Primary frequency modulation control parameter optimization and evaluation method for hydroelectric generating set

Abstract

The invention provides a hydroelectric generating set primary frequency modulation control parameter optimization and evaluation method, and the method comprises the following steps: 1, obtaining test data of a hydroelectric generating set, and carrying out the preprocessing of the data; 2, analyzing primary frequency modulation characteristics of the hydroelectric generating set, and constructing a hydroelectric generating set regulation system model; 3, an MDE-RBF algorithm is constructed, and the primary frequency modulation PID control parameters are optimized through the MDE-RBF algorithm; 4, evaluating the performance of the primary frequency modulation result through the ratio of the actual integral electric quantity to the theoretical integral electric quantity; according to the method, the optimal PID parameter combination is searched by using the MDE-RBF algorithm, the response speed and stability of primary frequency modulation of the hydroelectric generating set are effectively improved, effective control is realized under various operation conditions, the primary frequency modulation performance of the hydroelectric generating set is remarkably improved, the frequency stability of a power grid is enhanced, and powerful support is provided for safe and stable operation of a power system.

Description

Technical Field

[0001] The present invention relates to the technical field of hydropower generation regulation, and in particular to a method for optimizing and evaluating primary frequency regulation control parameters of a hydropower unit. Background Art

[0002] With the continuous expansion and increasing complexity of power systems, frequency stability and response speed have become crucial factors in grid operation. Hydropower units play a key role in primary frequency regulation due to their rapid response capabilities and regulatory flexibility. Hydropower units need to provide frequent frequency regulation auxiliary services to the grid, and the grid determines the amount of rewards and penalties for auxiliary services by evaluating the primary frequency regulation performance of the units. Primary frequency regulation specifically refers to maintaining grid frequency stability by rapidly adjusting the active power output of hydropower units when grid frequency fluctuates. Although traditional PID controllers perform well in many applications, their parameter optimization problems remain challenging in complex and changing grid environments. Summary of the Invention

[0003] In view of this, the object of the present invention is to provide a method for optimizing and evaluating the primary frequency regulation control parameters of a hydropower unit, so as to at least solve the above problems.

[0004] The technical solution adopted in the present invention is as follows: A method for optimizing and evaluating primary frequency regulation control parameters of a hydropower unit, comprising the following steps: Step 1: Obtain the test data of the hydropower unit and preprocess the data; Step 2: Analyze the primary frequency regulation characteristics of the hydropower unit and build a hydropower unit regulation system model; Step 3: Construct the MDE-RBF algorithm and optimize the primary frequency modulation PID control parameters through the MDE-RBF algorithm; Step 4: Evaluate the performance of the primary frequency modulation result by the ratio of the actual integrated power to the theoretical integrated power.

[0005] Furthermore, the data preprocessing in step 1 is specifically to remove the steady-state value of the data. The specific formula is:

[0006] in, is the processed signal sequence, is the original signal sequence, is the total number of data points in the time period before the frequency step perturbation, is the first points.

[0007] Furthermore, the open-loop transfer function of the hydropower unit regulation system in step 2 is:

[0008] in 、 、 They are the proportional gain, integral gain, and differential gain of the PID controller of the turbine control system of the hydraulic turbine. is the Laplace operator, is the inertia time constant, is the adjustment coefficient, is the frequency deviation amplification factor, is the time constant of the frequency measurement link.

[0009] Furthermore, step 3 is specifically as follows: using the mutation differential evolution algorithm to perform optimal search to generate the primary frequency modulation control parameters; introducing an exponential smoothing factor into the RBF neural network, combining the mutation differential evolution algorithm and the RBF neural network with fast and fine weights and deviations to construct the MDE-RBF algorithm, and optimizing the primary frequency modulation control parameters through the MDE-RBF algorithm.

[0010] Furthermore, the mutation differential evolution algorithm specifically includes the following steps: Step 51: Randomly generate the initial population:

[0011] in , , is the scale factor, For the The first generation Individuals, is the perturbation vector, To compare Generation and the first The optimal value determined by the generation group; The random index and The generation method is to extract twice, generating two sets of random integer indexes and , take the minimum value of each group as the final index:

[0012] Step 52: Generate a random factor and extract different strategies for search based on the value of the random factor:

[0013] in, is a random factor and , represents a random vector on the interval, whose length is equal to the number of variables, and are the upper and lower bounds of the objective function; Step 53: Generation Group Individuals and their variant intermediates Cross-processing between:

[0014] in, is the crossover probability, which is A random vector of for A random integer; Step 54: Use the greedy algorithm on the individuals obtained by crossover. If the individual after crossover is the optimal solution, the hybridized individual is added to the next generation. Otherwise, the original individual is added to the next generation and continues processing:

[0015] Furthermore, an exponential smoothing factor is introduced into the RBF neural network to quickly refine the weights and deviations. The specific steps include: Step 61: Introduce the conjugate gradient method to modify parameters in the RBF neural network 、 and , and set the maximum number of iterations and error threshold; is the node center of the password layer neuron, is the reference width, is the connection weight between the hidden layer and the output layer; Step 62: Introduce the smoothing factor and generate a new gradient search calculation formula:

[0016] in, For parameters No. The search direction vector of the iteration, is the neuron index, is the index of the iteration number, is the objective function Parameters The gradient, is the smoothing factor, when Time smoothing factor Determined by the Polak-Ribière formula:

[0017] in, is the objective function Parameters In the The gradient of the iteration, is the objective function Parameters In the The gradient of the iteration, express The transpose of express The transpose of Step 63: Calculate the adaptive step size:

[0018] in, is the correction multiplier and , Step 64: Iteratively update parameters:

[0019] in, 、 、 The three parameters are the learning rate of the node center, the expansion constant, the connection weight and , is the number of iterations.

[0020] Furthermore, optimizing the primary frequency modulation control parameters by using the MDE-RBF algorithm specifically includes the following steps: Step 71: In MDE and RBF, set the population , number of iterations , scale factor , crossover probability , the center point of the basis function , basis width vector , and the weights from the hidden layer to the output layer ; Step 72: Sample the reference input and output and calculate the system error ; Step 73: RBF neural network in time Normalize weights and increment PID related parameters; Step 74: Use the conjugate gradient method to calculate the center point of the basis function , base width parameter and the initial weights of RBF ; Step 75: Calculate the inverse of the mean square error as the fitness function; Step 76: Get the current time 、 、 And get the next update result; Step 77: Analyze the output values 、 、 Whether the requirements are met, if so, the optimization process ends; otherwise, the iterative parameter optimization process continues based on the MDE algorithm.

[0021] Furthermore, the calculation formula for the ratio of the actual integrated power to the theoretical integrated power in step 4 is:

[0022] in, is the primary frequency modulation integrated power ratio, which represents the ratio of actual integrated power to theoretical integrated power. The calculation formula is:

[0023] Among them, is the actual integrated power, is the theoretical integrated power, is the rated power of the unit, is the rated frequency, is the compensation ratio, is the instantaneous value of the unit’s active power, It is the initial steady-state reference value of the unit’s active power before a frequency regulation action. is the integration interval (s), is the integration time; is the effective frequency deviation, and the calculation formula is:

[0024] in, is the rated frequency, is the generator outlet frequency, This is the artificial frequency or speed dead zone.

[0025] Compared with the prior art, the present invention has the following beneficial effects: The present invention aims to rapidly adjust the active power output of a hydroelectric generator set to maintain grid frequency stability. This method collects and processes hydroelectric generator test data to construct a primary frequency regulation model for the hydroelectric generator set. The MDE-RBF algorithm is then used to effectively optimize the primary frequency regulation control parameters, improving overall model performance. The algorithm then evaluates the frequency regulation results, providing a method for optimizing and evaluating the primary frequency regulation control parameters of a hydroelectric generator set. This method improves the frequency regulation capability and operating efficiency of hydroelectric generator sets in complex grid environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only preferred embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0027] Figure 1 is a flow chart of a method provided by an embodiment of the present invention; Figure 2 Schematic diagram of a hydraulic turbine and a speed regulation closed-loop system thereof provided by an embodiment of the present invention; Figure 3 This is a flow chart of an improved RBF neural network provided by an embodiment of the present invention; Figure 4 is a flow chart of the MDE-RBF algorithm provided by an embodiment of the present invention; Figure 5 This is a comparison diagram of unit step response curves of the optimization algorithm provided by an embodiment of the present invention; Figure 6 This is a comparison diagram of the control effect of the input signal of the hydropower unit model provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The principles and features of the present invention are described below with reference to the accompanying drawings. The enumerated embodiments are only used to explain the present invention and are not used to limit the scope of the present invention.

[0029] Reference Figure 1 、 Figure 2 、 Figure 3 、 Figure 4 The present invention provides a method for optimizing and evaluating the primary frequency regulation control parameters of a hydropower unit, the method comprising: Step 1: Obtain the test data of the hydropower unit and preprocess the data; For example, the test data is mainly collected in real time through the data acquisition system of the unit speed governor, including the frequency, active power and guide vane opening of the hydro-generator unit. During the data processing process, the steady-state values ​​often contain a large amount of low-frequency noise or interference. In order to improve the signal-to-noise ratio of the signal, these steady-state values ​​need to be removed. Specifically, a frequency step disturbance is added to the frequency measurement input of the hydro-generator unit to collect the average value of the signal. By subtracting this steady-state value from the original signal, the noise level can be effectively reduced, thereby more accurately analyzing and evaluating the performance of the hydro-generator unit. The specific formula is as follows:

[0030] in, is the processed signal sequence, is the original signal sequence, is the total number of data points in the time period before the frequency step perturbation, is the first points.

[0031] Step 2: Analyze the primary frequency regulation characteristics of the hydropower unit and build a hydropower unit regulation system model; The hydropower unit is mainly responsible for the primary frequency regulation task, so the primary frequency regulation characteristics of the hydropower unit are mainly analyzed. The open-loop transfer function of the hydropower unit regulation system is:

[0032] in 、 、 They are the proportional gain, integral gain, and differential gain of the PID controller of the turbine control system of the hydraulic turbine. is the Laplace operator, is the inertia time constant, is the adjustment coefficient, is the frequency deviation amplification factor, is the time constant of the frequency measurement link.

[0033] Step 3: Construct the MDE-RBF algorithm and optimize the primary frequency modulation PID control parameters through the MDE-RBF algorithm; Step 3 is as follows: using the mutation differential evolution algorithm to perform optimal search and generate the primary frequency modulation control parameters; introducing an exponential smoothing factor into the RBF neural network, combining the mutation differential evolution algorithm and the RBF neural network with fast and fine weights and deviations to construct the MDE-RBF algorithm, and optimizing the primary frequency modulation control parameters through the MDE-RBF algorithm.

[0034] The differential evolution (DE) algorithm is a population-based evolutionary algorithm, but it has disadvantages such as slow convergence, poor accuracy, and easy to fall into local optimality. Therefore, the mutation differential evolution (MDE) algorithm is proposed. The mutation differential evolution algorithm specifically includes the following steps: Step 51: Randomly generate the initial population:

[0035] in , , is the scale factor, For the The first generation Individuals, is the perturbation vector, To compare Generation and the first The optimal value determined by the generation group, then Generation Group Mutation to generate intermediates ; The random index and The generation method is to extract twice, generating two sets of random integer indexes and , take the minimum value of each group as the final index:

[0036]

[0037] Using optimized random indexing 、 Select individuals for mutation operation.

[0038] Step 52: Generate a random factor and extract different strategies for searching based on the value of the random factor to avoid the search falling into a local optimum. The formula is:

[0039] in, is a random factor and , represents a random vector on the interval, whose length is equal to the number of variables, and are the upper and lower bounds of the objective function; The introduction of random factors and index secondary extraction strategies enables the algorithm to better retain dominant individuals, thereby improving the algorithm's convergence speed, accuracy, and global search capabilities. Step 53: Generation Group Individuals and their variant intermediates Cross-processing between:

[0040] in, is the crossover probability, which is A random vector of for A random integer; Step 54: Use the greedy algorithm on the individuals obtained by crossover. If the individual after crossover is the optimal solution, the hybridized individual is added to the next generation. Otherwise, the original individual is added to the next generation and continues processing:

[0041] Introducing an exponential smoothing factor into the RBF neural network to quickly refine weights and deviations involves the following steps: Step 61: Introduce the conjugate gradient method to modify parameters in the RBF neural network 、 and ,in is the node center of the password layer neuron, is the reference width, is the connection weight between the hidden layer and the output layer; and sets the maximum number of iterations and the loop stop condition, i.e. the objective function Not greater than the error value ; Step 62: Based on the original conjugate gradient descent method, the RBF neural network algorithm introduces a smoothing factor and superimposes the two phases in the search direction as , and the weight is less than or equal to 1, which effectively avoids data turbulence in the descent process of the objective function and improves the learning speed of the model; the new gradient search calculation formula is:

[0042] in, For parameters No. The search direction vector of the iteration, is the neuron index, is the index of the iteration number, is the objective function Parameters The gradient, is the smoothing factor, when Time smoothing factor Determined by the Polak-Ribière formula:

[0043] in, is the objective function Parameters In the The gradient of the iteration, is the objective function Parameters In the The gradient of the iteration, express The transpose of express The transpose of Step 63: Calculate the adaptive step size:

[0044] in, is the correction multiplier and , Step 64: Iteratively update parameters:

[0045] in, 、 、 The three parameters are the learning rate of the node center, the expansion constant, the connection weight and , is the number of iterations.

[0046] For example, in order to test the effectiveness and processing speed of the improved RBF neural network, the neural network and PID control are combined. The PID controller based on the RBF neural network adopts an incremental structure. The increment of time is;

[0047] in, for The output increment of the PID controller at time is the proportional gain, is the integral gain, is the differential gain, for Time system error, for Time system error, for Time system error; exist At the moment, the weight vector adjustment formula is:

[0048] in, for Moment The weight adjustment increment, for Moment weight values, for Moment weight values; The output of the neural network is:

[0049] in, For RBF neural network The predicted output value at time t, For the The connection weights from hidden layer nodes to the output layer, For the The output of a radial basis function (RBF); Introduce the error square function as the performance indicator function:

[0050] in, and As reference input and output respectively, for The actual output value of the system at time for Performance indicators at all times, is the model error; The weight correction is:

[0051] in, For learning efficiency, for Moment The amount of weight adjustment, The control increment for the neural network output The partial derivative of The neural network input features.

[0052] The optimization of primary frequency modulation control parameters using the MDE-RBF algorithm specifically includes the following steps: Step 71: In MDE and RBF, set the population , number of iterations , scale factor , crossover probability , the center point of the basis function , basis width vector , and the weights from the hidden layer to the output layer ; Step 72: Sample the reference input and output and calculate the system error ;in, for The actual output value of the system at time for The predicted output value of the system at that moment; Step 73: RBF neural network in time The following normalizes the weights and increments the PID related parameters: , , , , , , , ;in, is the time difference of the output of the first RBF neuron, For control quantity The instantaneous change of is the PID proportional coefficient of the previous moment, is the PID integral coefficient of the previous moment, is the PID differential coefficient of the previous moment, For the moment The systematic error of is the rate of change of error, is the acceleration of the error; Step 74: Use the conjugate gradient method to calculate the center point of the basis function , base width parameter and the initial weights of RBF ; Step 75: Calculate the inverse of the mean square error as the fitness function:

[0053] in, is the mean square error; Step 76: Get the current time 、 、 And get the next update result; Step 77: Analyze the output values 、 、 Whether the requirements are met, if so, the optimization process ends; otherwise, the iterative parameter optimization process continues based on the MDE algorithm; When Python is used for simulation and the input is a unit step signal, the response curves of DE-PID control, MDE-PID control and MDE-RBF-PID control are as follows: Figure 5 The original model output and the PID parameters tuned by the three algorithms and their main performances are shown in Table 1.

[0054] Table 1 PID parameter tuning results

[0055] Depend on Figure 5 As can be seen from Table 1, the main performance values ​​of MDE-RBF-PID control are better than those of the other two control algorithms.

[0056] To verify the effectiveness and feasibility of the proposed MDE-RBF-PID controller in quickly stabilizing when encountering disturbances, based on some constructed hydropower unit simulation models, the processed frequency signal is input into the model to simulate the MDE-RBF-PID control effect, such as Figure 6It can be seen that the frequency input response curve of the simulation model controlled by the traditional PID controller shows large fluctuations, while the response curve is smoother when controlled by the MDE-RBF-PID controller and can recover to a stable state more quickly. This shows that compared with the traditional PID controller, the MDE-RBF-PID controller adopted in the present invention has better dynamic performance and stability under frequency input disturbances. In addition, the smoothness of the response curve also indicates that the system has good robustness and is not easily affected by external disturbances and falls into an oscillatory or unstable state.

[0057] Step 4: Evaluate the performance of the primary frequency modulation result by the ratio of the actual integrated power to the theoretical integrated power.

[0058] The calculation formula for the ratio of actual integrated power to theoretical integrated power is:

[0059] in, is the primary frequency modulation integrated power ratio, which represents the ratio of actual integrated power to theoretical integrated power; Under the condition of stable operation of hydropower unit with load, when the absolute value of frequency step disturbance is not less than 0.07Hz, the primary frequency modulation integral power ratio The calculation formula is:

[0060] Among them, is the actual integrated power, is the theoretical integrated power, is the rated power of the unit, is the rated frequency, is the compensation ratio, is the instantaneous value of the unit’s active power, It is the initial steady-state reference value of the unit’s active power before a frequency regulation action. is the integration interval (s), is the integration time; is the effective frequency deviation, and the calculation formula is:

[0061] in, is the rated frequency, is the generator outlet frequency, This is the artificial frequency or speed dead zone.

[0062] When the grid frequency exceeds the dead-zone frequency threshold (0.05Hz), the duration exceeds the specified over-action delay (15s), and the intermediate frequency peak exceeds the evaluation threshold (0.07Hz), the hydropower unit participates in a frequency regulation action evaluation.

[0063] For small disturbances, within a frequency regulation evaluation cycle, only one assessment is conducted to see whether the actual frequency regulation output change matches the sign of the system frequency deviation value. If the high frequency output decreases or the low frequency output increases, it is considered qualified.

[0064] For large disturbances, a segmented assessment method is adopted, with 2 minutes as an assessment period. When the system frequency exceeds the primary frequency regulation dead zone within 30s, 60s, and 120s, if the actual output change and the system frequency deviation value have opposite signs and the ratio of the integrated power of the actual primary frequency regulation action to the integrated power of the theoretical action is greater than 30%, 50%, and 80% respectively, then the statistics are qualified, otherwise the evaluation is unqualified. The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for optimizing and evaluating primary frequency regulation control parameters of a hydropower unit, characterized in that: The method comprises the following steps: Step 1: Obtain the test data of the hydropower unit and preprocess the data; Step 2: Analyze the primary frequency regulation characteristics of the hydropower unit and build a hydropower unit regulation system model; Step 3: Construct the MDE-RBF algorithm and optimize the primary frequency modulation PID control parameters through the MDE-RBF algorithm; Step 4: Evaluate the performance of the primary frequency modulation result by the ratio of the actual integrated power to the theoretical integrated power.

2. A method for optimizing and evaluating primary frequency regulation control parameters of a hydropower unit according to claim 1, characterized in that: The data preprocessing in step 1 is to remove the steady-state value of the data. The specific formula is: in, is the processed signal sequence, is the original signal sequence, is the total number of data points in the time period before the frequency step perturbation, is the first points.

3. A method for optimizing and evaluating primary frequency regulation control parameters of a hydropower unit according to claim 1, characterized in that: The open-loop transfer function of the hydropower unit regulation system in step 2 is: in 、 、 They are the proportional gain, integral gain, and differential gain of the PID controller of the turbine control system of the hydraulic turbine. is the Laplace operator, is the inertia time constant, is the adjustment coefficient, is the frequency deviation amplification factor, is the time constant of the frequency measurement link.

4. A method for optimizing and evaluating primary frequency regulation control parameters of a hydropower unit according to claim 1, characterized in that: Step 3 is as follows: using the mutation differential evolution algorithm to perform optimal search and generate the primary frequency modulation control parameters; introducing an exponential smoothing factor into the RBF neural network, combining the mutation differential evolution algorithm and the RBF neural network with fast and fine weights and deviations to construct the MDE-RBF algorithm, and optimizing the primary frequency modulation control parameters through the MDE-RBF algorithm.

5. A method for optimizing and evaluating primary frequency regulation control parameters of a hydropower unit according to claim 4, characterized in that: The mutation differential evolution algorithm specifically includes the following steps: Step 51: Randomly generate the initial population: in , , is the scale factor, For the The first generation Individuals, is the perturbation vector, To compare Generation and the first The optimal value determined by the generation group; The random index and The generation method is to extract twice, generating two sets of random integer indexes and , take the minimum value of each group as the final index: Step 52: Generate a random factor and extract different strategies for search based on the value of the random factor: in, is a random factor and , represents a random vector on the interval, whose length is equal to the number of variables, and are the upper and lower bounds of the objective function; Step 53: Generation Group Individuals and their variant intermediates Cross-processing between: in, is the crossover probability, which is A random vector of for A random integer; Step 54: Use the greedy algorithm on the individuals obtained by crossover. If the individual after crossover is the optimal solution, the hybridized individual is added to the next generation. Otherwise, the original individual is added to the next generation and continues processing:

6. A method for optimizing and evaluating primary frequency regulation control parameters of a hydropower unit according to claim 5, characterized in that: Introducing an exponential smoothing factor into the RBF neural network to quickly refine weights and deviations involves the following steps: Step 61: Introduce the conjugate gradient method to modify parameters in the RBF neural network 、 and , and set the maximum number of iterations and error threshold; is the node center of the password layer neuron, is the reference width, is the connection weight between the hidden layer and the output layer; Step 62: Introduce the smoothing factor and generate a new gradient search calculation formula: in, For parameters No. The search direction vector of the iteration, is the neuron index, is the index of the iteration number, is the objective function Parameters The gradient, is the smoothing factor, when Time smoothing factor Determined by the Polak-Ribière formula: in, is the objective function Parameters In the The gradient of the iteration, is the objective function Parameters In the The gradient of the iteration, express The transpose of express The transpose of Step 63: Calculate the adaptive step size: in, is the correction multiplier and , Step 64: Iteratively update parameters: in, 、 、 The three parameters are the learning rate of the node center, the expansion constant, the connection weight and , is the number of iterations.

7. A method for optimizing and evaluating primary frequency regulation control parameters of a hydropower unit according to claim 6, characterized in that: The optimization of primary frequency modulation control parameters using the MDE-RBF algorithm specifically includes the following steps: Step 71: In MDE and RBF, set the population , number of iterations , scale factor , crossover probability , the center point of the basis function , basis width vector , and the weights from the hidden layer to the output layer ; Step 72: Sample the reference input and output and calculate the system error ; Step 73: RBF neural network in time Normalize weights and increment PID related parameters; Step 74: Use the conjugate gradient method to calculate the center point of the basis function , base width parameter and the initial weights of RBF ; Step 75: Calculate the inverse of the mean square error as the fitness function; Step 76: Get the current time 、 、 And get the next update result; Step 77: Analyze the output values 、 、 Whether the requirements are met, if so, the optimization process ends; otherwise, the iterative parameter optimization process continues based on the MDE algorithm.

8. A method for optimizing and evaluating primary frequency regulation control parameters of a hydropower unit according to claim 1, characterized in that: The calculation formula for the ratio of actual integrated power to theoretical integrated power in step 4 is: in, is the primary frequency modulation integrated power ratio, which represents the ratio of actual integrated power to theoretical integrated power. The calculation formula is: Among them, is the actual integrated power, is the theoretical integrated power, is the rated power of the unit, is the rated frequency, is the compensation ratio, is the instantaneous value of the unit’s active power, It is the initial steady-state reference value of the unit’s active power before a frequency regulation action. is the integration interval (s), is the integration time; is the effective frequency deviation, and the calculation formula is: in, is the rated frequency, is the generator outlet frequency, This is the artificial frequency or speed dead zone.

Images