Frequency control method of isolated network system under load disturbance based on safety and stability constraints

Abstract

The invention provides a safety and stability constraint-based frequency control method for an isolated network system under load disturbance, and relates to the technical field of power system stability and control. The method comprises the following steps: S1, establishing a synchronous machine and networking type energy storage system joint frequency response model; s2, quickly estimating load disturbance according to the response model and an estimation algorithm; s3, judging whether the power of the synchronous generator is out of limit or not, and solving the power which needs to be injected immediately; s4, allocating coordination power according to the capacity proportion in the multi-VSG system; and S5, adjusting the planned output set value of the synchronous generator according to the average active power of the output end of the synchronous generator. The method has the advantages that the problem of system oscillation caused by frequency instability can be avoided, the calculated amount is reduced, compared with reaction after disturbance or single-factor early warning, early warning factors are more comprehensive, early warning can be achieved earlier, the situation of adjustment failure of the system is avoided, and safety is improved.

Description

Technical Field

[0001] This invention relates to the field of power system stability and control technology, specifically to a frequency control method for islanded grid systems under load disturbances based on safety and stability constraints. Background Technology

[0002] When an isolated grid system is in operation, since it lacks the inertial support and power backup of a large power grid, its frequency regulation relies entirely on its internal power source. Therefore, its power and load balance is easily disrupted, and it is highly susceptible to load disturbances, causing significant frequency fluctuations.

[0003] Traditional islanded frequency regulation methods mostly involve testing after frequency fluctuations occur and then taking measures to stabilize the frequency. Although this can restore the system to stability, the voltage quality will be severely degraded during the fluctuation period. Electrical equipment may experience false triggering, performance degradation, or even damage. Furthermore, it is easy to trigger secondary fluctuations and may lead to a chain of risks such as power supply disconnection and partial system shutdown. Summary of the Invention

[0004] The main objective of this invention is to provide a frequency control method for an isolated network system under load disturbance based on safety and stability constraints, 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: a frequency control method for an isolated network system under load disturbance based on safety and stability constraints, comprising the following steps: S1. Establish a joint frequency response model for synchronous machine and grid-type energy storage system; S2. Quickly estimate load disturbances based on response models and estimation algorithms; S3. Determine whether the power of the synchronous generator exceeds the limit, and calculate the power that needs to be injected immediately; S4. In a multi-VSG system, coordinate power according to capacity ratio; S5. Adjust the planned output setting of the synchronous generator based on the average active power at the output terminal of the synchronous generator.

[0006] Furthermore, the response model includes a physical-side model and a grid-based energy storage model; The physical model is expressed as follows: (1); in, J The inertial constant of the synchronous generator, For rotor electrical angular velocity deviation, D The damping coefficient is... , These are the changes in mechanical power and the changes in electromagnetic power, respectively. R The adjustment rate, The servo time constant of the speed controller. , These are load disturbance and VSG output power change, respectively, and their relationship is as follows: (2); The expression for a grid-type energy storage model is as follows: (3); in, For virtual inertia coefficient, This is the reference change in VSG power. This is the virtual damping coefficient. The droop coefficient is... To coordinate power; Equivalent total inertia and equivalent total damping The expression is: (4); A virtual impedance is introduced to correct the voltage reference value. The correction process is as follows: (5); in, As a reference virtual impedance, S As an indicator of disturbance intensity, The change in frequency For the ideal no-load voltage vector, i The converter output current vector, This is the voltage reference after virtual impedance correction.

[0007] Furthermore, the estimation algorithm is the RLS-BP algorithm; The RLS algorithm is used to determine the initial value of the virtual inertia coefficient, and the BP neural network model is used to determine the change in the virtual inertia coefficient. The expression for the virtual inertia coefficient is: (6); in, The initial value for the virtual inertia coefficient. This represents the change in the virtual inertia coefficient.

[0008] The RLS algorithm determines the initial value of the virtual inertia coefficient as follows: (7); in, k For the number of recursions, For RLS parameter vectors, K This is the RLS gain vector. e For RLS prediction error, For frequency deviation, For RLS regression vectors, P Let covariance matrix be the variance matrix. Forgetting factor; A backpropagation (BP) neural network consists of an input layer, a hidden layer, and an output layer. The inputs to the input layer are frequency deviation, frequency deviation rate of change, total active power deviation on the grid side, and absolute value of RLS prediction error; The output of the output layer is the change in the virtual inertia coefficient.

[0009] Furthermore, load disturbance The maximum value is denoted as It needs to be corrected. The correction process is as follows: (8); in, This is the corrected estimate of the load disturbance. This represents locally measured load mutations. The confidence weight is expressed as follows: (9); in, The distance is Euclidean.

[0010] Furthermore, the detailed process of step S3 is as follows: First, determine the active power prediction value of the synchronous generator in the next second. Its expression is: (10); in, This represents the real-time power of the synchronous generator at the moment before the disturbance. VSG instantaneous available capacity; Frequency and active power prediction values ​​are monitored, and when they reach the triggering conditions, a stability boundary assessment is performed to provide early warning. The assessment process is as follows: The Lyapunov energy function is constructed as follows: (11); in, V For energy, It is the difference between the change in electromagnetic power and the change in mechanical power. The system synchronization power factor; Stability threshold Set to 0.05, when there is When the system tends towards instability, the required power injection value is calculated. To prevent system crashes due to frequency instability, its mathematical expression is: (12); in, This is the emergency power gain coefficient. As the upper limit of energy safety, Let be the energy gradient norm, expressed as: (13).

[0011] Furthermore, the triggering condition is: (14); (15); The triggering condition is met when one of the conditions in equations (14) and (15) is satisfied.

[0012] Furthermore, the process of step S4 is as follows: The expression for coordinated power is: (16); in, The planned output setting value for the synchronous generator; Determine the available capacity of each VSG device, the first i The available capacity of the equipment is denoted as Then the first i Weighting of devices The expression is: (17); in, n This represents the total number of VSG devices. No. i The power handled by this device is: (18); in, For the first i The power handled by the equipment.

[0013] Furthermore, the normal constraint range for the average active power at the output of the synchronous generator is: (19); in, The average active power at the output of the synchronous generator. This refers to the rated power of the synchronous generator.

[0014] Furthermore, regarding the average active power output of the synchronous generator, when it exceeds the normal range, the planned output setting of the synchronous generator is adjusted. The adjustment process is as follows: when When the conditions are within the normal range of constraints, no action is taken; when At that time, The value increases by 0.02Pr; when When this happens, it indicates that the synchronous generator is outputting too much power; at this time, [the generator should be switched off]. The value decreased by 0.02Pr.

[0015] Furthermore, the range of the judgment period for the adjustment period is 20S~60S.

[0016] Beneficial effects: (1) Provide early warnings before load disturbances occur and take measures in advance to stabilize the system frequency and avoid system oscillation problems caused by frequency instability; (2) RLS-BP can fit nonlinear mappings, and the offline BP model can reduce the amount of computation. (3) The factors for early warning include frequency change, frequency change rate and active power prediction. Compared with the reaction after the disturbance occurs or the early warning of a single factor, the early warning factors are more comprehensive and can achieve early warning earlier, thus improving safety. (4) Adjust the planned output setting value of the synchronous generator according to the constraint of the average active power output of the synchronous motor, so as to avoid the synchronous motor output becoming less and less, over-reliance on the output of the energy storage system, and ultimately the problem of regulation failure. Attached Figure Description

[0017] 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 present invention. Detailed Implementation

[0018] Example 1 like Figure 1 As shown, the frequency control method for an isolated network system under load disturbance based on security and stability constraints includes the following steps: S1. Establish a joint frequency response model for synchronous machine and grid-type energy storage system; The response models include: physical-side models and grid-based energy storage models; The physical-side model, i.e., the synchronous generator model, has the following mathematical expression: (1); in, J The inertial constant of the synchronous generator, This refers to the rotor's electrical angular velocity deviation, i.e., the rotor's actual electrical angular velocity. With set value The difference, D The damping coefficient is... , These are the changes in mechanical power and the changes in electromagnetic power, respectively. RThe droop rate represents the percentage increase in generator output for every 1% decrease in frequency. The servo time constant of the speed controller. , These are load disturbance and VSG output power change, respectively. The mathematical expression for the grid-based energy storage model, also known as the virtual synchronous generator (VSG) model, is as follows: (2); in, For virtual inertia coefficient, This is the reference change in VSG power. This is the virtual damping coefficient. The droop coefficient is... To coordinate power; Based on the above physical-side model and grid-type energy storage model, the equivalent total inertia and equivalent total damping can be obtained. and equivalent total damping The expression is: (3); A virtual impedance is introduced to correct the voltage reference value, thereby enhancing the system's resistance to load disturbances, voltage stability, and operational robustness. The correction process is as follows: (4); in, The reference virtual impedance is S, and the disturbance strength index is S. The change in frequency Let i be the ideal no-load voltage vector, and i be the converter output current vector. This is the voltage reference after virtual impedance correction.

[0019] S2. Quickly estimate load disturbances based on response models and estimation algorithms; First, the model is transformed to determine the upper limit of the load disturbance. The specific transformation process is as follows: For the system, its energy gap It can be determined by total energy and kinetic energy. The current kinetic energy W of the system is represented as follows: (5); The energy gap can be filled by the VSG, which then changes the kinetic energy of the VSG. It can be represented as: (6); Combining equations (5) and (6), and adding the electrical power output by the VSG, we get: (7); when As it approaches 0, equation (7) can be transformed into: (8) For equation (2), at the initial instant of the disturbance, the smallest value can be ignored. and those that have not yet responded At this point, if we only care about the magnitude relationship and do not consider the direction, equation (2) can be rewritten as: (9); Substituting equation (9) into equation (8) and deriving the result, we get: (10); From equation (10), it can be seen that when Take the maximum value hour, Take the maximum value , here This is the upper limit of load disturbance; Next, an RLS-BP algorithm is used to adjust the virtual inertia coefficient, thereby realizing the estimation of load disturbance; For the virtual inertia coefficient, its initial value is given by the RLS method, and its change value is given by the BP neural network, then: (11); in, The initial value for the virtual inertia coefficient. This represents the change in the virtual inertia coefficient; The RLS method updates the initial values ​​of the virtual inertia coefficients as follows: (12); in, k For the number of recursions, For RLS parameter vectors, K This is the RLS gain vector. e For RLS prediction error, Frequency deviation, which is the difference between the measured frequency and the rated frequency. is the RLS regression vector, where the elements are frequency deviation, frequency deviation change rate, and grid-side total active power deviation, respectively. Grid-side total active power deviation refers to the difference between the measured total active power at the grid connection point and the reference value 1 ms before the disturbance. P Let covariance matrix be the variance matrix. Forgetting factor; A backpropagation (BP) neural network consists of an input layer, a hidden layer, and an output layer. The input layer consists of 4 neurons, corresponding to the input dimension, and includes 4 feature parameters. Three of them are elements in the RLS regression vector, namely frequency deviation, frequency deviation rate of change, and total active power deviation on the network side. The other is the absolute value of the RLS prediction error. The hidden layer consists of 5 neurons, and the activation function used is the tanh function, whose expression is: (13); in, The hidden layer output vector. The hidden layer input vector, , These are the weight matrix and bias vector of the hidden layer, respectively; The output layer consists of one neuron, which outputs the change in the virtual inertia coefficient, and its expression is: (14); in, , These are the weight vector and the bias, respectively; The BP neural network model is trained offline, and its weights and biases are not updated when it is used. The offline training process of the model is as follows: A1. Collect feature parameter data and perform preprocessing operations on the collected data, including removing erroneous data, filling in missing data, and normalizing the data. A2. Determine the number of samples for the model to be 80,000, and divide them into training set and validation set in an 8:2 ratio. After one round of training on the training set, use the validation set for validation. The number of samples in each batch is 128. A3. Determine the hyperparameters and loss function of the model; The hyperparameters include: a maximum of 300 training epochs, a minimum of 100 training epochs, and a learning rate of 0.005 for the first 80% of batches and 0.001 for the last 20% of batches. The loss function chosen is the MSE function, and its expression is: (15); in ，m The number of samples in each batch. and These are the actual value and the predicted value, respectively. After iterating through the validation set, the average of the validation set loss function values ​​for all batches is calculated; this value is the weighted error. The model's validation set contains 16,000 samples, divided into 125 batches; therefore, the weighted error... The expression is: (16); A4. Input the training set into the input layer in batches for model training, and use stochastic gradient descent to update the weights and biases. That is, update the weights and biases after each batch of samples is trained. The update formula is as follows: (17); in, These are the updated values ​​for the weights and biases. For the weights and biases of each layer, The learning rate; A5. After one round of training on the training set, use the validation set for validation. A6. Repeat steps A4 to A5 until the maximum number of iterations is reached, at which point training stops. Set an early stopping threshold of 0.001. When the number of iterations reaches the minimum number of iterations, observe the weighted error. When the weighted error decreases by less than 0.001 for 20 consecutive iterations, end the training early. Substituting the initial value of the virtual inertia coefficient obtained by the RLS algorithm and the change value of the virtual inertia coefficient output by the BP neural network into equation (11) yields the virtual inertia coefficient. Substituting the virtual inertia coefficient into equation (10) yields the estimated value of the load disturbance. The estimated value obtained here corresponds to the maximum value, i.e. Here it should be rewritten as ; The estimated load disturbance values ​​need to be revised, as follows: (18); in, This is the corrected estimate of the load disturbance. This represents locally measured load mutations. The confidence weight is expressed as follows: (19); in, For Euclidean distance, here e The value used is the final error.

[0020] S3. Determine if the synchronous generator power exceeds the limit and calculate the power that needs to be injected immediately. The specific process is as follows: First, determine the active power prediction value of the synchronous generator in the next second. Its expression is: (20); in, This represents the real-time power of the synchronous generator at the moment before the disturbance. VSG instantaneous available capacity; Frequency and active power prediction values ​​are monitored, and subsequent calculations are performed when a trigger condition is met. The trigger condition is as follows: (twenty one); (twenty two); When one of the conditions in equations (21) and (22) is met, i.e. the triggering condition is met, a stability boundary assessment needs to be performed to provide early warning. The assessment process is as follows: The Lyapunov energy function is constructed as follows: (twenty three); in, V For energy, It is the difference between the change in electromagnetic power and the change in mechanical power. The system synchronization power factor; Stability threshold Set to 0.05, when there is When the system tends towards instability, the required power injection value is calculated. To prevent system crashes due to frequency instability, its mathematical expression is: (twenty four); in, This is the emergency power gain coefficient, with a value range of [0.5, 1]. As the upper limit for energy safety, a value of 0.1 is preferable. Let be the energy gradient norm, expressed as: (25).

[0021] S4. In a multi-VSG system, coordinate power according to capacity ratio; The expression for coordinated power is: (26); in, The planned output setting value for the synchronous generator; Determine the available capacity of each VSG device, the first i The available capacity of the equipment is denoted as Then the first i Weighting of devices The expression is: (27); in, n This represents the total number of VSG devices. Then the first i The power handled by this device is: (28); in, For the first iThe power handled by the equipment.

[0022] S5. Based on the average active power output of the synchronous generator. Planned output setpoint Adjustments are made; when subjected to prolonged load disturbances, to ensure frequency stability, the system continuously increases the output of the VSG while continuously decreasing the output of the synchronous generator. With power output constantly being reduced, the synchronous generator has been operating at low power for extended periods, while the energy storage battery is under high consumption. Therefore, it is necessary to... Make adjustments to ensure This is within the normal range of constraints. The normal range of constraints is: , among them It is the rated power of the synchronous generator, when When the value exceeds the normal range, adjustments are made. The adjustment process is as follows: A judgment is made every 30 seconds, when... When the conditions are within the normal range of constraints, no action is taken; when At that time, The value increases by 0.02Pr; when When this happens, it indicates that the synchronous generator is outputting too much power; at this time, [the generator should be switched off]. The value decreased by 0.02Pr.

[0023] 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 frequency control method for an isolated power system under load disturbance based on security stability constraints, characterized in that, Includes the following steps: S1. Establish a joint frequency response model for synchronous machine and grid-type energy storage system; S2. Quickly estimate load disturbances based on response models and estimation algorithms; S3. Determine whether the power of the synchronous generator exceeds the limit, and calculate the power that needs to be injected immediately; S4. In a multi-VSG system, coordinate power according to capacity ratio; S5. Adjust the planned output setting of the synchronous generator based on the average active power at the output terminal of the synchronous generator.

2. The control method according to claim 1, characterized by, The response model includes a physical-side model and a grid-based energy storage model; The physical model is expressed as follows: （1）； wherein, J is the inertia constant of the synchronous generator, is the rotor electrical angular speed deviation, D is the damping coefficient, , are the mechanical power variation and the electromagnetic power variation, respectively, R is the droop rate, is the governor servo time constant, , are the load disturbance and the VSG output electric power variation, respectively, and the relationship between them is as follows: （2）； The expression for a grid-type energy storage model is as follows: （3）； wherein, is a virtual inertia coefficient, is a VSG power reference change, is a virtual damping coefficient, is a droop coefficient, is a coordinated power; Equivalent total inertia and equivalent total damping is given by （4）； A virtual impedance is introduced to correct the voltage reference value. The correction process is as follows: （5）； wherein, is a reference virtual impedance, S is a disturbance intensity index, is a frequency variation amount, is an ideal no-load voltage vector, i is a converter output current vector, is a virtual impedance-corrected voltage reference.

3. The control method according to claim 2, characterized by, The estimation algorithm is the RLS-BP algorithm; The RLS algorithm is used to determine the initial value of the virtual inertia coefficient, and the BP neural network model is used to determine the change in the virtual inertia coefficient. The expression for the virtual inertia coefficient is: （6）； in, The initial value for the virtual inertia coefficient. This represents the change in the virtual inertia coefficient. The RLS algorithm determines the initial value of the virtual inertia coefficient as follows: （7）； in, k For the number of recursions, For RLS parameter vectors, K This is the RLS gain vector. e For RLS prediction error, For frequency deviation, For RLS regression vectors, P Let covariance matrix be the variance matrix. Forgetting factor; A backpropagation (BP) neural network consists of an input layer, a hidden layer, and an output layer. The inputs to the input layer are frequency deviation, frequency deviation rate of change, total active power deviation on the grid side, and absolute value of RLS prediction error; The output of the output layer is the change in the virtual inertia coefficient.

4. The control method according to claim 3, characterized in that, load disturbance The maximum value is denoted as It needs to be corrected. The correction process is as follows: （8）； in, This is the corrected estimate of the load disturbance. This represents locally measured load mutations. The confidence weight is expressed as follows: （9）； in, The distance is Euclidean.

5. The control method according to claim 4, characterized in that, The detailed process of step S3 is as follows: First, determine the active power prediction value of the synchronous generator in the next second. Its expression is: （10）； in, This represents the real-time power of the synchronous generator at the moment before the disturbance. VSG instantaneous available capacity; Frequency and active power prediction values ​​are monitored, and when they reach the triggering conditions, a stability boundary assessment is performed to provide early warning. The assessment process is as follows: The Lyapunov energy function is constructed as follows: （11）； in, V For energy, It is the difference between the change in electromagnetic power and the change in mechanical power. The system synchronization power factor; Stability threshold Set to 0.05, when there is When the system tends towards instability, the required power injection value is calculated. To prevent system crashes due to frequency instability, its mathematical expression is: （12）； in, This is the emergency power gain coefficient. As the upper limit of energy safety, Let be the energy gradient norm, expressed as: （13）。 6. The control method according to claim 5, characterized in that, The triggering condition is: （14）； （15）； The triggering condition is met when one of the conditions in equations (14) and (15) is satisfied.

7. The control method according to claim 5, characterized in that, The process of step S4 is as follows: The expression for coordinated power is: （16）； in, The planned output setting value for the synchronous generator; Determine the available capacity of each VSG device, the first i The available capacity of the equipment is denoted as Then the first i Weighting of devices The expression is: （17）； in, n This represents the total number of VSG devices. No. i The power handled by this device is: （18）； in, For the first i The power handled by the equipment.

8. The control method according to claim 7, characterized in that, The normal constraint range for the average active power at the output of a synchronous generator is: （19）； in, The average active power at the output of the synchronous generator. This refers to the rated power of the synchronous generator.

9. The control method according to claim 8, characterized in that, When the average active power output of the synchronous generator exceeds the normal range, the planned output setting of the synchronous generator is adjusted. The adjustment process is as follows: when When the conditions are within the normal range of constraints, no action is taken; when At that time, The value increases by 0.02Pr; when When this happens, it indicates that the synchronous generator is outputting too much power; at this time, [the generator should be switched off]. The value decreased by 0.02Pr.

10. The control method according to claim 9, characterized in that, The range of the adjustment period is 20S~60S.

Images