Optimal selection method for grouping primary frequency modulation strategy of small water motor group

Abstract

The invention provides a small water motor group primary frequency modulation strategy optimal selection method, and relates to the field of power system operation and control technology, and the method comprises the steps: S1, collecting operation data of five influence factors of a primary frequency modulation strategy; s2, determining a primary frequency modulation strategy to be used; s3, determining the influence weight of each influence factor according to the application scene; s4, determining standardized score values of five influence factors corresponding to each primary frequency modulation strategy, and judging whether the standardized score values meet constraint conditions or not; s5, calculating a comprehensive score of the primary frequency modulation strategy meeting the constraint condition according to the influence weight and the standardized score value, and taking the strategy with the highest score as an optimal strategy as a default execution strategy; and S6, finely adjusting the influence weight according to the actual operation state of the unit, calculating a comprehensive score of each primary frequency modulation strategy, obtaining a new optimal strategy, and switching the frequency modulation strategies. The method has the beneficial effects that the primary frequency modulation strategy can be more suitable for an application scene, and the operation state is considered.

Description

Technical Field

[0001] This invention relates to the field of power system operation and control technology, specifically to an optimal selection method for primary frequency regulation strategy of small hydropower unit. Background Technology

[0002] Small hydropower units have relatively weak primary frequency regulation capabilities under isolated grid operation or large disturbance conditions due to their smaller inertia, slower speed control system response, and limited adjustment margin. This poses a greater risk of excessive frequency deviation or even system instability.

[0003] Currently, commonly used primary frequency regulation methods include traditional PID control, feedforward compensation, energy storage coordinated regulation, virtual inertia injection, multi-machine coordinated optimization, and other strategies. Each strategy has its applicable scenarios and disadvantages. Therefore, how to select a suitable strategy according to different disturbances and frequency regulation requirements has become a difficult problem in the primary frequency regulation of small hydropower units. Summary of the Invention

[0004] The main objective of this invention is to provide an optimal selection method for the primary frequency regulation strategy of a small hydropower unit, thereby solving the problems mentioned in the background art.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: an optimal selection method for primary frequency regulation strategy of small hydropower unit, comprising the following steps: S1. Collect operational data of five influencing factors of the primary frequency modulation strategy, including: maximum frequency deviation, steady-state recovery time, peak overshoot power, control cost, and control system robustness. S2. Determine the primary frequency modulation strategy to be used; S3. Determine the influence weight of each influencing factor based on the application scenario; S4. Determine the standardized score values ​​of the five influencing factors corresponding to each primary frequency modulation strategy, and judge whether each strategy meets the constraints. S5. Calculate the comprehensive score of the primary frequency regulation strategy that meets the constraints based on the influence weight and standardized score value. Select the strategy with the highest comprehensive score as the optimal strategy and use it as the default execution strategy. S6. Fine-tune the influence weights according to the actual operating status of the unit, recalculate the comprehensive score of each primary frequency regulation strategy, and obtain a new optimal strategy. When the new optimal strategy is inconsistent with the currently executed optimal strategy, switch the frequency regulation strategy and execute the new optimal strategy.

[0006] Furthermore, in step S1, the influencing factors include: maximum frequency deviation, steady-state recovery time, peak overshoot power, control cost, and control system robustness.

[0007] Furthermore, in step S2, the primary frequency regulation strategy includes: traditional PID speed controller control strategy, enhanced PID plus feedforward compensation control strategy, energy storage cooperative frequency regulation control strategy, virtual inertia control strategy, and multi-machine coordinated joint frequency regulation strategy.

[0008] Furthermore, the application scenarios in step S3 include: cost-sensitive scenarios, power grid stability-first scenarios, multi-unit cluster scenarios, and high-reliability redundancy scenarios.

[0009] Furthermore, in step S4, the evaluation process for the standardized score value is as follows: S501. Build a primary frequency regulation model of a small hydropower unit in simulation software, simulate load disturbance scenarios, and record the frequency deviation curve, power response curve, and adjustment time of each strategy. In small hydropower units, actual machine tests were conducted, and the measured values ​​of the power grid frequency were recorded; S502. Extract the maximum values ​​of the five influencing factors. X max and minimum value X min ; S503. The five influencing factors are divided into positive and negative indicators, and the scoring process is as follows: For positive indicators, we have: (1); in, t Standardized score; For negative indicators, we have: (2); S504, When there is t When the value is greater than 10, its standardized score is constrained to be 10 points; When there is t When the value is less than 0, the standardized score of the constraint is 0.

[0010] Furthermore, in step S5, the formula for calculating the comprehensive score is as follows: (3); in, H For the overall score, x i The standardized score values ​​are for the corresponding influencing factors. w i The weight values ​​of the influencing factors.

[0011] Furthermore, in step S6, the BP neural network algorithm is used to fine-tune the influence weights; A backpropagation (BP) neural network consists of an input layer, a hidden layer, and an output layer. The input layer contains 5 neurons; The hidden layers include: hidden layer 1 and hidden layer 2; Hidden layer 1 has 40 neurons, and its activation function is the Leaky ReLU function, whose expression is: (4); in, α The activation coefficient should be a small positive number. The output of hidden layer 1 can then be expressed as: (5); in, x i For the first in a set of input data i One element, h j1 For the first hidden layer 1 j One output, w ij1 For the first hidden layer 1 i The input data corresponds to the first... j The weights of each neuron, b j1 For the first hidden layer 1 j Bias of each neuron; Hidden layer 2 has 32 neurons, and its activation function is also the swish function, whose expression is: (6); If the output of hidden layer 1 is used as the input of hidden layer 2, then the output of hidden layer 2 can be expressed as: (7); in, h j2 For the second hidden layer j One output, w ij2 For the second hidden layer i The input data corresponds to the first... j The weights of each neuron, b j2 For the second hidden layer j Bias of each neuron; The output layer contains 5 neurons, corresponding to the influence weights of the 5 influencing factors. The weighted input of the output layer can be represented as: (8); in, y in,j For the output layer's first j A weighted input, woutj For the output layer's first j The weights of each neuron, b outj For the output layer j Bias of each neuron; The softmax function is chosen as the activation function for the output layer, and its expression is as follows: (9); in, y out,j For the output layer's first j One output, t This is a temperature coefficient used to control the steepness of the output.

[0012] Furthermore, the training process of the BP neural network is as follows: S801. Collect historical operation data to obtain a sample set. Each sample set includes: <input features, output labels>: S802. Perform preprocessing operations on the data in the sample, including: outlier removal, data imputation, and data normalization. S803, Initialize weights and biases, and determine the values ​​of hyperparameters; S804. Perform batch training on the sample set to obtain the output; S805. After training all samples in the sample set once, calculate the loss value based on the output of the loss function. The expression for the loss function is: (10); (11); in, N Where is the total number of samples, and Loss is the total loss value. MSE j For the first j The loss value for each weight; They are the first k The first sample j The predicted value and label value for each weight; S806. Iteratively update the weights and biases, the process of which is as follows: (12); in, w , b These are the weights and biases of each layer in the neural network. or The learning rate; S807. Training stops after the set number of iterations has been reached. Training ends early when the total loss value and the weighted loss value meet the early stopping condition.

[0013] Furthermore, in step S4, the constraint condition is: Constraint 1: The adjustment process must not exceed the maximum and minimum power output range of the unit; Constraint 2: The response speed must not exceed the system's required settling time; Constraint 3: Satisfy the constraints of speed limiting of the governor guide vanes, structural oscillation, and stability of the hydraulic-mechanical coupling system.

[0014] Furthermore, for the cost-sensitive scenario, the frequency modulation requirements are: control cost > robustness > frequency deviation > recovery time > peak overshoot power; In scenarios prioritizing power grid stability, the frequency regulation requirements are: maximum frequency deviation > peak overshoot power > robustness > recovery time > cost; In multi-unit cluster scenarios, the frequency regulation requirements are: robustness > recovery time > frequency deviation > peak overshoot power > cost; In high-reliability redundancy scenarios, the frequency regulation requirements are: robustness > overshoot power > frequency deviation > recovery time > cost.

[0015] The beneficial effects of this invention are as follows: (1) Determine the weight of influencing factors according to different application scenarios and score the primary frequency modulation strategy to select the optimal solution, so that the primary frequency modulation strategy can be more suitable for the application scenario. (2) Artificial intelligence algorithms are used to fine-tune the influence weights based on information such as frequency offset, taking into account the actual operating status of the unit. Attached Figure Description

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart of the steps of the method of the present invention. Detailed Implementation

[0017] like Figure 1 As shown in the figure, a method for optimal selection of primary frequency regulation strategy for small hydropower unit is described below, with the following steps: S1. Determine the five influencing factors of the primary frequency regulation strategy, including: maximum frequency deviation, steady-state recovery time, peak overshoot power, control cost, and control system robustness.

[0018] S2. Determine the primary frequency regulation strategy to be used, including: traditional PID speed controller control strategy, enhanced PID plus feedforward compensation control strategy, energy storage cooperative frequency regulation control strategy, virtual inertia control strategy, and multi-machine coordinated joint frequency regulation strategy.

[0019] S3. Determine the influence weight of each influencing factor based on the application scenario; For cost-sensitive scenarios, the frequency regulation requirements are: cost control > robustness > frequency deviation > recovery time > peak overshoot power; In scenarios prioritizing power grid stability, the frequency regulation requirements are: maximum frequency deviation > peak overshoot power > robustness > recovery time > cost; In multi-unit cluster scenarios, the frequency regulation requirements are: robustness > recovery time > frequency deviation > peak overshoot power > cost; In high-reliability redundancy scenarios, the frequency regulation requirements are: robustness > overshoot power > frequency deviation > recovery time > cost.

[0020] S4. Determine the standardized score values ​​for the five influencing factors corresponding to each primary frequency modulation strategy; the evaluation process is as follows: A1. Build a primary frequency regulation model of a small hydropower unit in simulation software, simulate load disturbance scenarios, and record the frequency deviation curve, power response curve, and adjustment time of each strategy. In small hydropower units, actual machine tests were conducted, and the measured values ​​of the power grid frequency were recorded; A2. Extract the maximum values ​​of the five influencing factors. X max and minimum value X min ; A3. The five influencing factors are divided into positive and negative indicators, and the scoring process is as follows: For positive indicators, we have: (1); in, t Standardized score; For negative indicators, we have: (2); Of the five influencing factors, the maximum frequency deviation, steady-state recovery time, peak overshoot power, and control cost are negative indicators, while the robustness of the control system is a positive indicator. Specifically, regarding the robustness of the control system, if the strategy fails once out of 10 times, its value is 9; if it fails twice, its value is 8. When its maximum value is 10, its minimum value is 5, and the value corresponding to the virtual inertia control strategy is 8, then the standardized score for the robustness of the control system corresponding to the virtual inertia control strategy is: (3).

[0021] A4, when there is t When the value is greater than 10, its standardized score is constrained to be 10 points; When there is t When the value is less than 0, the standardized score of the constraint is 0. We determine whether each strategy satisfies the following constraints: Constraint 1: During the adjustment process, the output power should be within the range of the unit's maximum and minimum power output; Constraint 2: The response time must not exceed the system's required settling time; Constraint 3: Satisfy the constraints of speed limiting of the governor guide vanes, structural oscillation, and stability of the hydraulic-mechanical coupling system.

[0022] S5. Calculate the comprehensive score of the primary frequency regulation strategy that meets the constraints based on the influence weights and standardized score values. The calculation process is as follows: (4); in, H For the overall score, x i The standardized score values ​​are for the corresponding influencing factors. w i These are the weight values ​​of the influencing factors; The strategy with the highest overall score is selected as the optimal strategy and used as the default execution strategy. S6. Fine-tune the influence weights according to the actual operating status of the unit, recalculate the comprehensive score of each primary frequency regulation strategy, and obtain a new optimal strategy. When the new optimal strategy is inconsistent with the currently executed optimal strategy, switch the frequency regulation strategy and execute the new optimal strategy.

[0023] Example 2 To further illustrate with reference to Example 1, a method for optimal selection of primary frequency regulation strategy for small hydropower generator sets is as follows: In step S6, a BP neural network algorithm is used to fine-tune the influencing weights; A backpropagation (BP) neural network consists of an input layer, a hidden layer, and an output layer. The input layer contains 5 neurons, which correspond to the maximum frequency deviation, steady-state recovery time, peak overshoot power, control cost, and control system robustness, respectively. The hidden layers include: hidden layer 1 and hidden layer 2; Hidden layer 1 has 40 neurons, and its activation function is the Leaky ReLU function, whose expression is: (5); in, α The activation coefficient is usually a small positive number; 0.01 is a good choice here. The output of hidden layer 1 can then be expressed as: (6); in, x i For the first in a set of input datai One element, h j1 For the first hidden layer 1 j One output, w ij1 For the first hidden layer 1 i The input data corresponds to the first... j The weights of each neuron, b j1 For the first hidden layer 1 j Bias of each neuron; Hidden layer 2 has 32 neurons, and its activation function is also the swish function, whose expression is: (7); If the output of hidden layer 1 is used as the input of hidden layer 2, then the output of hidden layer 2 can be expressed as: (8); in, h j2 For the second hidden layer j One output, w ij2 For the second hidden layer i The input data corresponds to the first... j The weights of each neuron, b j2 For the second hidden layer j Bias of each neuron; The output layer contains 5 neurons, corresponding to the influence weights of the 5 influencing factors. The weighted input of the output layer can be represented as: (9); in, y in,j For the output layer's first j A weighted input, w outj For the output layer's first j The weights of each neuron, b outj For the output layer j Bias of each neuron; The softmax function is chosen as the activation function for the output layer, and its expression is as follows: (10); in, y out,j For the output layer's first j One output, t This is a temperature coefficient used to control the steepness of the output.

[0024] The training process is as follows: S801. Collect historical operation data to obtain a sample set. Each sample set includes: <input features, output labels>: S802. Perform preprocessing operations on the data in the sample, including: outlier removal, data imputation, and data normalization. S803. Initialize weights and biases, and determine the values ​​of hyperparameters, including: 500 iterations, learning rate of 0.001, decay coefficient of 0.5, learning rate decays once every 100 iterations, batch size of 32, regularization coefficient of 0.001, and temperature coefficient of 0.8. S804. Perform batch training on the sample set to obtain the output; S805. After training all samples in the sample set once, calculate the loss value based on the output of the loss function. The expression for the loss function is: (11); (12); in, N Where is the total number of samples, and Loss is the total loss value. MSE j For the first j The loss value for each weight; They are the first k The first sample j The predicted value and label value of each weight, l It is the regularization coefficient; S806. Iteratively update the weights and biases, the process of which is as follows: (13); in, w , b These are the weights and biases of each layer in the neural network. or The learning rate; S807. Training stops after the set number of iterations has been reached. Training ends early when the total loss value and the weighted loss value meet the early stopping condition.

[0025] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method for optimal selection of primary frequency regulation strategy for small hydropower unit assembly, characterized in that: Includes the following steps: S1. Collect operational data of five influencing factors of the primary frequency modulation strategy, including: maximum frequency deviation, steady-state recovery time, peak overshoot power, control cost, and control system robustness. S2. Determine the primary frequency modulation strategy to be used; S3. Determine the influence weight of each influencing factor based on the application scenario; S4. Determine the standardized score values ​​of the five influencing factors corresponding to each primary frequency modulation strategy, and judge whether each strategy meets the constraints. S5. Calculate the comprehensive score of the primary frequency regulation strategy that meets the constraints based on the influence weight and standardized score value. Select the strategy with the highest comprehensive score as the optimal strategy and use it as the default execution strategy. S6. Fine-tune the influence weights according to the actual operating status of the unit, recalculate the comprehensive score of each primary frequency regulation strategy, and obtain a new optimal strategy. When the new optimal strategy is inconsistent with the currently executed optimal strategy, switch the frequency regulation strategy and execute the new optimal strategy.

2. The optimal selection method for primary frequency regulation strategy of a small hydropower unit according to claim 1, characterized in that: In step S1, the influencing factors include: maximum frequency deviation, steady-state recovery time, peak overshoot power, control cost, and control system robustness.

3. The optimal selection method for primary frequency regulation strategy of a small hydropower unit as described in claim 1, characterized in that: In step S2, the primary frequency regulation strategy includes: traditional PID speed controller control strategy, enhanced PID plus feedforward compensation control strategy, energy storage cooperative frequency regulation control strategy, virtual inertia control strategy, and multi-machine coordinated joint frequency regulation strategy.

4. The optimal selection method for primary frequency regulation strategy of a small hydropower unit according to claim 1, characterized in that: The application scenarios in step S3 include: cost-sensitive scenarios, power grid stability priority scenarios, multi-unit cluster scenarios, and high-reliability redundancy scenarios.

5. The optimal selection method for primary frequency regulation strategy of a small hydropower unit as described in claim 1, characterized in that: In step S4, the process of evaluating the standardized score is as follows: S501. Build a primary frequency regulation model of a small hydropower unit in simulation software, simulate load disturbance scenarios, and record the frequency deviation curve, power response curve, and adjustment time of each strategy. In small hydropower units, actual machine tests were conducted, and the measured values ​​of the power grid frequency were recorded; S502. Extract the maximum values ​​of the five influencing factors. X max and minimum value X min ; S503. The five influencing factors are divided into positive and negative indicators, and the scoring process is as follows: For positive indicators, we have: （1）； in, t Standardized score; For negative indicators, we have: （2）； S504, When there is t When the value is greater than 10, its standardized score is constrained to be 10 points; When there is t When the value is less than 0, the standardized score of the constraint is 0.

6. The optimal selection method for primary frequency regulation strategy of a small hydropower unit as described in claim 1, characterized in that: In step S5, the formula for calculating the comprehensive score is as follows: （3）； in, H For the overall score, x i The standardized score values ​​are for the corresponding influencing factors. w i The weight values ​​of the influencing factors.

7. The optimal selection method for primary frequency regulation strategy of a small hydropower unit as described in claim 1, characterized in that: In step S6, the BP neural network algorithm is used to fine-tune the influence weights; A backpropagation (BP) neural network consists of an input layer, a hidden layer, and an output layer. The input layer contains 5 neurons; The hidden layers include: hidden layer 1 and hidden layer 2; Hidden layer 1 has 40 neurons, and its activation function is the Leaky ReLU function, whose expression is: （4）； in, α The activation coefficient should be a small positive number. The output of hidden layer 1 can then be expressed as: （5）； in, x i For the first in a set of input data i One element, h j1 For the first hidden layer 1 j One output, w ij1 For the first hidden layer 1 i The input data corresponds to the first... j The weights of each neuron, b j1 For the first hidden layer 1 j Bias of each neuron; Hidden layer 2 has 32 neurons, and its activation function is also the swish function, whose expression is: （6）； If the output of hidden layer 1 is used as the input of hidden layer 2, then the output of hidden layer 2 can be expressed as: （7）； in, h j2 For the second hidden layer j One output, w ij2 For the second hidden layer i The input data corresponds to the first... j The weights of each neuron, b j2 For the second hidden layer j Bias of each neuron; The output layer contains 5 neurons, corresponding to the influence weights of the 5 influencing factors. The weighted input of the output layer can be represented as: （8）； in, y in,j For the output layer's first j A weighted input, w outj For the output layer's first j The weights of each neuron, b outj For the output layer j Bias of each neuron; The softmax function is chosen as the activation function for the output layer, and its expression is as follows: （9）； in, y out,j For the output layer's first j One output, τ This is a temperature coefficient used to control the steepness of the output.

8. The optimal selection method for primary frequency regulation strategy of small hydropower unit as described in claim 7, characterized in that: The training process of the BP neural network is as follows: S801. Collect historical operation data to obtain a sample set. Each sample set includes: <input features, output labels>: S802. Perform preprocessing operations on the data in the sample, including: outlier removal, data imputation, and data normalization. S803, Initialize weights and biases, and determine the values ​​of hyperparameters; S804. Perform batch training on the sample set to obtain the output; S805. After training all samples in the sample set once, calculate the loss value based on the output of the loss function. The expression for the loss function is: （10）； （11）； in, N Where is the total number of samples, and Loss is the total loss value. MSE j For the first j The loss value for each weight; They are the first k The first sample j The predicted value and label value for each weight; S806. Iteratively update the weights and biases, the process of which is as follows: （12）； in, w , b These are the weights and biases of each layer in the neural network. η The learning rate; S807. Training stops after the set number of iterations has been reached. Training ends early when the total loss value and the weighted loss value meet the early stopping condition.

9. The optimal selection method for primary frequency regulation strategy of small hydropower unit as described in claim 1, characterized in that: In step S4, the constraint condition is: Constraint 1: The adjustment process must not exceed the maximum and minimum power output range of the unit; Constraint 2: The response speed must not exceed the system's required settling time; Constraint 3: Satisfy the constraints of speed limiting of the governor guide vanes, structural oscillation, and stability of the hydraulic-mechanical coupling system.

10. The optimal selection method for primary frequency regulation strategy of small hydropower unit as described in claim 4, characterized in that: For the cost-sensitive scenario, the frequency modulation requirements are: control cost > robustness > frequency deviation > recovery time > peak overshoot power; In scenarios prioritizing power grid stability, the frequency regulation requirements are: maximum frequency deviation > peak overshoot power > robustness > recovery time > cost; In multi-unit cluster scenarios, the frequency regulation requirements are: robustness > recovery time > frequency deviation > peak overshoot power > cost; In high-reliability redundancy scenarios, the frequency regulation requirements are: robustness > overshoot power > frequency deviation > recovery time > cost.

Images