i =(x i1 ,x i2 ,···,x iN ) represents the N-dimensional position vector in the i-th set of constants, i = (1, 2, ..., m) represents a set of constants, P i =(P i1 ,P i2 ,···,P iN ) represents the performance corresponding to the currently searched optimal fixed value, P g =(P g1 ,P g2 ,···,P gN ) represents the performance corresponding to the optimal fixed value searched so far, c1 and c2 are learning factors, r1 and r2 are random numbers between 0 and 1, and w represents the inertia weight factor. Represents the value of the constant group before the current iteration, Represents the value of the constant group after the current iteration, Among them, the improved particle swarm algorithm includes: a. Control the number m of the constant value groups currently being searched and adjust them dynamically during the iteration process. The formula is: In formula 2, m min is the minimum population size set; m max is the maximum population size set; n is the number of protected species required to calculate the constant value; n max The total number of protections for calculating the fixed value; b. Control the inertia weight factor w to perform global and local searches and make dynamic adjustments. The formula is: In formula 3, w max represents the maximum inertia weight factor, w min represents the minimum inertia weight factor, k max is the maximum number of iterations, k is the current number of iterations, c. The optimal set R obtained by the improved particle swarm algorithm of step a and step b is input into the relay protection setting calculation and tuning system to obtain the performance parameters of average action time, number of mismatch points, average sensitivity, and number of set values to be adjusted. The objective function of optimizing relay protection performance is: In formula 4, T mean is the average action time, Num sp is the number of mismatch points, K lm-mean is the average sensitivity, Num tz To adjust the number, When comparing the performance of relay protection, after finding all the optimal constant value group sets R through the improved particle swarm algorithm, the first target f1 is optimized to obtain the constant value group set R1 with the shortest average action time, and then the second target f2 is optimized to obtain the constant value group set R2 with the largest number of mismatch points, and then the third target f3 is optimized to obtain the constant value group set R3 with the lowest average sensitivity, and then the fourth target f4 is optimized to obtain the constant value group x with the least number of adjustments, and finally the optimized constant value group is obtained.
2. The line backup constant value tuning method based on improved particle swarm algorithm according to claim 1 is characterized in that: In step b, the maximum inertia weight factor w max Obtained through the following methods, In formula 5, is the lower limit of the maximum inertia weight factor, is the upper limit of the maximum inertia weight factor; is the lower limit of the minimum inertia weight factor, is the upper limit of the minimum inertia weight factor.
3. The line backup constant value tuning method based on improved particle swarm algorithm according to claim 1 is characterized in that: In step b, the minimum inertia weight factor w min Obtained through the following methods, In formula 6, is the lower limit of the maximum inertia weight factor, is the upper limit of the maximum inertia weight factor; is the lower limit of the minimum inertia weight factor, is the upper limit of the minimum inertia weight factor.
4. The line backup constant value tuning method based on improved particle swarm algorithm according to claim 1 is characterized in that: The set value includes a range set value and a time set value, and the value of N is 2.
5. The line backup constant value tuning method based on improved particle swarm algorithm according to claim 1 is characterized in that: If there are multiple fixed value groups after optimization, just randomly select one group.
Citation Information
Cited By
Power plant relay protection setting value optimization method and system based on optimization algorithm
CN119813079A
A power plant relay protection setting optimization method and system based on optimization algorithm
CN119813079B
