Response driving frequency stability control method based on SFR model identification and control sensitive point mining

Abstract

The invention discloses a response driving frequency stability control method based on SFR model identification and control sensitive point mining, and relates to the technical field of power system frequency control. The method comprises the following steps: firstly, determining a frequency response model parameter containing a stability control module, obtaining a power disturbance quantity through a frequency change rate, and iteratively calculating a minimum load shedding quantity; constructing a frequency control sensitive point mining model based on Transform, training by using an off-line data set, and inputting disturbed key bus frequency time sequence data to obtain a load shedding sensitivity sequence of each control position; and finally distributing the load shedding amount according to the sensitivity sequence and issuing a control instruction. The method can rapidly and accurately formulate load shedding measures, reduces the calculation amount, improves the frequency improvement amount, and effectively guarantees the safety and stability of the frequency of a complex power system.

Description

A response-driven frequency stability control method based on SFR model identification and control sensitivity point mining Technical Field

[0001] This invention belongs to the field of power system frequency control technology, and relates to a response-driven frequency stability control method based on SFR model identification and control sensitivity point mining. Background Technology

[0002] As the proportion of renewable energy output continues to increase, the complexity and uncertainty of power grid operation will continue to rise. The complexity and variability of operating modes, the uncertainty of disturbances and faults, the weak immunity of power electronic equipment, and the high degree of freedom in control make system-level stability control more difficult, posing a huge challenge to ensuring the safe and efficient operation of ultra-large AC / DC hybrid power grids.

[0003] Emergency load shedding is an important measure in a stable defense technology system. The purpose of emergency load shedding is to prevent system parameters from exceeding limits, maintain system safety and stability, prevent large-scale power outages by proactively disconnecting loads, prepare the system for recovery control, and maintain system integrity and security.

[0004] Application content

[0005] To address the aforementioned issues, this invention proposes a response-driven frequency stability control method based on SFR model identification and control sensitivity point mining. This method improves the accuracy of control strategies when power systems experience disturbances, reduces the computational burden of solving power system control strategies, increases the frequency boost after the implementation of power system control strategies, and enhances frequency security and stability under complex power systems.

[0006] The objective of this invention is mainly achieved through the following technical solution: This invention discloses a response-driven frequency stability control method based on SFR model identification and control sensitivity point mining, comprising the following steps:

[0007] The frequency stabilization measures total quantity generation module is used to read power system status information and frequency response data, identify the system frequency response model, and calculate the total quantity of stabilization control measures that meet the frequency stability requirements.

[0008] The frequency control sensitive point mining module is used to input the frequency time series data of the key bus frequency of the power system into the frequency stability control sensitive point mining network, and to output the sensitivity ranking of each control position as the output of the frequency stability control sensitive point mining network.

[0009] The frequency stabilization and control measures generation module is used to read the total amount of stabilization and control measures that meet the system frequency stability requirements calculated by the total amount of frequency stabilization and control measures generation module, as well as the sensitivity ranking of each control position output by the frequency control sensitive point mining module. When a disturbance occurs in the power system, corresponding emergency load shedding measures are formulated to restore the system frequency to a safe range.

[0010] Preferably, the power system status information and frequency response data are the system equivalent inertia, the primary frequency regulation aggregation parameters of each unit, and the system inertia center frequency, respectively. The system equivalent inertia is obtained through the aggregation calculation of the inertia of each operating component, such as the inertia time constant of a synchronous generator. If the component inertia cannot be obtained through model parameters, a data-driven inertia evaluation method is used to obtain the component inertia. For the primary frequency regulation aggregation parameters, a detailed frequency regulation model is established based on the actual operating generating units with primary frequency regulation functions. The aggregation parameters are determined by reducing the overall order of the detailed model.

[0011] Preferably, sensitivity is a parameter that measures the response of the power system to different load shedding measures, and is quantified according to the magnitude of the increase in the minimum point of the system inertia center frequency under the same load shedding amount and different load shedding positions.

[0012] Preferably, the system frequency stability requirement is that the system inertia center frequency is not lower than a certain value, which is set to meet the system frequency stability requirement if the system inertia center frequency is not lower than 49.2Hz.

[0013] Preferably, it includes:

[0014] Step 1: Determine the frequency response model parameters including the stability control module, including the system equivalent inertia and the primary frequency regulation parameters of each unit;

[0015] Step 2: Obtain the power disturbance of the system based on the frequency change rate. Use the system disturbance and load shedding as inputs to the frequency response model, and the lowest frequency point as the output of the frequency response model. Obtain the minimum load shedding amount that meets the system frequency stability requirements through iterative calculation.

[0016] Step 3: Select several bus nodes in the power system as observation nodes and several load shedding positions as load shedding sorting positions. Extract the bus frequency time series data before the power system is disturbed as network input and the load shedding sensitivity sorting of each control position as network output to construct a frequency control sensitive point mining dataset.

[0017] Step 4: Construct a frequency control sensitive point mining model based on Transformer, and train the network using the frequency control sensitive point mining dataset generated offline.

[0018] Step 5: Input the key bus frequency time series data after the disturbance occurs into the frequency control sensitive point mining model, and output the load shedding sensitivity ranking of each control position in the current scenario;

[0019] Step 6: Using the obtained load shedding amount that meets the system frequency stability requirements and the ranking of load shedding sensitivity at each control position under the current scenario, formulate and generate control measures, and send the output instructions to the power system for execution, so that the system can provide corresponding load shedding measures online for different operating conditions to keep the system stable.

[0020] Preferably, the equivalent inertia of the system in step one is obtained by aggregating the inertia of each operating generator in the coordinate system of the center of inertia, and the calculation formula is as follows:

[0021]

[0022] Where: M sys M is the equivalent inertia of the system. i Let S be the inertia of the i-th generator. N,i Let be the capacity of the i-th unit; if the component inertia cannot be obtained through model parameters, then a data-driven inertia evaluation method is used to obtain the component inertia.

[0023] Preferably, the formula for calculating the power disturbance in step two is:

[0024]

[0025] Where: ΔP e,o1 Let f be the system disturbance power, f be the system frequency, and t0 be the time when the disturbance occurs.

[0026] Preferably, the quantitative index of load shedding sensitivity in step three is the increase in the minimum point of the system's center frequency of inertia under the same load shedding amount and different load shedding positions. The sensitivity calculation formula is:

[0027]

[0028] In the formula: λ is the load shear sensitivity; f 1min f is the point where the system's center frequency of inertia is lowest before load shedding. 2min ΔP represents the lowest point of the system's center frequency of inertia after load shedding; ΔP represents the amount of load shedding; Δf represents the increase in the lowest point of the system's center frequency of inertia before and after load shedding.

[0029] Beneficial effects

[0030] This invention uses a frequency response model and Transformer as the decision-making body, and the actual power system as the environment. It identifies the frequency response model by extracting bus frequency data from the power system node model, and then calculates the load shedding amount that meets the system frequency stability requirements under the current operating conditions. At the same time, it uses a frequency control sensitive point mining model to output the ranking of sensitive points at each control position under the current operating conditions, and formulates corresponding frequency stability control measures. It can quickly and accurately provide corresponding load shedding measures when the power system is disturbed, so that the system frequency can be restored to a safe range. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in this invention or the prior art, the accompanying drawings involved in the embodiments or the prior art are briefly described below. Obviously, these drawings illustrate several embodiments of the present invention, and those skilled in the art can derive other possible drawings based on these drawings without creative effort. The purpose of the drawings is limited to illustrating specific embodiments and does not limit the scope of the present invention.

[0032] Figure 1 is a flowchart of the response-driven frequency stability control based on SFR model identification and control sensitivity point mining proposed in this invention.

[0033] Figure 2 is a 39-node system example used in the demonstration;

[0034] Figure 3 is a heatmap of the predicted and actual rankings on the test set.

[0035] Figure 4 shows the power system frequency recovery curve under the response-driven frequency stability control method based on SFR model identification and control sensitivity point mining proposed in this invention. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.

[0037] The response-driven frequency stability control method based on SFR model identification and control sensitivity point mining, as shown in Figures 1 to 4, includes the following steps:

[0038] The frequency stabilization measures total quantity generation module is used to read power system status information and frequency response data, identify the system frequency response model, and calculate the total quantity of stabilization control measures that meet the frequency stability requirements.

[0039] The frequency control sensitive point mining module is used to input the frequency time series data of the key bus frequency of the power system into the frequency stability control sensitive point mining network, and to output the sensitivity ranking of each control position as the output of the frequency stability control sensitive point mining network.

[0040] The frequency stabilization and control measures generation module is used to read the total amount of stabilization and control measures that meet the system frequency stability requirements calculated by the total amount of frequency stabilization and control measures generation module, as well as the sensitivity ranking of each control position output by the frequency control sensitive point mining module. When a disturbance occurs in the power system, corresponding emergency load shedding measures are formulated to restore the system frequency to a safe range.

[0041] Preferably, the power system status information and frequency response data are the system equivalent inertia, the primary frequency regulation aggregation parameters of each unit, and the system inertia center frequency, respectively. The system equivalent inertia is obtained through the aggregation calculation of the inertia of each operating component, such as the inertia time constant of a synchronous generator. If the component inertia cannot be obtained through model parameters, a data-driven inertia evaluation method is used to obtain the component inertia. For the primary frequency regulation aggregation parameters, a detailed frequency regulation model is established based on the actual operating generating units with primary frequency regulation functions. The aggregation parameters are determined by reducing the overall order of the detailed model.

[0042] Sensitivity is a parameter that measures the response of a power system to different load shedding measures. It is quantified by the magnitude of the increase in the minimum point of the system inertia center frequency under the same load shedding amount and different load shedding locations.

[0043] The system frequency stability requirement is that the system inertia center frequency is not lower than a certain value. It is set that the system frequency stability requirement is met if the system inertia center frequency is not lower than 49.2Hz.

[0044] 1. Determine the frequency response model parameters including the stability control module, including the system equivalent inertia and the primary frequency regulation parameters of each unit. The system equivalent inertia can be obtained by aggregating the inertia of each operating generator in the coordinate system of the center of inertia. The calculation formula is as follows:

[0045]

[0046] For the primary frequency regulation aggregation parameters, a detailed frequency regulation model is established based on the actual operating generator units that have primary frequency regulation functions. In a multi-machine power system, the primary frequency regulation process is jointly regulated by each synchronous generator and converter. Among them, the generators equipped with primary frequency regulation and the converters with control respond to frequency deviations by changing mechanical power and output power. Since the detailed frequency regulation model is relatively complex, from the perspective of input-output equivalence, the aggregation parameters are obtained by replacing the high-order model with a low-order transfer function.

[0047] 2. Obtain the system power disturbance based on the frequency change rate. Use the system disturbance and load shedding as inputs to the frequency response model, and the lowest frequency point as the output. Obtain the minimum load shedding amount that satisfies the system frequency stability requirements through iterative calculation. The formula for calculating the disturbance power is as follows:

[0048]

[0049] Where: ΔP e,o1 Let f be the system disturbance power, f be the system frequency, and t0 be the time when the disturbance occurs.

[0050] 3. Select several bus nodes in the power system as observation nodes and several load shedding locations as load shedding sorting locations. Extract the bus frequency time series data before the power system is disturbed as the network input, and the load shedding sensitivity ranking at each control location as the network output, to construct a frequency control sensitive point mining dataset. The load shedding sensitivity index is as follows:

[0051]

[0052] In the formula: λ is the load shear sensitivity; f 1min f is the point where the system's center frequency of inertia is lowest before load shedding. 2min ΔP represents the lowest point of the system's center frequency of inertia after load shedding; ΔP represents the amount of load shedding; Δf represents the increase in the lowest point of the system's center frequency of inertia before and after load shedding.

[0053] 4. Construct a frequency control sensitivity point mining model based on Transformer, utilizing offline generation.

[0054] The network was trained using a frequency-controlled sensitive point mining dataset. Key parameters in the model included training batch size, learning rate, optimizer, number of encoder layers, input feature dimension, number of attention heads, and dimension of the hidden layers in the feedforward neural network.

[0055] 5. Read the frequency deviation time series data of the key bus within 1 second after the disturbance occurs, input it into the frequency control sensitive point mining model, and output the sensitivity ranking of each control position.

[0056] 6. Using the obtained load shedding amount that meets the system frequency stability requirements and the load shedding sensitivity ranking of each control position in the current scenario, control measures are formulated and generated. The total load shedding amount is carried out in order of control position sensitivity ranking, with priority given to cutting load nodes with high sensitivity, until the total load shedding amount is not less than the load shedding amount that meets the system frequency stability requirements.

[0057] 7. The generated frequency stabilization control measures are distributed to the power system for execution, enabling the system to provide corresponding load shedding measures online for different operating conditions to maintain system stability.

[0058] Example Demonstration

[0059] A 39-node power system was constructed as shown in Figure 2, with the following settings:

[0060] 1. Optimizer: Adam optimizer;

[0061] 2: Number of encoder layers: 6;

[0062] 3: Input feature dimension size: 100;

[0063] 4: Fault occurrence time: 1 second;

[0064] 5: Number of heads to focus on: 5;

[0065] 6: Hidden layer dimension of feedforward neural network: 2048;

[0066] 7: Buses to be observed: Bus01, Bus03, Bus05, Bus07, Bus09, Bus11, Bus13, Bus15, Bus17, Bus19, Bus21, Bus23, Bus25, Bus27, Bus29, Bus31, Bus33, Bus35, Bus37, Bus39;

[0067] 8. Control positions: Load03, Load04, Load07, Load08, Load15, Load16, Load18, Load20, Load21, Load23, Load25, Load26, Load27, Load29, Load39;

[0068] 9. Deep learning parameters: learning rate 0.001, batch size 64, number of iterations 1000

[0069] Figure 3 shows a heatmap of the predicted sensitivity ranking and the actual sensitivity ranking of the frequency control sensitive point mining model test set. The frequency control sensitive point mining model can achieve a sensitivity ranking accuracy of 93.5%. Figure 4 shows the generation and implementation of a response-driven frequency stability control strategy based on SFR model identification and control sensitive point mining, demonstrating the effectiveness of the frequency stability control method proposed in this invention for power system frequency stability restoration.

[0070] Those skilled in the art should understand that the above embodiments are merely illustrative of the content of this disclosure and do not limit its scope. The system capacity, voltage, line parameters, etc., shown may vary depending on the specific circumstances of the power electronic grid-connected generator set and its grid connection. Based on this disclosure, those skilled in the art can make other changes or adjustments, and these changes still fall within the scope of this disclosure.

Claims

1. A response-driven frequency stability control method based on SFR model identification and control sensitivity point mining, characterized in that, Includes the following steps: The frequency stabilization and control measures total quantity generation module is used to read power system status information and frequency response data, identify the system frequency response model, and calculate the total quantity of stabilization control measures that meet frequency stability requirements. The frequency control sensitive point mining module is used to input the frequency time series data of key power system bus frequencies into the frequency stabilization and control sensitive point mining network, and output the sensitivity ranking of each control position as the output of the frequency stabilization and control sensitive point mining network. The frequency stabilization and control measures generation module is used to read the total quantity of stabilization control measures that meet system frequency stability requirements calculated by the frequency stabilization and control measures total quantity generation module and the sensitivity ranking of each control position output by the frequency control sensitive point mining module, and formulate corresponding emergency load shedding measures when the power system experiences disturbances to restore the system frequency to a safe range.

2. The response-driven frequency stabilization control method based on SFR model identification and control sensitivity point mining according to claim 1, characterized in that: The power system status information and frequency response data are the system equivalent inertia, the primary frequency regulation aggregation parameters of each unit, and the system inertia center frequency, respectively. The system equivalent inertia is obtained by aggregating the inertia of each operating component, such as the inertia time constant of a synchronous generator. If the component inertia cannot be obtained through model parameters, a data-driven inertia evaluation method is used to obtain the component inertia. For the primary frequency regulation aggregation parameters, a detailed frequency regulation model is established based on the actual operating generating units with primary frequency regulation functions. The aggregation parameters are determined by reducing the overall equivalent order of the detailed model.

3. The response-driven frequency stabilization control method based on SFR model identification and control sensitivity point mining according to claim 2, characterized in that: Sensitivity is a parameter that measures the response of a power system to different load shedding measures. It is quantified by the magnitude of the increase in the minimum point of the system inertia center frequency under the same load shedding amount and different load shedding locations.

4. The response-driven frequency stability control method based on SFR model identification and control sensitivity point mining according to claim 3, characterized in that: The system frequency stability requirement is that the system inertia center frequency is not lower than a certain value. It is set that the system frequency stability requirement is met if the system inertia center frequency is not lower than 49.2Hz.

5. The response-driven frequency stability control method based on SFR model identification and control sensitivity point mining according to claim 1, characterized in that, include: Step 1: Determine the parameters of the frequency response model containing the stability control module, including the system equivalent inertia and the primary frequency regulation parameters of each unit. Step 2: Obtain the system power disturbance based on the frequency change rate. Use the system disturbance and load shedding as inputs to the frequency response model, and the lowest frequency point as the output. Iteratively calculate the minimum load shedding amount that meets the system frequency stability requirements. Step 3: Select several bus nodes in the power system as observation nodes and several load shedding locations as load shedding sorting locations. Extract the bus frequency time-series data before the disturbance as network input, and sort the load shedding sensitivity at each control location as network output. This data is used to construct a frequency control sensitive point mining mechanism. Step 4: Construct a frequency control sensitivity mining model based on Transformer, and train the network using the offline-generated frequency control sensitivity mining dataset; Step 5: Input the key bus frequency time series data after the disturbance occurs into the frequency control sensitivity mining model, and output the load shedding sensitivity ranking of each control position under the current scenario; Step 6: Use the obtained load shedding amount that meets the system frequency stability requirements and the load shedding sensitivity ranking of each control position under the current scenario to formulate and generate control measures, and issue the output instructions to the power system for execution, so that the system can provide corresponding load shedding measures online for different operating conditions to keep the system stable.

6. The response-driven frequency stability control method based on SFR model identification and control sensitivity point mining according to claim 5, characterized in that, In step one, the system's equivalent inertia is calculated by aggregating the inertia of each operating generator in the coordinate system of the center of inertia. The calculation formula is as follows: Where: M sys M is the equivalent inertia of the system. i Let S be the inertia of the i-th generator. N,i Let be the capacity of the i-th unit; if the component inertia cannot be obtained through model parameters, then a data-driven inertia evaluation method is used to obtain the component inertia.

7. The response-driven frequency stability control method based on SFR model identification and control sensitivity point mining according to claim 5, characterized in that, The formula for calculating the power disturbance in step two is: Where: ΔP e,o1 Let f be the system disturbance power, f be the system frequency, and t0 be the time when the disturbance occurs.

8. The response-driven frequency stability control method based on SFR model identification and control sensitivity point mining according to claim 5, characterized in that, In step three, the quantitative index of load shedding sensitivity is the increase in the minimum point of the system's center frequency of inertia under the same load shedding amount and different load shedding positions. The sensitivity calculation formula is: In the formula: λ is the load shear sensitivity; f 1min f is the point where the system's center frequency of inertia is lowest before load shedding. 2min ΔP represents the lowest point of the system's center frequency of inertia after load shedding; ΔP represents the amount of load shedding; Δf represents the increase in the lowest point of the system's center frequency of inertia before and after load shedding.

Images