An improved hybrid genetic algorithm-based whole-line grounding state evaluation method

Abstract

The application provides a full-line grounding state evaluation method based on an improved hybrid genetic algorithm, belongs to the grounding resistance monitoring problem of a power system, and comprises the following steps: measuring each grounding resistance; establishing an objective function of the improved hybrid genetic algorithm; constructing a suitable BP neural network; training the BP neural network; and judging whether the error reaches the specified requirement or not.The application improves the grounding resistance measurement mode, realizes real-time monitoring of the grounding resistance, performs trend analysis and early warning, reduces the workload of manual measurement, and improves the work efficiency.

Description

Technical Field

[0001] This invention relates to the problem of monitoring grounding resistance in power systems, and in particular to a method for evaluating the grounding status of the entire power line based on an improved hybrid genetic algorithm. Background Technology

[0002] Currently, China's power transmission lines are widely distributed, long-distance, and crisscrossed. Many are located in mountainous areas with complex terrain and variable climates. During operation, these transmission lines are easily susceptible to direct lightning strikes, which can cause tripping accidents and serious damage to the power system. Therefore, transmission lines need a good discharge channel.

[0003] As a key component of the current discharge channel, the grounding device of transmission towers plays a crucial role in the stable operation of transmission lines. During a lightning strike, the lightning current flows into the earth through the grounding device. If the grounding resistance is too high, the lightning arrester will generate a backflash overvoltage, causing the line to trip. The measurement of the grounding resistance of transmission line towers is affected by numerous factors, including the shape of the grounding electrode, its inductive component, the tower's own inductance, the lightning arrester's inductance, and the surrounding environment. Therefore, how to quickly and accurately measure the grounding resistance of transmission line towers is one of the most pressing problems to be solved in the power industry.

[0004] Currently, the measurement of tower grounding resistance mainly employs the megohmmeter method and the clamp meter method. The megohmmeter method is accurate, but cumbersome to operate, requiring the setup of voltage and current electrodes and the disconnection of the grounding down conductor during measurement, significantly increasing the workload of workers and resulting in low efficiency. The clamp meter method is simple and convenient, requiring only the clamp jaws to hold the grounding down conductor for measurement. When the tower has only one grounding wire, it doesn't need to be disconnected; however, when there are multiple grounding down conductors, the others must be disconnected, leaving only the conductor being measured, which adds inconvenience and increases measurement error. Both methods share the drawback of requiring periodic on-site manual measurement, necessitating the disconnection of the grounding down conductor during measurement, consuming significant manpower and resources, and easily causing bolt loosening or poor contact, leading to increased tower grounding resistance. Therefore, there is an urgent need to improve grounding resistance measurement methods to achieve real-time monitoring of grounding resistance, trend analysis and early warning, reduce manual measurement workload, and improve work efficiency. Summary of the Invention

[0005] This invention provides a method for evaluating the grounding status of the entire line based on an improved hybrid genetic algorithm, which reduces the workload of manual measurement and improves work efficiency.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0007] A method for assessing the grounding status of an entire power line based on an improved hybrid genetic algorithm is proposed. This method employs an improved hybrid genetic algorithm and a BP neural network for analysis and prediction to achieve real-time monitoring of grounding resistance. The method includes the following steps:

[0008] Step 1: Measure the resistance of each grounding element;

[0009] Step 2: Establish the objective function for the improved hybrid genetic algorithm;

[0010] Step 3: Perform crossover and mutation operations on the remaining individuals;

[0011] Step 4: Calculate the population center It is expressed as follows:

[0012]

[0013] Step 5: As the initial search point for the conjugate gradient method, set the iteration count N0, and execute to obtain the result.

[0014] Step 6: Put Add it to the population and replace the individual with the lowest fitness in the population. Proceed to step 2. Terminate the algorithm when the set conditions are met to obtain a local optimum.

[0015] Step 7: Use the local optimal solution obtained in Step 6 as the initial value for the sequential quadratic programming algorithm to construct the Lagrange function;

[0016] Step 8: Obtain the Lagrange factor θ and determine the step size;

[0017] Step 9: Obtain the new iteration point χ θ+1 Update the Hessian matrix;

[0018] Step 10: Determine χ θ+1 Check if the iteration termination condition is met. If it is, output the final result; otherwise, let θ = θ + 1 and repeat steps 7 to 9.

[0019] Step 11: Construct a BP neural network with 3 layers: an input layer, a hidden layer, and an output layer.

[0020] Step 12: Initialize the parameters;

[0021] Step 13: Train the BP neural network;

[0022] Step 14: Determine whether the error meets the specified requirements. If it does, output the data. If it does not meet the specified requirements, return to step 13.

[0023] Further: Step 1 specifically includes:

[0024] Step 1.1: Define the apparent resistance at the first grounding electrode in the circuit as r. s1Its calculation expression is as follows:

[0025]

[0026] In the formula: u is the measured voltage; i1 is the measured current of the first grounding electrode; r1 is the grounding resistance value of the first grounding electrode;

[0027] Step 1.2: Measure the resistance r of each grounding electrode circuit. χ The apparent grounding resistance r of each grounding electrode is obtained. sχ , means as follows:

[0028]

[0029] In the formula: r χ =u χ / i n ,χ=1,2,…,n; This represents the total parallel resistance of the remaining branches;

[0030] Step 1.3: Percentage error in grounding resistance measurement It is expressed as follows:

[0031]

[0032] Further: Step 2 specifically includes;

[0033] Step 2.1: Set the population size M and crossover probability P for the traditional genetic algorithm. mc Probability of mutation P mu Number of iterations N;

[0034] Step 2.2: Generate the initial population;

[0035] Step 2.3: Regarding the difference ε between the apparent grounding resistance value and the actual measured value n , means as follows:

[0036] ε n =r m -r sn =r m -r χ -1 / (1 / r² + 1 / r³ + ... + 1 / r) χ-1 +1 / r χ+1 +…+1 / r n )

[0037] In the formula: r m The actual measured value for each grounding electrode;

[0038] Step 2.4: Based on Step 2.3, calculate the difference ε between the apparent grounding resistance value and the actual measured value.n The smaller the value, the lower the resistance value r. m The closer it is to the true value, the better. Based on step 2.1, the objective function F of the grounding resistance is obtained, which is the fitness of the individual, as follows:

[0039]

[0040] Further: Step 13 specifically includes:

[0041] Step 13.1: Calculate the hidden layer. The output vector Ξ of the hidden layer is equal to the input vector Γ = {Γ1, Γ2, ..., Γ}. E Connection weights between input layer and hidden layer and hidden layer threshold The relationship is represented as follows:

[0042]

[0043] In the formula: Ξ is the hidden layer output; g is the activation function of the hidden layer;

[0044] Step 13.2: BP neural network output Θ and hidden layer output Ξ, hidden layer-output layer connection weights. and output layer threshold The relationship is represented as follows:

[0045]

[0046] Step 13.3: Calculate the error and predict the error. The relationship between the expected output Λ and the predicted output Θ is expressed as follows:

[0047]

[0048] Step 13.4: Update the weights and use the BP prediction error. Update network connection weights and It is expressed as follows:

[0049]

[0050]

[0051] In the formula: For learning rate;

[0052] Step 13.5: Update the threshold and use the BP prediction error. Update node threshold and It is expressed as follows:

[0053]

[0054]

[0055] A method for evaluating the grounding status of the entire line based on an improved hybrid genetic algorithm includes the following steps:

[0056] Step 1: Measure the resistance of each grounding element;

[0057] Step 1.1: Define the apparent resistance at the first grounding electrode in the circuit as r. s1 Its calculation expression is as follows:

[0058]

[0059] In the formula: u is the measured voltage; i1 is the measured current of the first grounding electrode; r1 is the grounding resistance value of the first grounding electrode;

[0060] Step 1.2: Measure the resistance r of each grounding electrode circuit. χ The apparent grounding resistance r of each grounding electrode can be obtained. sχ , means as follows:

[0061]

[0062] In the formula: r χ =u χ / i n ,χ=1,2,…,n; This represents the total parallel resistance of the remaining branches;

[0063] Step 1.3: Percentage error in grounding resistance measurement It is expressed as follows:

[0064]

[0065] Step 2: Establish the objective function F of the improved hybrid genetic algorithm;

[0066] Step 2.1: Set the population size M and crossover probability P for the traditional genetic algorithm. mc Probability of mutation P mu Number of iterations N;

[0067] Step 2.2: Generate the initial population;

[0068] Step 2.3: Regarding the difference ε between the apparent grounding resistance value and the actual measured value n , means as follows:

[0069]

[0070] In the formula: r mThe actual measured value for each grounding electrode;

[0071] Step 2.4: As shown in Step 2.3, the difference ε between the apparent grounding resistance value and the actual measured value is... n The smaller the value, the lower the resistance value r. m The closer it is to the true value, the better. Based on step 2.1, the objective function F of the grounding resistance, i.e., the fitness of the individual, can be obtained, as follows:

[0072]

[0073] Step 3: Perform crossover and mutation operations on the remaining individuals;

[0074] Step 4: Calculate the population center It is expressed as follows:

[0075]

[0076] Step 5: As the initial search point for the conjugate gradient method, set the iteration count N0, and execute to obtain the result.

[0077] Step 6: Put Add it to the population and replace the individual with the lowest fitness in the population. Proceed to step 2. Terminate the algorithm when the set conditions are met to obtain a local optimum.

[0078] Step 7: Use the local optimal solution obtained in Step 6 as the initial value for the sequential quadratic programming algorithm to construct the Lagrange function;

[0079] Step 8: Obtain the Lagrange factor θ and determine the step size;

[0080] Step 9: Obtain the new iteration point χ θ+1 Update the Hessian matrix;

[0081] Step 10: Determine χ θ+1 Check if the iteration termination condition is met. If it is, output the final result; otherwise, let θ = θ + 1 and repeat steps 7 to 9.

[0082] Step 11: Construct a suitable BP neural network with 3 layers: an input layer, a hidden layer, and an output layer.

[0083] Step 12: Initialize the parameters;

[0084] Step 13: Train the BP neural network;

[0085] Step 13.1: Calculate the hidden layer. The output vector Ξ of the hidden layer is equal to the input vector Γ = {Γ1, Γ2, ..., Γ}.E Connection weights between input layer and hidden layer and hidden layer threshold The relationship is represented as follows:

[0086]

[0087] In the formula: Ξ is the hidden layer output; g is the activation function of the hidden layer;

[0088] Step 13.2: BP neural network output Θ and hidden layer output Ξ, hidden layer-output layer connection weights. and output layer threshold The relationship is represented as follows:

[0089]

[0090] Step 13.3: Calculate the error and predict the error. The relationship between the expected output Λ and the predicted output Θ is expressed as follows:

[0091]

[0092] Step 13.4: Update the weights and use the BP prediction error. Update network connection weights and It is expressed as follows:

[0093]

[0094]

[0095] In the formula: For learning rate;

[0096] Step 13.5: Update the threshold and use the BP prediction error. Update node threshold and It is expressed as follows:

[0097]

[0098]

[0099] Step 14: Determine whether the error meets the specified requirements. If it does, output the data. If it does not meet the specified requirements, return to step 13.

[0100] Furthermore, the improved hybrid genetic algorithm described in steps 2 to 10 is based on the traditional genetic algorithm, and combines the sequential quadratic programming algorithm and the conjugate gradient method to improve the grounding resistance measurement method.

[0101] Furthermore, the BP neural network described in steps 11 to 14 can effectively analyze and predict the trend of grounding resistance value, thereby realizing real-time monitoring and trend early warning.

[0102] Furthermore, step 1 improves the grounding resistance measurement method to achieve real-time monitoring of grounding resistance and perform trend analysis and early warning, thereby reducing the workload of manual measurement and improving work efficiency.

[0103] The advantages and beneficial effects of this invention are:

[0104] This invention can be widely applied to real-time monitoring and trend assessment of grounding resistance of transmission line towers, replacing traditional manual on-site measurements. It improves the grounding resistance measurement method by combining a hybrid genetic algorithm based on traditional genetic algorithms with sequential quadratic programming and the conjugate gradient method. Through a BP neural network, it can effectively analyze and predict the trend of grounding resistance values, thereby achieving real-time monitoring and trend early warning. This invention improves the grounding resistance measurement method, enabling real-time monitoring of grounding resistance and trend analysis and early warning, reducing the workload of manual measurements and improving work efficiency.

[0105] By using multiple sensors to synchronously measure the grounding resistance of multiple monitoring points on the same line, a multi-information fusion mathematical model is established. An improved hybrid genetic algorithm is used for comprehensive calculation, analysis and management, which enables rapid and accurate measurement of the grounding resistance of transmission line towers. Furthermore, a BP neural network algorithm is introduced to predict the measured data, evaluate the grounding status of each tower in real time, and generate trend analysis curves of grounding resistance at each point, providing data support for timely detection of potential faults. Attached Figure Description

[0106] Figure 1 This is a flowchart of the full-line grounding status assessment method based on the improved hybrid genetic algorithm in this invention.

[0107] Figure 2 This is the equivalent model of each grounding resistor in this invention.

[0108] Figure 3 This is the equivalent model of multiple grounding resistors and multiple loops in this invention.

[0109] Figure 4 This is a diagram of the topology of the three-layer BP neural network in this invention.

[0110] Figure 5 This is a comparison chart of the error results between the basic genetic algorithm and the improved hybrid genetic algorithm in this invention. Detailed Implementation

[0111] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0112] This invention presents a method for assessing the grounding status of an entire power line based on an improved hybrid genetic algorithm. It combines a basic genetic algorithm with the conjugate gradient method and a sequential quadratic programming algorithm, and introduces a backpropagation (BP) neural network algorithm to predict the measured data. The specific process is as follows: Figure 1 As shown, it includes the following steps:

[0113] Step 1: Measure each grounding resistance, and its equivalent model is as follows. Figure 2 As shown;

[0114] Step 1.1: Define the apparent resistance at the first grounding electrode in the circuit as r. s1 Its calculation expression is:

[0115]

[0116] Step 1.2: Measure the resistance r of each grounding electrode circuit. χ The apparent grounding resistance r of each grounding electrode can be obtained. sχ :

[0117]

[0118] Step 1.3: Percentage error in grounding resistance measurement When the number of grounding electrodes n approaches infinity, theoretically r can be approximated. sχ =r χ The error percentage approaches 0 infinitely. In practical systems, the number of grounding resistors is finite, and there are many loops. The specific equivalent model is as follows: Figure 3 As shown:

[0119]

[0120] Step 2: Based on the traditional genetic algorithm, establish the objective function of the improved hybrid genetic algorithm;

[0121] Step 2.1: Set the population size M and crossover probability P for the traditional genetic algorithm. mc Probability of mutation P mu Number of iterations N;

[0122] Step 2.2: Generate the initial population, i.e., the measured resistance of each grounding electrode;

[0123] Step 2.3: Regarding the difference ε between the apparent grounding resistance value and the actual measured value n Its expression is:

[0124]

[0125] Step 2.4: As shown in Step 2.3, the difference ε between the apparent grounding resistance value and the actual measured value is... n The smaller the value, the lower the resistance value r. m The closer it is to the true value, the better. Based on step 2.1, the objective function F of the grounding resistance, i.e., the fitness of the individual, can be obtained, and its expression is:

[0126]

[0127] Step 3: Perform crossover and mutation operations on the remaining individuals;

[0128] Step 4: Calculate the population center

[0129]

[0130] Step 5: As the initial search point for the conjugate gradient method, set the iteration count N0, and execute to obtain the result.

[0131] Step 6: Put Add it to the population and replace the individual with the lowest fitness in the population. Proceed to step 2. Terminate the algorithm when the set conditions are met to obtain a local optimum.

[0132] Step 7: Use the local optimal solution obtained in Step 6 as the initial value for the sequential quadratic programming algorithm to construct the Lagrange function;

[0133] Step 8: Obtain the Lagrange factor θ and determine the step size;

[0134] Step 9: Obtain the new iteration point χ θ+1 Update the Hessian matrix;

[0135] Step 10: Determine χ θ+1 Check if the iteration termination condition is met. If it is, output the final result; otherwise, let θ = θ + 1 and repeat steps 7 to 9.

[0136] Step 11: Construct a suitable BP neural network. This invention sets the number of layers to 3: an input layer, a hidden layer, and an output layer. The specific three-layer BP neural network topology diagram is shown below. Figure 4 As shown;

[0137] Step 12: Initialize the measured parameters;

[0138] Step 12.1: The number of nodes E in the input layer and the number of nodes F in the output layer are set to correspond to the sequence dimensions of the input and output (Γ,Λ);

[0139] Step 12.2: The number of nodes L in the hidden layer is related to E and F;

[0140] Step 12.3: Connection weights between input layer, hidden layer, and output layer and Hidden layer threshold Output layer threshold All of them need to be initialized;

[0141] Step 12.4: Then determine the learning rate and neuron activation function;

[0142] Step 13: Train the BP neural network;

[0143] Step 13.1: Calculate the hidden layer. The output vector Ξ of the hidden layer is equal to the input vector Γ = {Γ1, Γ2, ..., Γ}. E Connection weights between input layer and hidden layer and hidden layer threshold The relationship is:

[0144]

[0145] Step 13.2: BP neural network output Θ and hidden layer output Ξ, hidden layer-output layer connection weights. and output layer threshold The relationship is:

[0146]

[0147] Step 13.3: Calculate the error and predict the error. The relationship between the expected output Λ and the predicted output Θ is as follows:

[0148]

[0149] Step 13.4: Update the weights and use the BP prediction error. Update network connection weights and

[0150]

[0151]

[0152] Step 13.5: Update the threshold and use the BP prediction error. Update node threshold and

[0153]

[0154]

[0155] Step 14: Determine whether the error meets the specified requirements. If it does, output the data. If it does not meet the specified requirements, return to step 13.

[0156] This invention utilizes an improved hybrid genetic algorithm for comprehensive calculation, analysis, and management, enabling rapid and accurate measurement of the grounding resistance of transmission line towers. From... Figure 5 As can be seen, this invention collected 10 grounding resistances for verification, and the error percentage using the improved hybrid genetic algorithm was significantly smaller than the error percentage using the basic genetic algorithm.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.

Claims

1. A method for evaluating the grounding status of the entire line based on an improved hybrid genetic algorithm, characterized in that: An improved hybrid genetic algorithm is used to analyze and predict grounding resistance through a BP neural network, enabling real-time monitoring of grounding resistance. The steps include: Step 1: Measure the resistance of each grounding element; Step 2: Establish the objective function for the improved hybrid genetic algorithm, specifically including: Step 2.1: Set the population size M and crossover probability for the traditional genetic algorithm. ,Mutations probability Number of iterations N; Step 2.2: Generate the initial population; Step 2.3: Regarding the difference between the apparent grounding resistance value and the actual measured value , means as follows: ， In the formula: The actual measured value for each grounding electrode; Step 2.4: Based on Step 2.3, calculate the difference between the apparent grounding resistance value and the actual measured value. The smaller the value, the lower the resistance value. The closer the value is to the true value, the closer the objective function F of the grounding resistance is to the individual's fitness, which is expressed as follows: ； Step 3: Perform crossover and mutation operations on the remaining individuals; Step 4: Calculate the population center , means as follows: ， Step 5: As the initial search point for the conjugate gradient method, the number of iterations is set. The execution yields the result. ; Step 6: [The sentence is incomplete and requires more context to be translated accurately.] Add it to the population and replace the individual with the lowest fitness in the population. Proceed to step 2. Terminate the algorithm when the set conditions are met to obtain a local optimum. Step 7: Use the local optimal solution obtained in Step 6 as the initial value for the sequential quadratic programming algorithm to construct the Lagrange function; Step 8: Obtain the Lagrange factor And determine the step size; Step 9: Derive the new iteration point Update the Hessian matrix; Step 10: Determine If the iteration termination condition is met, output the final result. If so, otherwise Repeat steps 7 through 9; Step 11: Construct a BP neural network with 3 layers: an input layer, a hidden layer, and an output layer. Step 12: Initialize the parameters; Step 13: Train the BP neural network, specifically including: Step 13.1: Calculate the hidden layer and output the hidden layer. With input vector Connection weights between input layer and hidden layer and hidden layer threshold The relationship is represented as follows: ， In the formula: Output for hidden layer; Here is the activation function for the hidden layer; Step 13.2: Output of the BP neural network With hidden layer output Hidden layer-output layer connection weights and output layer threshold The relationship is represented as follows: ， Step 13.3: Calculate the error and predict the error. With expected output and predicted output The relationship is represented as follows: ， Step 13.4: Update the weights and use the BP prediction error. Update network connection weights and , means as follows: ， ， In the formula: For learning rate; Step 13.5: Update the threshold and use the BP prediction error. Update the threshold of the node and , means as follows: ， ； Step 14: Determine whether the error meets the specified requirements. If it does, output the data. If it does not meet the specified requirements, return to step 13.

2. The method for evaluating the grounding status of the entire line based on an improved hybrid genetic algorithm according to claim 1, characterized in that: Step 1 specifically includes: Step 1.1: Define the apparent resistance at the first grounding electrode in the circuit as... Its calculation expression is as follows: ， In the formula: u is the measured voltage; The measured current is for the first grounding electrode; This is the grounding resistance value of the first grounding electrode; Step 1.2: Measure the resistance of each grounding electrode circuit separately. The apparent grounding resistance of each grounding electrode is obtained. , means as follows: ， In the formula: ; This represents the total parallel resistance of the remaining branches; Step 1.3: Percentage error in grounding resistance measurement , means as follows: 。

Images