CN112765882B - CVT equivalent parameter identification method of AFSA and L-M fusion algorithm - Google Patents

Abstract

The present application provides a CVT equivalent parameter identification method of the AFSA and L‑M fusion algorithm, including: algorithm initialization, assigning initial values to artificial fish; calculating the fitness value of each artificial fish and recording the global optimal artificial fish state; averaging the state of the artificial fish and simulating the execution of the corresponding behavior; updating the global optimal state of the artificial fish group and assigning the optimal value to the bulletin board; if the termination condition is met, assigning the initial value of the L‑M algorithm and setting the step size; iterating and accurately solving until the convergence condition is met; identifying the optimal solution for the parameters of the CVT electromagnetic unit. Based on the research on vector impedance load, the present application provides an important foundation and guarantee for subsequent CVT model simulation, error analysis, fault diagnosis and other work through intelligent algorithm identification, which can effectively improve the parameter identification accuracy of the CVT electromagnetic unit and reduce the equipment modification and operation and maintenance costs.

Description

CVT equivalent parameter identification method based on AFSA and L-M fusion algorithm

Technical Field

The present application relates to the technical field of research on equivalent model parameter identification of capacitive voltage transformer, and in particular to a CVT equivalent parameter identification method of AFSA and L-M fusion algorithm.

Background technique

Voltage transformers convert primary voltage into low voltage at a certain ratio and are mainly used to measure voltage and power in high and ultra-high voltage systems. At present, CVT (capacitor voltage transformer) accounts for up to 90% of the substations of 110kV and above due to its good insulation performance, light weight, small size and low cost. Its measurement results are an important basis for secondary metering, relay protection and monitoring equipment.

However, CVT has a complex structure and a changeable operating environment, and its failure rate is relatively high, about 5 times that of traditional PT. CVT insulation defects mainly include concentrated defects and distributed defects. Compared with distributed defects, concentrated defects include capacitor unit breakdown, compensation reactor failure, and short circuit between turns of intermediate transformer windings, which are more fatal to the safe operation of CVT. Accurately identifying the internal equivalent parameters of the CVT electromagnetic unit is an important prerequisite for evaluating and diagnosing CVT concentrated defects.

Domestic and foreign identification methods are based on known CVT circuit model parameters (using pure resistive load) and use traditional nonlinear optimization algorithms including gradient descent, Gauss-Newton algorithm (G-N algorithm for short), Levenberg-Marquarat algorithm (L-M algorithm for short) and least squares method. In practical applications, due to the limitations of traditional optimization algorithms in terms of convergence speed and initial value sensitivity, their application scope is greatly restricted.

Summary of the invention

The present application provides a CVT equivalent parameter identification method of AFSA and L-M fusion algorithm to solve the problem that traditional identification methods cannot guarantee global search capability and inaccurate solutions.

The CVT equivalent parameter identification method of the AFSA and L-M fusion algorithm provided in this application includes:

Initialize the algorithm and assign initial values to the artificial fish;

Calculate the fitness value of each artificial fish and record the global optimal artificial fish state;

Average the state of the artificial fish and simulate the execution of corresponding behaviors;

Update the global optimal state of the artificial fish school and assign the optimal value to the bulletin board;

If the termination condition is met, the initial value of the L-M algorithm is assigned and the step size is set;

Iterate and accurately solve the problem until the convergence condition is met;

The optimal solution of the parameters of the CVT electromagnetic unit is obtained by identification.

Optionally, the algorithm is initialized and the step of assigning initial values to the artificial fish includes: initializing the population size N, the initial position of each artificial fish, the field of vision Visual of the artificial fish, the step size step, the crowding factor δ, and the number of repetitions Trynumber (variables L1, R1, Lm, Rm, L2, R2), initial randomization settings, and setting of the objective function.

Optionally, the steps of averaging the state of the artificial fish and simulating the execution of corresponding behaviors include: evaluating each individual and selecting the behaviors to be executed, including foraging, flocking, chasing and randomness.

Optionally, the step of updating the global optimal artificial fish school state and assigning the optimal value to the bulletin board includes:

Randomly execute the behavior of artificial fish and update and generate new fish schools;

Evaluate all individuals, and if an individual is better than the billboard, update the billboard to that individual.

Optionally, if the termination condition is met, the steps of assigning an initial value to the L-M algorithm and setting a step size include:

When the optimal solution on the bulletin board reaches the convergence condition, the algorithm ends;

The optimal solution (L1, R1, Lm, Rm, L2, R2) obtained by the AFSA algorithm is used as the initial value and step size of the L-M algorithm.

Optionally, the step of identifying and obtaining the optimal solution of the parameters of the CVT electromagnetic unit further includes:

The identification results were evaluated by the goodness of fit R;

The goodness of fit R is defined as:

Among them, yi is the measured data, To fit the data, is the average value of the measured data. The goodness of fit R ranges from (0 to 1), and the closer it is to 1, the better the fitting effect.

This application proposes a fusion optimization algorithm of artificial fish school algorithm and L-M algorithm on a capacitive voltage transformer based on vector impedance load, which ensures the global search capability of the algorithm and avoids the problem of inaccurate solution. This application provides a CVT equivalent parameter identification method of AFSA and L-M fusion algorithm, including: algorithm initialization, artificial fish assignment of initial value; calculation of fitness value of each artificial fish, recording of global optimal artificial fish state; average artificial fish state, simulation execution of corresponding behavior; update global optimal artificial fish state, assign the optimal value to the bulletin board; if the termination condition is met, assign L-M algorithm initial value and set step size; iterate and accurately solve until the convergence condition is met; identify and obtain the optimal solution of the parameters of the CVT electromagnetic unit. Based on the research on vector impedance load, this application provides an important foundation and guarantee for subsequent CVT model simulation, error analysis, fault diagnosis and other work through intelligent algorithm identification, which can effectively improve the parameter identification accuracy of the CVT electromagnetic unit and reduce the equipment transformation and operation and maintenance costs.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solution of the present application, the drawings required for use in the embodiments are briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

FIG1 is a schematic diagram of a CVT load error characteristic test circuit;

Figure 2 is a schematic flow chart of a CVT equivalent parameter identification method of an AFSA and L-M fusion algorithm provided in this application.

Detailed ways

The following embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following embodiments do not represent all implementations consistent with the present application. They are only examples of systems and methods consistent with some aspects of the present application as detailed in the claims.

In order to accurately identify the internal parameters of CTV, the current identification methods at home and abroad are based on the known CVT circuit model parameters (using pure resistive load), using traditional nonlinear optimization algorithms including gradient descent, Gauss-Newton algorithm (G-N algorithm for short), Levenberg-Marquarat algorithm (L-M algorithm for short) and least squares method. In practical applications, due to the limitations of traditional optimization algorithms in terms of convergence speed and initial value sensitivity, their application scope is greatly limited. Intelligent algorithms are engineering technologies for solving mathematical optimization problems that have been explored and derived by scholars at home and abroad in recent years, mainly including: Genetic Algorithm (GA), Simulated Annealing (SA), Ant Colony Algorithm (ACO), Particle Swarm Optimization (PSO), Artificial Neural Network (ANN), etc.

AFSA is an efficient intelligent algorithm that simulates the cluster foraging behavior of fish in nature. It adopts a bottom-up optimization model and enables the group to achieve the goal of optimal selection through collaboration with individuals in the group. In the artificial fish swarm algorithm, foraging behavior lays the foundation for algorithm convergence; clustering behavior enhances the stability of algorithm convergence; tail-chasing behavior enhances the rapidity and globality of algorithm convergence; and its evaluation behavior also provides a guarantee for the speed and stability of algorithm convergence. AFSA has the advantages of strong global search capabilities and low requirements on parameters, initial values, and objective functions, but it will have defects such as reduced convergence speed in the later stage, and it is not easy to accurately converge to the optimal solution.

The L-M algorithm (Levenberg-Marquardt, L-M for short) is the most widely used nonlinear least squares method. Its basic idea is to use the gradient to find the maximum (minimum) value, and to perform analysis and evaluation by adding a positive definite matrix to the Hessian matrix. It is a kind of "hill climbing" method. The L-M algorithm has the advantages of fast convergence speed and strong local search ability, but its accuracy is more dependent on the initial value of the calculation. If the initial value is not properly selected, the algorithm may not be able to search for the optimal solution. Therefore, this application uses the Artificial Fish Swarm Algorithm (AFSA for short) and the L-M fusion algorithm to perform parameter identification of the CVT electromagnetic unit on a capacitive voltage transformer based on a vector impedance load.

In order to better solve the optimization problem, this application proposes an optimization method that combines the AFSA algorithm with the L-M algorithm. When running the algorithm, the AFSA algorithm can be used to perform a global search of the feasible domain to find an approximate solution to the optimal solution. Then, the L-M algorithm is used to search for the optimal solution near the approximate solution. The use of this algorithm can not only ensure the global search capability of the algorithm, but also avoid inaccurate solutions.

Since the transfer function of the CVT output voltage is fixed, when the secondary load parameters or electromagnetic unit parameters change, the ratio difference and angle difference will change accordingly. The ratio difference and angle difference can be regarded as multivariate functions of the CVT electromagnetic unit parameters, and the values of the electromagnetic unit parameters can be solved by the optimization algorithm.

The present application provides a CVT equivalent parameter identification method of an AFSA and L-M fusion algorithm, comprising:

The algorithm is initialized and the artificial fish is assigned initial values.

Initialize the population size N, the initial position of each artificial fish, the visual field Visual of the artificial fish, the step size step, the crowding factor δ, and the number of repetitions Trynumber (initial randomization settings of variables L1, R1, Lm, Rm, L2, R2, and setting of the objective function).

Calculate the fitness value of each artificial fish and record the global optimal artificial fish state.

Average the states of the artificial fish and simulate execution of corresponding behaviors.

Each individual is evaluated and given a choice of behaviors to perform, including foraging, flocking, chasing, and random.

Update the global optimal state of the artificial fish school and assign the optimal value to the bulletin board.

Randomly execute the behavior of artificial fish and update and generate new fish schools;

Evaluate all individuals, and if an individual is better than the billboard, update the billboard to that individual.

If the termination condition is met, the initial value of the L-M algorithm is assigned and the step size is set.

When the optimal solution on the bulletin board reaches the convergence condition, the algorithm ends;

The optimal solution (L1, R1, Lm, Rm, L2, R2) obtained by the AFSA algorithm is used as the initial value and step size of the L-M algorithm.

Iterate and accurately solve the problem until the convergence condition is met.

The optimal solution of the parameters of the CVT electromagnetic unit is obtained by identification.

The step of identifying and obtaining the optimal solution of the parameters of the CVT electromagnetic unit also includes:

The identification results were evaluated by the goodness of fit R;

The goodness of fit R is defined as:

Among them, yi is the measured data, To fit the data, is the average value of the measured data. The goodness of fit R ranges from (0 to 1), and the closer it is to 1, the better the fitting effect.

The following is an embodiment provided by the present application.

Figure 1 is a schematic diagram of a CVT load error characteristic test circuit. Figure 2 is a flow chart of a CVT equivalent parameter identification method using an AFSA and L-M fusion algorithm provided by the present application.

First get the parameters:

U _s is accurately obtained through a standard voltage transformer (PT).

The equivalent capacitance C _H , C _L and dielectric loss tangent tanδ of the high and low voltage arms of CVT can be measured and obtained by dielectric loss meter. _RH and _RL can be converted into power frequency equivalent resistance by the formula R＝l/(ωC·tanδ). _Kn is the known rated voltage ratio.

Finally, six unknown parameters need to be determined: _Rm is the equivalent resistance of the intermediate transformer excitation branch, _Lm is the equivalent inductance of the intermediate transformer excitation branch, _L1 is the sum of the equivalent inductance of the compensation reactor and the leakage inductance of the primary side of the transformer, _R1 is the sum of the equivalent resistance of the compensation reactor and the equivalent resistance of the primary side of the transformer, _L2 is the leakage inductance of the secondary side of the transformer converted to the primary side, and _R2 is the equivalent resistance of the secondary side of the transformer converted to the primary side. CVT obtains the output voltage of the primary side of the intermediate transformer through capacitor voltage division (primary voltage reduction) and then through the intermediate transformer (secondary voltage reduction).

Among them, the output voltage of the secondary side load voltage converted to the primary side of the intermediate transformer is,

_Z0 is the input impedance of the electromagnetic unit, and its expression is,

_Z1 is the impedance of the high-voltage arm capacitor, and its calculation formula is shown in formula (8), where _tanδH is the dielectric loss tangent of the high-voltage arm capacitor,

Z ₁ =1/( _{ωCH tanδH} + _jωCH ) (3)

_Z2 is the parallel impedance of the low-voltage arm capacitor and the electromagnetic unit, and its calculation formula is shown in formula (6), where _tanδL is the dielectric loss tangent of the low-voltage arm capacitor,

Z ₂ ＝Z ₀ /(1+jωCLZ ₀ + _{ωCLtan δLZ 0 )} (4)

The ratio error refers to the error between the CVT secondary output voltage and the primary voltage amplitude after conversion. The specific expression is as follows:

Where: U _s is the amplitude of the primary voltage; U ₂ is the amplitude of the CVT secondary output voltage.

The angular difference refers to the phase difference between the CVT secondary output voltage and the primary voltage:

δ＝δ ₂ -δ _s (6)

Where: δ _s is the phase angle of the primary voltage; δ ₂ is the phase angle of the CVT secondary output voltage.

The existing standard stipulates that when δ ₂ is greater than δ _s , the secondary output voltage is defined as leading the primary voltage. At this time, the angle difference is positive, otherwise it is negative. Usually the angle difference is expressed in minutes or centiradians.

Through the CVT load characteristic error test, multiple sets of ratio difference ε and angle difference δ are obtained. From the following equations (7) to (8), it can be known that the ratio difference ε and angle difference δ are functions of L ₁ , R ₁ , L _m , R _m , L ₂ , and R ₂ , as follows:

ε＝f(L ₁ ，R ₁ ，L _m ，R _m ，L ₂ ，R ₂ ) (7)

δ = g(L ₁ , R ₁ , L _m , R _m , L ₂ , R ₂ ) (8)

The ratio difference ε and angle difference δ obtained by the experiment are identified to obtain L′ ₁ , R′ ₁ , L′ _m , R′ _m , L′ ₂ , and R′ ₂ , and then ε′ and δ′ are calculated according to the following equations (9) to (10).

ε′＝f(L′ ₁ ，R′ ₁ ，L′ _m ，R′ _m ，L′ ₂ ，R′ ₂ ) (9)

δ′＝g(L′ ₁ ，R′ ₁ ，L′ _m ，R′ _m ，L′ ₂ ，R′ ₂ ) (10)

The objective function can be set as the following equations (11) and (12). The purpose of parameter identification is to minimize the objective function, thereby determining the "optimal solution" of the electromagnetic unit parameters.

After obtaining the objective function, the equivalent parameter identification method of the CVT electromagnetic unit of the AFSA and L-M fusion algorithm is run. When running the algorithm, the feasible domain can be globally searched by the AFSA algorithm to find the approximate solution of the optimal solution. Then, the L-M algorithm is used to search for the optimal solution near the approximate solution. The specific algorithm steps include: algorithm initialization, assigning initial values to the artificial fish; calculating the fitness value of each artificial fish and recording the global optimal artificial fish state; averaging the state of the artificial fish and simulating the execution of the corresponding behavior; updating the global optimal state of the artificial fish group and assigning the optimal value to the bulletin board; if the termination condition is met, assigning the initial value of the L-M algorithm and setting the step size; iterating and accurately solving until the convergence condition is met; identifying the optimal solution of the parameters of the CVT electromagnetic unit. The use of this algorithm can not only ensure the global search capability of the algorithm, but also avoid inaccurate solutions.

It can be seen from the above technical solutions that the present application proposes a fusion optimization algorithm of artificial fish school algorithm and L-M algorithm on the capacitive voltage transformer based on vector impedance load, which ensures the global search capability of the algorithm and avoids the problem of inaccurate solution. The present application provides a CVT equivalent parameter identification method of AFSA and L-M fusion algorithm, including: algorithm initialization, artificial fish assignment of initial value; calculation of fitness value of each artificial fish, recording of global optimal artificial fish state; average artificial fish state, simulation execution of corresponding behavior; update global optimal artificial fish state, assign the optimal value to the bulletin board; if the termination condition is met, assign L-M algorithm initial value and set step size; iterate and accurately solve until the convergence condition is met; identify and obtain the optimal solution of the parameters of the CVT electromagnetic unit. Based on the research on vector impedance load, the present application provides an important foundation and guarantee for subsequent CVT model simulation, error analysis, fault diagnosis and other work through intelligent algorithm identification, which can effectively improve the parameter identification accuracy of the CVT electromagnetic unit and reduce the equipment transformation and operation and maintenance costs.

Similar parts between the embodiments provided in this application can be referenced to each other. The specific implementation methods provided above are only a few examples under the general concept of this application and do not constitute a limitation on the protection scope of this application. For those skilled in the art, any other implementation methods expanded based on the scheme of this application without creative work belong to the protection scope of this application.

Claims (1)

1. A CVT equivalent parameter identification method of an AFSA and L-M fusion algorithm is characterized by comprising the following steps:

   Initializing an algorithm, and giving an initial value to the artificial fish; the algorithm is initialized, and the initial value given by the artificial fish comprises the following components: initializing the population scale N, the initial position of each artificial fish, the Visual field of the artificial fish, the step length step, the crowding factor delta and the repetition number Trynumber, performing initial random setting on preset variables, and setting an objective function; the preset variables comprise L ₁、R₁、L_m、R_m、L₂、R₂,R_m which is the equivalent resistance of an intermediate transformer excitation branch, L _m which is the equivalent inductance of the intermediate transformer excitation branch, L ₁ which is the sum of the equivalent inductance of the compensating reactor and the primary side leakage inductance of the transformer, R ₁ which is the sum of the equivalent resistance of the compensating reactor and the primary side equivalent resistance of the transformer, L ₂ which is the secondary side leakage inductance of the transformer converted to the primary side, and R ₂ which is the secondary side equivalent resistance of the transformer converted to the primary side;

   calculating the fitness value of each artificial fish, and recording the global optimal artificial fish state;

   evaluating each individual, selecting the behavior to be performed by the individual, including foraging, clustering, rear-end collision and randomization;

   Updating the global optimal artificial fish swarm state and assigning an optimal value to the bulletin board; the updating the global optimal artificial fish school status, assigning an optimal value to the bulletin board comprises: randomly executing the behavior of the artificial fish, and updating to generate a new fish swarm; evaluating all individuals, and if a certain body is better than the bulletin board, updating the bulletin board to the individual;

   if the termination condition is met, giving an initial value of the L-M algorithm and setting a step length; if the termination condition is met, the steps of giving the initial value of the L-M algorithm and setting the step length comprise the following steps: when the optimal solution on the bulletin board reaches the convergence condition, ending the algorithm; taking the optimal solution of the preset variable obtained by the AFSA algorithm as an initial value of the L-M algorithm and setting a step length;

   Iterative accurate solution is carried out until convergence conditions are met;

   identifying to obtain the optimal solution of the parameters of the CVT electromagnetic unit; the identifying to obtain the optimal solution of the parameters of the CVT electromagnetic unit comprises the following steps:

   Evaluating the identification result through the fitting goodness R;

   The goodness of fit R is defined as:

   Wherein yi is the actual measurement data, To fit data,/>For the average value of the measured data, the value range of the goodness of fit R is (0, 1), wherein the closer the goodness of fit R is to 1, the better the fitting effect is.