Integrated causal data driven energy storage power centralized control method and system

Through the centralized control method of energy storage power driven by comprehensive causal data, the neural network is used to correct the SFR model and adjust the control coefficient, which solves the problem that traditional control methods are difficult to adapt to the frequency changes of the power grid, and reduces the output power and absorbs power during the frequency recovery period. , ensure the safety of the power grid frequency.

CN119029920APending Publication Date: 2024-11-26STATE GRID ELECTRIC POWER RES INST +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410944321.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-15
Publication Date
2024-11-26

AI Technical Summary

Technical Problem

The traditional frequency correction control method is difficult to adapt to the frequency drop and recovery of the power grid, and cannot operate effectively, resulting in further decrease of the power grid frequency. The fixed control strategy cannot find a compromise solution, and cannot reduce the output power, absorb the power during the frequency recovery period, and ensure the power grid. Frequency safe.

Method used

The centralized control method of energy storage power driven by comprehensive causal data is adopted. By obtaining the power grid operation mode data, forming simulation data, fitting the SFR model, modifying parameters using the neural network prediction model, adding energy storage sag control links, adjusting the control coefficients, according to The power adjustment is made to ensure the stability of the grid frequency.

Benefits of technology

It realizes the reduction of output power during frequency recovery, absorbs power when frequency overshoots, ensures the safety of the grid frequency, improves the adaptability and accuracy of the control strategy, and adapts to short-term power impact scenarios in a multi-grid environment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119029920A_ABST
    Figure CN119029920A_ABST
Patent Text Reader

Abstract

The invention discloses a comprehensive causal data driven energy storage power centralized control method and system, and relates to the technical field of power system automation, and the method comprises the steps: obtaining power grid mode data through an online device; performing fault simulation of power shortage to obtain a power grid frequency curve and performing parameter fitting of the SFR model; constructing a neural network prediction model in an off-line manner, predicting the parameter correction of the SFR model according to the response condition of the power grid, and correcting the SFR model; adding an energy storage droop control link, and adjusting a droop control coefficient to enable the power grid frequency stability margin to be maximum; and issuing the energy storage control coefficient to each energy storage unit for execution, and performing power adjustment according to the power grid frequency drop condition. The method can set the power regulation capability of energy storage in real time according to the current power grid state, improves the adaptability of power grid frequency control in a novel power system environment, and guarantees the safety and stability of a power grid.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of power system automation, and in particular to a comprehensive causal data-driven energy storage power centralized control method and system. Background Art

[0002] As the proportion of new energy continues to increase, the insufficient redundancy of the power grid causes the frequency to slowly decrease, and the high-power short-term power shock causes the frequency to drop and recover quickly. The traditional fixed strategy frequency correction control method is difficult to adapt.

[0003] Conventional power sources in the power grid are gradually being replaced, frequency regulation resources are scattered and insufficient, and the volatility of new energy sources causes the frequency to slowly decrease, significantly deteriorating the overall frequency characteristics. At present, the round setting of frequency correction control cannot be effectively operated, and can only dispatch and pull the power. It is urgent to put flexible control resources into action in advance to prevent further frequency decline in the scenario of power supply and demand imbalance. Due to the continuous increase in the scale of power electronic components, their switching control characteristics are prone to single failures with a high probability in the power grid, which may cause short-term shocks such as low penetration of new energy and DC commutation failure. After the power grid fails and the fault disappears, the high power change impacts the power grid frequency twice, and the fixed frequency correction control cannot find a compromise control solution. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: how to set a flexible control strategy to reduce the output power during frequency recovery and absorb power when the frequency overshoots, so as to ensure the safety of the power grid frequency.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides a comprehensive causal data driven energy storage power centralized control method, comprising:

[0008] Acquire grid operation mode data to form grid operation mode simulation data, perform power shortage fault simulation based on the simulation data, obtain grid frequency curve, and perform parameter fitting to obtain an SFR model;

[0009] The historical data of the power grid, the error between the initial time frequency after the fault and the predicted value of the SFR model are used as input, and the correction amount of the SFR model parameter is used as the output of the neural network prediction model;

[0010] When the real-time grid frequency is less than a set value, the input information of the neural network prediction model is collected to obtain the correction amount of the SFR model parameter, and the current SFR model is corrected;

[0011] Based on the modified SFR model, an energy storage droop control link is added, and the droop control coefficient is calculated and adjusted to maximize the grid frequency stability margin. The control coefficient is sent to each energy storage unit for execution, and power is adjusted according to the grid frequency drop.

[0012] As a preferred solution of the comprehensive causal data driven energy storage power centralized control method, wherein:

[0013] The acquiring of grid operation mode data, forming grid operation mode simulation data, performing power shortage fault simulation based on the simulation data, obtaining a grid frequency curve, and performing parameter fitting to obtain an SFR model comprises:

[0014] Generate online simulation calculation data based on the state estimation data issued by the power grid dispatching system;

[0015] Performing power shortage simulation on the data of the method, according to the scale of the power grid, using the method of DC bipolar blocking fault and power plant unit shutdown to simulate the power shortage of the power grid, and obtaining the frequency simulation curve of the power grid under the simulated power shortage of the power grid;

[0016] The least square method is used to fit the power grid SFR model, and the SFR model under the current power shortage is obtained, which is expressed as:

[0017]

[0018] Where Δf is the grid frequency deviation, ΔP d is the power shortage of the power grid, A0, A1, A2, A3, B0, B1 are the coefficients to be fitted of the SFR model, and s is the Laplace operator.

[0019] As a preferred solution of the comprehensive causal data driven energy storage power centralized control method, wherein:

[0020] The method of taking the error between the historical data of the power grid, the initial time frequency after the fault and the predicted value of the SFR model as input and the correction amount of the SFR model parameter as the output of the neural network prediction model includes:

[0021] A neural network prediction model is constructed with the grid equivalent inertia, generator damping coefficient, the error between the measured frequency 0.5s after the fault and the SFR model prediction value in historical data as input and the SFR model parameter correction as output.

[0022] As a preferred solution of the comprehensive causal data driven energy storage power centralized control method, wherein:

[0023] The method of taking the error between the historical data of the power grid, the initial time frequency after the fault and the predicted value of the SFR model as input and the correction amount of the SFR model parameter as the output of the neural network prediction model also includes:

[0024] The initial training parameters of the neural network include the number of hidden layers of the neural network, the number of neurons in the hidden layer, the number of training times, the learning rate, the weights and thresholds of each layer, and the tanh function is used as the activation function. The error gradient descent method is used to correct the output layer weights and thresholds, the weights and thresholds of each hidden layer in turn, and then compared with the simulation results to obtain the standard deviation of the error between the model prediction results and the simulation results.

[0025] As a preferred solution of the comprehensive causal data driven energy storage power centralized control method, wherein:

[0026] When the real-time grid frequency is less than a fixed value, collecting the input information of the neural network prediction model, obtaining the correction amount of the SFR model parameter, and correcting the current SFR model includes:

[0027] According to the grid equivalent inertia of the current grid, the generator damping coefficient, and the error between the measured frequency 0.5s after the actual frequency drop and the SFR model prediction value, the neural network prediction model is used to obtain the parameter correction amount of the current grid SFR model, modify the SFR model parameters, and correct the current SFR model.

[0028] As a preferred solution of the comprehensive causal data driven energy storage power centralized control method, wherein:

[0029] The energy storage droop control link added based on the modified SFR model includes:

[0030] Set the frequency offset threshold f cr The duration of the frequency anomaly that deviates from this given value is t cr The binary table [f cr ,t cr ], the gradient method is used to adjust the energy storage droop control coefficient K droop , making the grid frequency safety and stability margin the highest;

[0031] K droop It is sent to each energy storage unit for execution, and power is adjusted according to the grid frequency drop.

[0032] The power adjustment according to the grid frequency drop includes:

[0033] The specific power adjustment calculation formula is:

[0034] ΔP es =K droop ×(f0-f real)

[0035] Among them, ΔP es is the energy storage power increase amount, if it is a negative value, it is a decrease amount, f0 is the frequency rating, f real It is the measured value of the frequency at the grid connection point of the energy storage unit.

[0036] In a second aspect, an embodiment of the present invention provides a comprehensive causal data driven energy storage power centralized control system, including:

[0037] An SFR model building module is used to obtain power grid operation mode data, form power grid operation mode simulation data, perform power shortage fault simulation based on the simulation data, obtain a power grid frequency curve, and perform parameter fitting to obtain an SFR model;

[0038] A neural network prediction model establishment module is used to take the error between the power grid historical data, the initial time frequency after the fault and the predicted value of the SFR model as input, and the correction amount of the SFR model parameter as the output of the neural network prediction model;

[0039] A correction module, used for collecting the input information of the neural network prediction model when the real-time power grid frequency is less than a set value, obtaining the correction amount of the SFR model parameter, and correcting the current SFR model;

[0040] The power adjustment module is used to add a storage droop control link, calculate and adjust the droop control coefficient to maximize the grid frequency stability margin, send the control coefficient to each storage unit for execution, and adjust the power according to the grid frequency drop.

[0041] As a preferred solution for the energy storage power centralized control system driven by comprehensive causal data, in which:

[0042] The SFR model building module generates online simulation calculation mode data according to the state estimation data issued by the power grid dispatching system;

[0043] Performing power shortage simulation on the data of the method, according to the scale of the power grid, using the method of DC bipolar blocking fault and power plant unit shutdown to simulate the power shortage of the power grid, and obtaining the frequency simulation curve of the power grid under the simulated power shortage of the power grid;

[0044] The least square method is used to fit the power grid SFR model, and the SFR model under the current power shortage is obtained, which is expressed as:

[0045]

[0046] Where Δf is the grid frequency deviation, ΔP dis the power shortage of the power grid, A0, A1, A2, A3, B0, B1 are the coefficients to be fitted of the SFR model, and s is the Laplace operator;

[0047] The neural network prediction model building module uses the grid equivalent inertia, generator damping coefficient, and the error between the measured frequency 0.5s after the fault and the SFR model prediction value in the historical data as input, and the SFR model parameter correction value as output to build a neural network prediction model;

[0048] The initial training parameters of the neural network include the number of hidden layers of the neural network, the number of neurons in the hidden layer, the number of training times, the learning rate, the weights and thresholds of each layer, and the activation function uses the tanh function; the error gradient descent method is used to correct the weights and thresholds of the output layer, the weights and thresholds of each hidden layer in turn, and compared with the simulation results, to obtain the standard deviation of the error between the model prediction results and the simulation results;

[0049] The correction module uses the neural network prediction model to obtain the parameter correction amount of the current power grid SFR model according to the power grid equivalent inertia, the generator damping coefficient of the current power grid, and the error between the measured frequency 0.5s after the actual frequency drops and the SFR model prediction value, and modifies the SFR model parameters to correct the current SFR model;

[0050] The power adjustment module sets the frequency offset threshold value f cr The duration of the frequency anomaly that deviates from this given value is t cr The binary table [f cr ,t cr ], the gradient method is used to adjust the energy storage droop control coefficient K droop , making the grid frequency safety and stability margin the highest;

[0051] K droop It is sent to each energy storage unit for execution, and the power is adjusted according to the frequency drop of the power grid;

[0052] The power adjustment according to the grid frequency drop includes:

[0053] The specific power adjustment calculation formula is:

[0054] ΔP es =K droop ×(f0-f real )

[0055] Among them, ΔP es is the energy storage power increase amount, if it is a negative value, it is a decrease amount, f0 is the frequency rating, f real It is the measured value of the frequency at the grid connection point of the energy storage unit.

[0056] In a third aspect, an embodiment of the present invention provides a computing device, including:

[0057] Memory and processor;

[0058] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the one or more programs are executed by the one or more processors, the one or more processors implement the comprehensive causal data-driven energy storage power centralized control method as described in any embodiment of the present invention.

[0059] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the comprehensive causal data-driven energy storage power centralized control method.

[0060] Beneficial effects of the present invention: Currently, frequency control all adopts a pre-established control strategy, and discrete control is performed according to the frequency drop situation; the present invention adopts a method of formulating a control strategy in real time. When the system frequency drops, the measured data between the beginning of the drop and the strategy threshold value is used to predict the subsequent frequency curve, thereby more accurately ensuring the effectiveness of frequency control; traditional methods all adopt discrete control, but the present invention adopts a continuous control method for energy storage, which ensures the adaptability of the control strategy to variable power grid modes. In the case of short-term power impact, the output power can be reduced during frequency recovery, and power can be absorbed when the frequency overshoots to ensure the safety of the power grid frequency. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0062] Figure 1 It is an overall flow chart of the comprehensive causal data driven energy storage power centralized control method described in the first embodiment of the present invention;

[0063] Figure 2 It is a fitting result and an actual curve diagram in a simulation example of the comprehensive causal data-driven energy storage power centralized control method described in the second embodiment of the present invention. DETAILED DESCRIPTION

[0064] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0065] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0066] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0067] Example 1

[0068] Reference Figure 1 , which is the first embodiment of the present invention, and provides a comprehensive causal data driven energy storage power centralized control method, comprising:

[0069] S1: Acquire grid operation mode data to form grid operation mode simulation data, perform power shortage fault simulation based on the simulation data, obtain grid frequency curve, and perform parameter fitting to obtain SFR model;

[0070] In the embodiment of the present application, obtaining grid operation mode data, forming grid operation mode simulation data, performing power shortage fault simulation based on the simulation data, obtaining a grid frequency curve, and performing parameter fitting to obtain an SFR model include:

[0071] Generate online simulation calculation data based on the state estimation data issued by the power grid dispatching system;

[0072] Performing power shortage simulation on the data of the method, according to the scale of the power grid, using the method of DC bipolar blocking fault and power plant unit shutdown to simulate the power shortage of the power grid, and obtaining the frequency simulation curve of the power grid under the simulated power shortage of the power grid;

[0073] The least square method is used to fit the power grid SFR model, and the SFR model under the current power shortage is obtained, which is expressed as:

[0074]

[0075] Where Δf is the grid frequency deviation, ΔP d is the power shortage of the power grid, A0, A1, A2, A3, B0, B1 are the coefficients to be fitted of the SFR model, and s is the Laplace operator.

[0076] It should be noted that the SFR model is a system frequency response model. In the power system, frequency response refers to the ability of the system to automatically adjust the power generation to maintain the stability of the system frequency when facing load changes or generator failures. As a synthetic frequency response model, the SFR model is used to describe the frequency response characteristics of the system under different operating conditions. It can help predict how the grid frequency changes over time and adjust the power output of the energy storage system based on the prediction results to ensure that the grid frequency operates within a safe range.

[0077] S2: taking the error between the historical data of the power grid, the initial time frequency after the fault and the predicted value of the SFR model as input, and the correction amount of the SFR model parameter as the output of the neural network prediction model;

[0078] In an embodiment of the present application, offline construction of a neural network prediction model includes:

[0079] A neural network prediction model is constructed with the grid equivalent inertia, generator damping coefficient, the error between the measured frequency 0.5s after the fault and the SFR model prediction value in historical data as input and the SFR model parameter correction as output.

[0080] The initial training parameters of the neural network include the number of hidden layers of the neural network, the number of neurons in the hidden layer, the number of training times, the learning rate, the weights and thresholds of each layer, and the tanh function is used as the activation function. The error gradient descent method is used to correct the output layer weights and thresholds, the weights and thresholds of each hidden layer in turn, and then compared with the simulation results to obtain the standard deviation of the error between the model prediction results and the simulation results.

[0081] S3: When the real-time grid frequency is less than a set value, the input information of the neural network prediction model is collected to obtain the correction amount of the SFR model parameter, and the current SFR model is corrected;

[0082] In the embodiment of the present application, when the real-time grid frequency is less than a set value, the neural network prediction model input information is collected to obtain the correction amount of the SFR model parameter, and the current SFR model is corrected, including:

[0083] According to the grid equivalent inertia of the current grid, the generator damping coefficient, and the error between the measured frequency 0.5s after the actual frequency drop and the SFR model prediction value, the neural network prediction model is used to obtain the parameter correction amount of the current grid SFR model, modify the SFR model parameters, and correct the current SFR model.

[0084] It should be noted that the SFR model parameters under different power shortages are different, so the model parameters need to be corrected according to the real-time operation of the power grid. However, since parameter fitting requires a complete frequency curve to be completed, and the actual project cannot wait until the frequency response is complete before taking measures, an artificial intelligence method is used to construct a neural network prediction model with the error between the equivalent inertia of the power grid, the damping coefficient of the generator, and the measured frequency 0.5s after the fault in the historical data and the SFR model prediction value as input, and the SFR model parameter correction as output.

[0085] S4: Based on the modified SFR model, add the energy storage droop control link, calculate and adjust the droop control coefficient to maximize the grid frequency stability margin, send the control coefficient to each energy storage unit for execution, and adjust the power according to the grid frequency drop.

[0086] In the embodiment of the present application, based on the modified SFR model, adding the energy storage droop control link includes:

[0087] Set the frequency offset threshold f cr The duration of the frequency anomaly that deviates from this given value is t cr The binary table [f cr ,t cr ], the gradient method is used to adjust the energy storage droop control coefficient K droop , making the grid frequency safety and stability margin the highest;

[0088] K droop It is sent to each energy storage unit for execution, and power is adjusted according to the grid frequency drop.

[0089] Power adjustment based on grid frequency drop includes:

[0090] The specific power adjustment calculation formula is:

[0091] ΔP es =K droop ×(f0-f real )

[0092] Among them, ΔP es is the energy storage power increase amount, if it is a negative value, it is a decrease amount, f0 is the frequency rating, f real It is the measured value of the frequency at the grid connection point of the energy storage unit.

[0093] Specifically, when the number of energy storage units is determined, when K droop When K is larger, the overall control amount will be higher, which may cause frequency over-regulation, reducing the safety and stability margin of the power grid frequency. droopWhen it is small, the overall control amount is small, which will cause frequency under-regulation and low stability margin. Therefore, the gradient method is used to adjust K droop , so that the power grid frequency safety and stability margin is the highest. The calculation of frequency safety and stability margin is relatively mature and can be calculated with the help of FASTEST software. The frequency offset threshold value f needs to be set. cr The duration of the frequency anomaly that deviates from this given value is t cr The binary table [f cr ,t cr ], which can be set according to the target power grid requirements.

[0094] The above is a schematic scheme of the comprehensive causal data driven energy storage power centralized control method of this embodiment. It should be noted that the technical scheme of the comprehensive causal data driven energy storage power centralized control system and the technical scheme of the above comprehensive causal data driven energy storage power centralized control method belong to the same concept, and the details of the technical scheme of the comprehensive causal data driven energy storage power centralized control system in this embodiment that are not described in detail can all be referred to the description of the technical scheme of the above comprehensive causal data driven energy storage power centralized control method.

[0095] The energy storage power centralized control system driven by comprehensive causal data in this embodiment includes:

[0096] An SFR model building module is used to obtain power grid operation mode data, form power grid operation mode simulation data, perform power shortage fault simulation based on the simulation data, obtain a power grid frequency curve, and perform parameter fitting to obtain an SFR model;

[0097] A neural network prediction model establishment module is used to take the error between the power grid historical data, the initial time frequency after the fault and the predicted value of the SFR model as input, and the correction amount of the SFR model parameter as the output of the neural network prediction model;

[0098] A correction module, used for collecting the input information of the neural network prediction model when the real-time power grid frequency is less than a set value, obtaining the correction amount of the SFR model parameter, and correcting the current SFR model;

[0099] The power adjustment module is used to add a storage droop control link, calculate and adjust the droop control coefficient to maximize the grid frequency stability margin, send the control coefficient to each storage unit for execution, and adjust the power according to the grid frequency drop.

[0100] In the embodiment of the present application, the SFR model building module generates online simulation calculation mode data according to the state estimation data issued by the power grid dispatching system;

[0101] Performing power shortage simulation on the data of the method, according to the scale of the power grid, using the method of DC bipolar blocking fault and power plant unit shutdown to simulate the power shortage of the power grid, and obtaining the frequency simulation curve of the power grid under the simulated power shortage of the power grid;

[0102] The least square method is used to fit the power grid SFR model, and the SFR model under the current power shortage is obtained, which is expressed as:

[0103]

[0104] Where Δf is the grid frequency deviation, ΔP d is the power shortage of the power grid, A0, A1, A2, A3, B0, B1 are the coefficients to be fitted of the SFR model, and s is the Laplace operator;

[0105] The neural network prediction model building module uses the grid equivalent inertia, generator damping coefficient, and the error between the measured frequency 0.5s after the fault and the SFR model prediction value in the historical data as input, and the SFR model parameter correction value as output to build a neural network prediction model;

[0106] The initial training parameters of the neural network include the number of hidden layers of the neural network, the number of neurons in the hidden layer, the number of training times, the learning rate, the weights and thresholds of each layer, and the activation function uses the tanh function; the error gradient descent method is used to correct the weights and thresholds of the output layer, the weights and thresholds of each hidden layer in turn, and compared with the simulation results, to obtain the standard deviation of the error between the model prediction results and the simulation results;

[0107] The correction module uses the neural network prediction model to obtain the parameter correction amount of the current power grid SFR model according to the power grid equivalent inertia, the generator damping coefficient of the current power grid, and the error between the measured frequency 0.5s after the actual frequency drops and the SFR model prediction value, and modifies the SFR model parameters to correct the current SFR model;

[0108] The power adjustment module sets the frequency offset threshold value f cr The duration of the frequency anomaly that deviates from this given value is t cr The binary table [f cr ,t cr ], the gradient method is used to adjust the energy storage droop control coefficient K droop , making the grid frequency safety and stability margin the highest;

[0109] K droop It is sent to each energy storage unit for execution, and the power is adjusted according to the frequency drop of the power grid;

[0110] The power adjustment according to the grid frequency drop includes:

[0111] The specific power adjustment calculation formula is:

[0112] ΔP es =K droop ×(f0-f real )

[0113] Among them, ΔP es is the energy storage power increase amount, if it is a negative value, it is a decrease amount, f0 is the frequency rating, f real It is the measured value of the frequency at the grid connection point of the energy storage unit.

[0114] This embodiment further provides a computing device, which is applicable to the case of a centralized energy storage power control method driven by comprehensive causal data, including:

[0115] Memory and processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the comprehensive causal data driven energy storage power centralized control method proposed in the above embodiment.

[0116] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the comprehensive causal data-driven energy storage power centralized control method proposed in the above embodiment is implemented.

[0117] The storage medium proposed in this embodiment and the comprehensive causal data-driven energy storage power centralized control method proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0118] Example 2

[0119] Reference Figure 2 Tables 1 and 2 are an embodiment of the present invention, which provides a comprehensive causal data-driven energy storage power centralized control method. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0120] SFR model fitting based on simulation data:

[0121] The online device is used to obtain the grid operation mode data and form the grid operation mode simulation data; the simulation data is subjected to the power shortage fault simulation to obtain the grid frequency curve.

[0122] SFR model parameter fitting:

[0123] In this embodiment, the SFR model is fitted using the least squares method, and the following is an example of the fitting result:

[0124] The least square method is used to fit the power grid SFR model, and the SFR model under the current power shortage is obtained, which is expressed as:

[0125]

[0126] Where Δf is the grid frequency deviation, ΔP d is the power shortage of the power grid, A0, A1, A2, A3, B0, B1 are the coefficients to be fitted of the SFR model, and s is the Laplace operator.

[0127] The results are shown in Table 1 and Figure 2 As shown, Figure 2 The blue one is the actual curve and the gray one is the fitting result.

[0128] Table 1 Fitting results

[0129]

[0130] These parameters will be used to describe the grid frequency response characteristics under different power shortages.

[0131] Neural network prediction model construction:

[0132] The grid equivalent inertia, generator damping coefficient and the error between the measured frequency 0.5 seconds after the fault and the SFR model prediction value in the historical data are used as input. The output is the correction value of the SFR model parameters, which is used to adjust the SFR model to more accurately predict the grid frequency response.

[0133] Set the parameters of the neural network as follows:

[0134] Number of hidden layers: 2

[0135] Number of neurons in the hidden layer: 10

[0136] Number of trainings: 1000

[0137] Learning rate: 0.01

[0138] Activation function: tanh function

[0139] The neural network is trained by the error gradient descent method to reduce the prediction error and optimize the ability to correct the SFR model parameters.

[0140] Real-time frequency response control:

[0141] When the real-time grid frequency is lower than the set value, the input information of the neural network prediction model is collected. The neural network prediction model is used to calculate the parameter correction of the current grid SFR model, and the SFR model is adjusted in real time to cope with the change of grid frequency.

[0142] Energy storage droop control strategy:

[0143] Based on the modified SFR model, the droop control link is designed. The droop control coefficient is adjusted by the gradient method to maximize the grid frequency stability margin. The optimized control coefficient is sent to each energy storage unit for execution. The power output of each energy storage unit is adjusted according to the grid frequency drop to ensure that the grid frequency operates stably within a safe range.

[0144] Table 2 Simulation results

[0145]

[0146] The above table shows the simulation results based on different power shortages, including the grid frequency deviation curve, the SFR model parameters obtained by fitting, the correction amount predicted by the neural network, and the final control coefficient adjustment results. These data intuitively demonstrate the actual effect and application potential of the proposed method in dealing with grid frequency changes.

[0147] Through the detailed description and simulation results of this embodiment, the feasibility and superiority of the comprehensive causal data-driven energy storage power centralized control method in practical applications are verified, especially its effectiveness in improving grid frequency stability and responding to emergencies.

[0148] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A comprehensive causal data driven energy storage power centralized control method, characterized in that: include: Acquire grid operation mode data to form grid operation mode simulation data, perform power shortage fault simulation based on the simulation data, obtain grid frequency curve, and perform parameter fitting to obtain an SFR model; The historical data of the power grid, the error between the initial time frequency after the fault and the predicted value of the SFR model are used as input, and the correction amount of the SFR model parameter is used as the output of the neural network prediction model; When the real-time grid frequency is less than a set value, the input information of the neural network prediction model is collected to obtain the correction amount of the SFR model parameter, and the current SFR model is corrected; Based on the modified SFR model, an energy storage droop control link is added, and the droop control coefficient is calculated and adjusted to maximize the grid frequency stability margin. The control coefficient is sent to each energy storage unit for execution, and power is adjusted according to the grid frequency drop.

2. The comprehensive causal data driven energy storage power centralized control method according to claim 1, characterized in that: The acquiring of grid operation mode data, forming grid operation mode simulation data, performing power shortage fault simulation based on the simulation data, obtaining a grid frequency curve, and performing parameter fitting to obtain an SFR model comprises: Generate online simulation calculation data based on the state estimation data issued by the power grid dispatching system; Performing power shortage simulation on the data of the method, according to the scale of the power grid, using the method of DC bipolar blocking fault and power plant unit shutdown to simulate the power shortage of the power grid, and obtaining the frequency simulation curve of the power grid under the simulated power shortage of the power grid; The least square method is used to fit the power grid SFR model, and the SFR model under the current power shortage is obtained, which is expressed as: Where Δf is the grid frequency deviation, ΔP d is the power shortage of the power grid, A0, A1, A2, A3, B0, B1 are the coefficients to be fitted of the SFR model, and s is the Laplace operator.

3. The comprehensive causal data driven energy storage power centralized control method according to claim 2, characterized in that: The method of taking the error between the historical data of the power grid, the initial time frequency after the fault and the predicted value of the SFR model as input and the correction amount of the SFR model parameter as the output of the neural network prediction model includes: A neural network prediction model is constructed with the grid equivalent inertia, generator damping coefficient, the error between the measured frequency 0.5s after the fault and the SFR model prediction value in historical data as input and the SFR model parameter correction as output.

4. The comprehensive causal data driven energy storage power centralized control method according to claim 3, characterized in that: The method of taking the error between the historical data of the power grid, the initial time frequency after the fault and the predicted value of the SFR model as input and the correction amount of the SFR model parameter as the output of the neural network prediction model also includes: The initial training parameters of the neural network include the number of hidden layers of the neural network, the number of neurons in the hidden layer, the number of training times, the learning rate, the weights and thresholds of each layer, and the tanh function is used as the activation function. The error gradient descent method is used to correct the output layer weights and thresholds, the weights and thresholds of each hidden layer in turn, and then compared with the simulation results to obtain the standard deviation of the error between the model prediction results and the simulation results.

5. The comprehensive causal data driven energy storage power centralized control method according to claim 4, characterized in that: When the real-time grid frequency is less than a fixed value, collecting the input information of the neural network prediction model, obtaining the correction amount of the SFR model parameter, and correcting the current SFR model includes: According to the grid equivalent inertia of the current grid, the generator damping coefficient, and the error between the measured frequency 0.5s after the actual frequency drop and the SFR model prediction value, the neural network prediction model is used to obtain the parameter correction amount of the current grid SFR model, modify the SFR model parameters, and correct the current SFR model.

6. The comprehensive causal data driven energy storage power centralized control method according to claim 5, characterized in that: The energy storage droop control link added based on the modified SFR model includes: Set the frequency deviation threshold f cr The duration of the frequency anomaly that deviates from this given value is t cr The binary table [f cr ,t cr ], the gradient method is used to adjust the energy storage droop control coefficient K droop , making the grid frequency safety and stability margin the highest; K droop It is sent to each energy storage unit for execution, and the power is adjusted according to the frequency drop of the power grid; The power adjustment according to the grid frequency drop includes: The specific power adjustment calculation formula is: ΔP es =K droop ×(f0-f real ) Among them, ΔP es is the energy storage power increase amount, if it is a negative value, it is a decrease amount, f0 is the frequency rating, f real It is the measured value of the frequency at the grid connection point of the energy storage unit.

7. A comprehensive causal data driven energy storage power centralized control system, characterized in that: include: An SFR model building module is used to obtain power grid operation mode data, form power grid operation mode simulation data, perform power shortage fault simulation based on the simulation data, obtain a power grid frequency curve, and perform parameter fitting to obtain an SFR model; A neural network prediction model establishment module is used to take the error between the power grid historical data, the initial time frequency after the fault and the predicted value of the SFR model as input, and the correction amount of the SFR model parameter as the output of the neural network prediction model; A correction module, used for collecting the input information of the neural network prediction model when the real-time power grid frequency is less than a set value, obtaining the correction amount of the SFR model parameter, and correcting the current SFR model; The power adjustment module is used to add a storage droop control link, calculate and adjust the droop control coefficient to maximize the grid frequency stability margin, send the control coefficient to each storage unit for execution, and adjust the power according to the grid frequency drop.

8. The comprehensive causal data driven energy storage power centralized control system according to claim 7, characterized in that: include: The SFR model building module generates online simulation calculation mode data according to the state estimation data issued by the power grid dispatching system; Performing power shortage simulation on the data of the method, according to the scale of the power grid, using the method of DC bipolar blocking fault and power plant unit shutdown to simulate the power shortage of the power grid, and obtaining the frequency simulation curve of the power grid under the simulated power shortage of the power grid; The least square method is used to fit the power grid SFR model, and the SFR model under the current power shortage is obtained, which is expressed as: Where Δf is the grid frequency deviation, ΔP d is the power shortage of the power grid, A0, A1, A2, A3, B0, B1 are the coefficients to be fitted of the SFR model, and s is the Laplace operator; The neural network prediction model building module uses the grid equivalent inertia, generator damping coefficient, and the error between the measured frequency 0.5s after the fault and the SFR model prediction value in the historical data as input, and the SFR model parameter correction value as output to build a neural network prediction model; The initial training parameters of the neural network include the number of hidden layers of the neural network, the number of neurons in the hidden layer, the number of training times, the learning rate, the weights and thresholds of each layer, and the activation function uses the tanh function; the error gradient descent method is used to correct the weights and thresholds of the output layer, the weights and thresholds of each hidden layer in turn, and compared with the simulation results, to obtain the standard deviation of the error between the model prediction results and the simulation results; The correction module uses the neural network prediction model to obtain the parameter correction amount of the current power grid SFR model according to the power grid equivalent inertia, the generator damping coefficient of the current power grid, and the error between the measured frequency 0.5s after the actual frequency drops and the SFR model prediction value, and modifies the SFR model parameters to correct the current SFR model; The power adjustment module sets the frequency offset threshold value f cr The duration of the frequency anomaly that deviates from this given value is t cr The binary table [f cr ,t cr ], the gradient method is used to adjust the energy storage droop control coefficient K droop , making the grid frequency safety and stability margin the highest; K droop It is sent to each energy storage unit for execution, and the power is adjusted according to the frequency drop of the power grid; The power adjustment according to the grid frequency drop includes: The specific power adjustment calculation formula is: ΔP es =K droop ×(f0-f real ) Among them, ΔP es is the energy storage power increase amount, if it is a negative value, it is a decrease amount, f0 is the frequency rating, f real It is the measured value of the frequency at the grid connection point of the energy storage unit.

9. A computing device comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the comprehensive causal data-driven energy storage power centralized control method described in any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the comprehensive causal data-driven energy storage power centralized control method as described in any one of claims 1 to 6.