Method for identifying power grid frequency signal assessment standard

Abstract

The invention provides a method for identifying a power grid frequency signal assessment standard, and the method comprises the steps: constructing a system architecture composed of a master station, a substation and a communication network, and achieving the initialization of equipment through a finite-state machine; the method comprises the following steps: collecting power grid frequency, power generation side frequency difference and communication state multi-dimensional data, processing signals by applying median filtering, a sliding window and an EWMA algorithm, introducing fuzzy logic to evaluate a system state, and dynamically adjusting a response strategy; an improved PID algorithm is adopted to track frequency change, feed-forward compensation and PSO parameter optimization are combined, a hidden Markov model is utilized to predict a communication fault, and local mode switching is triggered in advance; main station, mixing and local three-state switching is carried out according to communication states and load changes, and a frequency modulation dead zone is automatically adjusted according to power grid loads through a dynamic threshold value self-adaptive algorithm; the control effect is verified by using a digital twinning technology; and the safety and reliability of the system are guaranteed by optional quantum encryption and active-active redundancy design. The whole scheme effectively reduces the power loss of power generation enterprises, and maintains the safe and stable operation of a power grid.

Description

Technical Field

[0001] The present invention relates to the field of power grid frequency signal assessment, and in particular to a method for identifying power grid frequency signal assessment standards. Background Art

[0002] With the reform of the power market and the addition of new energy projects, the primary frequency regulation assessment burden on power generation has increased. However, due to the different sampling locations of the frequency signals used by the power grid for assessment and the grid frequency signals collected by the power generation side, the assessment standards are inconsistent. This imposes severe assessment power requirements on power generation companies in terms of primary frequency regulation performance, AGC regulation performance, and power market transactions. This places a particularly heavy operational burden on low-profit thermal power companies. Domestic power grids assess power generation companies' primary frequency regulation performance using the grid frequency reference point of decentralized frequency transmitters within the regional grid. However, power generation companies can only collect frequency signals on the high-voltage side of the main transformer. This results in significant discrepancies between the assessed and actual operation times. Inconsistent operation signals can trigger AGC lockout and power assessment of units.

[0003] The existing detection method based on the frequency transmitter on the power generation side cannot be consistent with the real-time assessment point of the power grid. The deviation of the operation assessment is likely to lead to the power assessment, which is not conducive to the real-time response of the unit's frequency regulation performance, the operation of the power generation enterprise and the safe and stable operation of the power grid. Therefore, there is an urgent need for a solution that optimizes the impact of the frequency acquisition signal on the unit's AGC regulation performance and power trading while ensuring that the unit's primary frequency regulation function is not affected. Summary of the Invention

[0004] The main purpose of the present invention is to provide a method for identifying the grid frequency signal assessment standard, so as to solve the problem in the prior art that the detection method based on the frequency transmitter on the power generation side cannot be consistent with the grid assessment point in real time, and the operation assessment deviation is easy to lead to the power assessment, which is not conducive to the real-time response of the unit to the frequency regulation performance, the operation of the power generation enterprise and the safe and stable operation of the power grid.

[0005] To solve the above technical problems, the technical solution adopted by the present invention is: a method for identifying the assessment standard of the power grid frequency signal, the method comprising: S1, constructing a system architecture consisting of a master station, a substation and a communication network, and implementing device initialization through a finite state machine; S2, collects multi-dimensional data on grid frequency, power generation side frequency difference, and communication status, applies median filtering, sliding window, and EWMA algorithms to process the signals, introduces fuzzy logic to evaluate system status and dynamically adjust response strategies; S3: Use an improved PID algorithm to track frequency changes, combine feedforward compensation with PSO parameter optimization, and use a hidden Markov model to predict communication failures and trigger local mode switching in advance; S4: Switch between the master, hybrid, and local states according to the communication status and load changes. The dynamic threshold adaptive algorithm automatically adjusts the frequency regulation dead zone according to the grid load. S5. Use digital twin technology to verify the control effect; S6, optional quantum encryption and active-active redundancy design ensure system security and reliability.

[0006] In the preferred solution, in step S1, a system architecture consisting of a master station, a substation, and a communication network is constructed. The master station deploys a frequency acquisition server, an instruction generation unit, and a dispatching data network core switch. The substation includes an edge computing gateway, a hard-wired interface unit, a synchronous timing terminal, and an AC / DC redundant power supply module. The communication network adopts a dedicated dispatching data network with a dual-ring network structure and is configured with the OSPF dynamic routing protocol and the TCP transmission protocol. Use a finite state machine to initialize the substation equipment, define a state set including cold start state, hardware self-test state, safety authentication state, communication handshake state, standby operation state, and fault lock state, and calculate the probability of hardware self-test components passing using an exponential distribution failure rate model; The M / M / 1 model of queuing theory is used to calculate the time consumption of security authentication, the first-order autoregressive model AR(1) is used to predict the communication link delay, and fault detection and state switching are realized based on multi-dimensional logical expressions. Among them, the hardware self-test component is calculated through the probability formula: , the time-consuming formula for security authentication is , the communication delay prediction formula is ; The fault judgment formula is: , through the model and formula, the state verification and conversion of the substation equipment from cold start to standby operation are realized to ensure the normal operation of the system components.

[0007] In the preferred solution, in step S2, the grid frequency, the power generation side frequency difference, and the communication status data are collected, and the median filter algorithm is applied to denoise the grid frequency signal. The filtered signal is obtained by establishing a sliding window of length N and taking the median value of the sampled values ​​in the window. The formula is: ; And determine whether to trigger signal redundancy check by calculating the standard deviation of the three-way filtered signal; The frequency difference change rate on the power generation side is calculated using the sliding window algorithm, and the frequency difference is fitted with a quadratic polynomial within a sliding window of length M. , solve the coefficient to get the rate of change , and introduce the forgetting factor α to perform exponential weighted averaging to obtain ; The EWMA algorithm based on the Kalman filter framework is used to predict the communication state, and the state transition equation and the observation equation Perform recursive calculations to dynamically predict the round-trip delay and packet loss rate of TCP links; Fuzzy logic is introduced to evaluate the system status. By defining the fuzzy sets and membership functions of the power grid frequency difference, frequency difference change rate, and communication delay, a rule base is constructed and the Mamdani reasoning algorithm and center of gravity method are used. Perform reasoning and defuzzification, and use the formula according to the fuzzy output results Dynamically adjust local switching thresholds and instruction transmission strategies.

[0008] In the preferred solution, in step S3, an improved PID algorithm is used to track frequency changes, which introduces a feedforward compensation term To respond to frequency mutations in advance and combine with anti-saturation mechanism Suppress the integral accumulation when the deviation is large, and use the incomplete differential filtering formula Smoothed differential output; Use particle swarm optimization algorithm to adjust PID parameters online, and update the speed formula through And the position update formula Minimize the comprehensive cost function ; Dynamically adjust the primary frequency modulation dead zone threshold through the formula Update the adjustment coefficient by combining the least squares method ; Using the hidden Markov model to predict communication failures, the forward probability formula Calculate state probability based on risk index Trigger local mode smooth switching in advance, and use the weight function when switching Realize linear transition of control signal.

[0009] In the preferred solution, in step S4, the master, hybrid, and local states are switched according to the communication state and load changes, and a three-dimensional state space ( , , ) Determine the state transition condition, when 50ms≤ <100ms and 1%≤ <5% triggers the master mode to switch to hybrid mode. ≥100ms or ≥5% or >0.9 triggers the switch from hybrid mode to local mode; Using weighted fusion formula in hybrid mode Calculate the action signal, where the weight , ; The dynamic threshold adaptive algorithm is based on the formula Automatically adjust the frequency modulation dead zone, every 100ms according to the latest collected data through the formula Update threshold; Handle switching conflicts through priority decision rules when a communication interruption or substation hardware failure is detected or When the value is >0.95, local mode is triggered directly. In non-emergency switching, state transfer is performed through a 500ms buffer period. Using the formula Calculate the switching delay by the formula Assess load shock and use the exponential moving average formula Store and analyze historical switching data.

[0010] In the preferred solution, in step S5, the control effect is verified by using digital twin technology, a reinforcement learning model is constructed to continuously optimize the control strategy, and the control effect and system safety are verified based on the digital twin technology; The digital twin model includes a grid-side model based on the power system power flow equation and frequency response characteristics, a communication network model based on queuing theory and network delay distribution, and a replicated controller digital twin. A Bayesian optimization algorithm is used to minimize the deviation between the twin model and the physical system to calibrate the model parameters. The reinforcement learning model adopts Markov decision process modeling. The state space includes frequency, frequency deviation, communication delay, packet loss rate, load rate and operation mode. The action space includes PID parameters, frequency modulation dead zone threshold and mixed mode weight. The reward function comprehensively considers frequency tracking error, power fluctuation, communication risk and control smoothness. A proximal strategy optimization algorithm is used to train the control strategy, and an entropy regularization term is added to the objective function to enhance exploration. A control effect verification framework is constructed, including frequency quality indicators, assessment compliance rate, and switching performance indicators. A closed-loop optimization process is formed through offline training, online deployment and adaptive update mechanism; a risk assessment model is used to calculate the system operation risk, and a safety penalty term is added to the reinforcement learning objective function to ensure that the optimization process does not violate the safety boundary.

[0011] In a preferred embodiment, the grid-side model includes a set of differential algebraic equations: ; in, is the generator power angle, is the angular velocity, is the inertia time constant, is the mechanical power, is the electromagnetic power, is the damping coefficient; the queuing delay in the communication network model is expressed by the formula Calculate, where is the arrival rate, is the service rate; the Bayesian optimization algorithm is implemented by formula Calibrate model parameters; the reward function is expressed as Calculation: The objective function of the proximal strategy optimization algorithm is calculated by the formula ; Calculation: The root mean square error in the frequency quality index is calculated by the formula Calculation; the risk assessment model is calculated by the formula Calculate system operation risks.

[0012] In the preferred solution, in step S6, quantum encryption communication and active-active redundant architecture are used to ensure system security and reliability; the quantum encryption communication is implemented based on the decoy state BB84 protocol, and the decoy state strength formula is optimized. Improve the single photon detection rate and use the formula The final key rate is calculated to quantify the security, according to the formula Dynamically update keys; The active-active redundant architecture calculates synchronization delay through the active-standby synchronization protocol , using conflict resolution algorithms Determine switching priority based on dynamic load distribution formula Achieve traffic balance.

[0013] In the preferred solution, in step S6, the quantum encryption and active-active architecture synergy mechanism includes: Assign encryption paths to data flows, using the formula Select the optimal encrypted transmission path; the reliability of the active-active system is determined by the Markov model state transition rate matrix and system availability formula Evaluation, combined with the key cache synchronization formula Ensure key synchronization security through key consumption prediction formula and performance consistency evaluation formula Implement anomaly detection and security auditing.

[0014] The present invention provides a method for identifying the assessment standard of power grid frequency signals. Through the master station-substation architecture and multi-algorithm collaboration, the reliable transmission of the assessment frequency signal from the grid side to the power generation side is achieved, solving the assessment deviation problem caused by sampling position differences. By using an improved PID combined with PSO optimization, fuzzy logic dynamic adjustment threshold and three-state switching control, the frequency tracking accuracy and system response flexibility are improved, overshoot is suppressed and unnecessary frequency modulation actions are reduced. The introduction of quantum encryption and active-active redundant architecture ensures signal transmission security and high system availability, and reduces the risk of communication failure. The application of digital twins and reinforcement learning realizes the automated optimization and verification of control strategies, and improves the adaptability and safety of the system. The overall solution effectively reduces the power loss of power generation enterprises during assessment, maintains the safe and stable operation of the power grid, and is particularly suitable for precise frequency control in scenarios with high penetration of new energy, balancing assessment indicators and unit operation economy. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The present invention will be further described below with reference to the accompanying drawings and examples: Figure 1 This is a flow chart of the assessment standard for identifying power grid frequency signals according to the present invention. DETAILED DESCRIPTION

[0016] Example 1 like Figure 1 As shown, a method for identifying a grid frequency signal assessment standard includes: S1, constructing a system architecture consisting of a master station, a substation, and a communication network, and initializing equipment through a finite state machine; S2, collects multi-dimensional data on grid frequency, power generation side frequency difference, and communication status, applies median filtering, sliding window, and EWMA algorithms to process the signals, introduces fuzzy logic to evaluate system status and dynamically adjust response strategies; S3: Use an improved PID algorithm to track frequency changes, combine feedforward compensation with PSO parameter optimization, and use a hidden Markov model to predict communication failures and trigger local mode switching in advance; S4: Switch between the master, hybrid, and local states according to the communication status and load changes. The dynamic threshold adaptive algorithm automatically adjusts the frequency regulation dead zone according to the grid load. S5. Use digital twin technology to verify the control effect; S6, optional quantum encryption and active-active redundancy design ensure system security and reliability.

[0017] (1) Construct a system architecture consisting of a master station, a substation, and a communication network, and use a finite state machine to complete device initialization to ensure the normal operation of all system components; (2) Collect multi-dimensional data such as grid frequency, power generation side frequency difference, and communication status, use median filtering, sliding window, and EWMA algorithms to process the signals, evaluate the system status through fuzzy logic, and then dynamically adjust the response strategy; (3) Use an improved PID algorithm combined with feedforward compensation and PSO parameter optimization to achieve accurate tracking of frequency changes, and use the hidden Markov model to predict communication failures and trigger local mode switching in advance; (4) Switch between the master station, hybrid, and local modes based on communication status and load changes, and automatically adjust the frequency regulation dead zone according to the grid load through a dynamic threshold adaptive algorithm; (5) Use digital twin technology to build a virtual environment to verify the control effect; (6) Ensure the safety and reliability of the system through optional quantum encryption and active-active redundancy design.

[0018] This method solves the problem of frequency measurement deviation between the grid and the generator side caused by sampling location discrepancies, ensuring reliable transmission of the grid measurement frequency signal to the generator side. An improved PID algorithm, combined with multiple optimization strategies, enhances frequency tracking accuracy and system response speed, suppresses overshoot, and reduces ineffective frequency modulation. Three-state switching and dynamic threshold adjustment enable the system to better adapt to communication status and load changes. Digital twin technology ensures the effectiveness of the control strategy, while quantum encryption and active-active redundancy enhance system security and reliability. Ultimately, this reduces power loss during measurement for power generation companies and maintains stable grid operation. It is suitable for precise frequency control in scenarios with high penetration of renewable energy, balancing measurement indicators with unit operating economics.

[0019] Example 2 In step S1, a system architecture consisting of a master station, a substation, and a communication network is constructed. The master station deploys a frequency acquisition server, an instruction generation unit, and a dispatching data network core switch. The substation includes an edge computing gateway, a hard-wired interface unit, a synchronous timing terminal, and an AC / DC redundant power supply module. The communication network adopts a dedicated dispatching data network with a dual-ring network structure and is configured with the OSPF dynamic routing protocol and the TCP transmission protocol. Use a finite state machine to initialize the substation equipment, define a state set including cold start state, hardware self-test state, safety authentication state, communication handshake state, standby operation state, and fault lock state, and calculate the probability of hardware self-test components passing using an exponential distribution failure rate model; The M / M / 1 model of queuing theory is used to calculate the time consumption of security authentication, the first-order autoregressive model AR(1) is used to predict the communication link delay, and fault detection and state switching are realized based on multi-dimensional logical expressions. Among them, the hardware self-test component is calculated through the probability formula: , the time-consuming formula for security authentication is , the communication delay prediction formula is ; The fault judgment formula is: , through the model and formula, the state verification and conversion of the substation equipment from cold start to standby operation are realized to ensure the normal operation of the system components.

[0020] First, the system architecture is constructed, and the frequency acquisition server, instruction generation unit, and dispatching data network core switch are deployed at the main station; the substation is equipped with an edge computing gateway, a hard-wired interface unit, a synchronous timing terminal, and an AC / DC redundant power supply module; the communication network adopts a dedicated dispatching data network with a dual-ring network structure, and is configured with the OSPF dynamic routing protocol and the TCP transmission protocol. Then, a finite state machine is used to initialize the substation equipment and define a state set, which includes the cold start state, hardware self-test state, security authentication state, communication handshake state, standby operation state, and fault lock state. In the hardware self-test link, the probability of component passing is calculated according to the exponential distribution failure rate model, and the formula is: ,in is the component passing probability, is the component failure rate, To test the time consumption, we can evaluate the reliability of hardware self-test. We use the queuing theory M / M / 1 model to calculate the time consumption of security authentication. The formula is: , It takes time for security authentication. is the authentication processing rate, is the arrival rate of authentication requests, which is used to estimate the time required for authentication and avoid system congestion. The first-order autoregressive model AR(1) is used to predict the communication link delay. The formula is: , is the communication delay predicted for the next moment, is the communication delay at the current moment, is the autoregressive coefficient, is Gaussian white noise to predict communication stability in advance; fault detection and state switching are realized based on multi-dimensional logical expressions, and the fault judgment formula is , To determine the fault result, this formula is used to achieve complete status verification and conversion of substation equipment from cold start to standby operation, ensuring the normal operation of system components.

[0021] This solution establishes a layered system architecture, clearly defining the functions and configurations of the master station, substations, and communication network, ensuring the system's fundamental stability and data transmission efficiency. Utilizing finite state machines and various mathematical models to initialize and manage substation equipment, this approach enables quantitative assessment of hardware reliability, prediction of security authentication time and communication latency, timely fault detection, and precise state transitions. This prevents system anomalies caused by device initialization issues or operational failures, ensuring proper coordination between components during startup and operation, and laying a solid foundation for subsequent grid frequency signal identification and processing. This improves the overall system reliability and stability, while reducing operational risks and maintenance costs.

[0022] The actual deployment steps are as follows: Master station deployment: Configure a high-performance server in the master station computer room as the frequency acquisition server, install the Linux operating system (such as CentOS 8), deploy a frequency acquisition program developed in Python, call the `pyserial` module to implement serial port communication with the frequency transmitter, and complete the grid frequency signal acquisition. The instruction generation unit can use MATLAB / Simulink to build the control algorithm model, compile the optimized control strategy into an executable file and deploy it on the server. Use Huawei CloudEngine S12700 series as the core switch of the dispatching data network, configure the OSPF dynamic routing protocol and TCP / IP transmission protocol parameters through the command line interface, and establish the communication link between the master station and the substation.

[0023] Substation deployment: The edge computing gateway uses Advantech's UNO-3083G industrial computer, pre-installed with Windows Embedded Standard 7, runs a C-based data processing program, and integrates the `Newtonsoft.Json` module to process communication data. The hard-wired interface unit uses a Weidmüller terminal block to achieve electrical isolation and connection of signals. The synchronization terminal deploys a Beidou / GPS dual-mode clock server (such as the Beidou Navigation BDS-2000) to synchronize time with the substation equipment through the NTP protocol. The AC / DC redundant power supply module is installed with a Phoenix QUINT-PS power supply to ensure the stability of the power supply for the substation equipment.

[0024] Communication network deployment: A dedicated dispatching data network with a dual-ring structure was constructed using optical fiber. At the network layer, the Huawei eNSP simulator was used for network topology design and configuration simulation. During actual deployment, the OSPF protocol was configured through the Huawei switch command line, and different VLANs were divided to achieve network isolation. At the transport layer, a data transmission program was developed based on the TCP protocol, and the `socket` module was used to establish reliable communication connections between the master station and substations.

[0025] Software tools: The finite state machine modeling tool Stateflow (integrated in MATLAB / Simulink) is used to design the substation device state transition logic, and the model is generated into C code and burned into the edge computing gateway for operation; in the hardware self-test phase, a Python script is used to call the `psutil` module to monitor the device hardware status (such as CPU temperature and memory usage) in real time, and the probability of component passing is calculated by combining the exponential distribution failure rate model; the OpenSSL library is used for security authentication to implement digital certificate verification and encrypted communication, and the security authentication time is calculated using the queuing theory M / M / 1 model; the communication delay prediction uses Python's `statsmodels` library to implement the first-order autoregressive model AR(1), and historical communication delay data is analyzed and predicted; the fault detection and state switching logic is implemented by writing a program in C, which judges the device status in real time based on multi-dimensional logical expressions and triggers the corresponding state switching action.

[0026] Debugging tools: Use Wireshark network analysis software to capture packets and analyze data transmission between the master station and substation to ensure that the communication protocol is correct and the data is complete. Use MATLAB's data visualization tool to draw frequency signal change curves to evaluate system initialization and data acquisition and processing effects. Use the logging function of the edge computing gateway to view the substation device status switching process and fault information for problem troubleshooting and optimization.

[0027] Operation monitoring: Deploy the Zabbix monitoring system to monitor the main site server, sub-site equipment and communication network in real time, set thresholds for monitoring indicators such as CPU usage, memory usage, and communication delay, and send alarms via email or SMS when indicators are abnormal; regularly back up the configuration files and operation data of the main site and sub-site, and use the Rsync tool to achieve data synchronization and backup to ensure system data security.

[0028] Example 3 In step S2, the grid frequency, the power generation side frequency difference, and the communication status data are collected, and the median filter algorithm is applied to denoise the grid frequency signal. The filtered signal is obtained by establishing a sliding window of length N and taking the median value of the sampled values ​​in the window. The formula is: ; And determine whether to trigger signal redundancy check by calculating the standard deviation of the three-way filtered signal; The frequency difference change rate on the power generation side is calculated using the sliding window algorithm, and the frequency difference is fitted with a quadratic polynomial within a sliding window of length M. , solve the coefficient to get the rate of change , and introduce the forgetting factor α to perform exponential weighted averaging to obtain ; The EWMA algorithm based on the Kalman filter framework is used to predict the communication state, and the state transition equation and the observation equation Perform recursive calculations to dynamically predict the round-trip delay and packet loss rate of TCP links; Fuzzy logic is introduced to evaluate the system status. By defining the fuzzy sets and membership functions of the power grid frequency difference, frequency difference change rate, and communication delay, a rule base is constructed and the Mamdani reasoning algorithm and center of gravity method are used. Perform reasoning and defuzzification, and use the formula according to the fuzzy output results Dynamically adjust local switching thresholds and instruction transmission strategies.

[0029] Usage: First, collect the grid frequency, power generation side frequency difference, and communication status data. For the grid frequency signal, apply the median filter algorithm to remove noise, and establish a length of The sliding window of the three frequency transmitters is integrated to obtain the sampling values ​​of the three frequency transmitters. Calculate, where for The grid frequency signal after filtering at all times, The three original frequency signals are taken as the intermediate values ​​to obtain the filtered signals, and the standard deviation of the three filtered signals is calculated. When the standard deviation exceeds the set threshold, the signal redundancy check is triggered.

[0030] When calculating the frequency difference change rate on the power generation side, the sliding window algorithm is used. A quadratic polynomial is fitted to the frequency difference within the sliding window of ,in for Frequency difference on the power generation side at every moment, is the polynomial coefficient, and the coefficient is solved by the least square method to obtain the rate of change , for The rate of change of the time frequency difference; then the forgetting factor is introduced , through the formula Perform exponentially weighted averaging, for The frequency difference change rate after exponential weighted average at each moment, It is the frequency difference change rate after exponential weighted averaging at the previous moment, which is used to suppress high-frequency noise and highlight trend changes.

[0031] For communication state prediction, the EWMA algorithm based on the Kalman filter framework is used to calculate the state transition equation. and the observation equation Perform recursive calculation, where for The state vector at that moment (including TCP link round-trip delay, packet loss rate, etc.), is the state transition matrix, is the process noise, for The moment observation vector (measured communication status data), is the observation matrix, To observe noise, dynamic prediction of TCP link round-trip delay and packet loss rate is achieved.

[0032] Finally, fuzzy logic is introduced to evaluate the system status. The fuzzy sets and membership functions of power grid frequency difference, frequency difference change rate, and communication delay are defined first, and a rule base is constructed. The Mamdani reasoning algorithm and the center of gravity method are used. Perform reasoning and defuzzification, is the exact output value of fuzzy inference, For fuzzy sets The membership degree at the location is calculated based on the fuzzy output result through the formula Dynamically adjust local switching thresholds and instruction transmission strategies, is the adjusted primary frequency modulation dead zone threshold, is the baseline threshold, is the adjustment coefficient, is the predicted round-trip delay of communication.

[0033] This solution utilizes multiple algorithms to collaboratively process multi-dimensional data, significantly improving data accuracy and reliability. The median filter algorithm effectively removes impulse noise from the grid frequency signal. Combined with a redundant check mechanism based on standard deviation, it ensures frequency signal stability and accuracy, providing a reliable basis for subsequent frequency control. A sliding window algorithm, combined with quadratic polynomial fitting and an exponentially weighted average algorithm, accurately calculates the frequency difference change rate, effectively suppresses noise interference, and clearly captures frequency trends on the generator side, improving the timeliness and accuracy of frequency control. The EWMA algorithm, based on the Kalman filter framework, dynamically predicts communication status, enabling the system to proactively detect communication anomalies and enhance its resilience to communication failures. A fuzzy logic evaluation system comprehensively assesses system status using multi-dimensional data and dynamically adjusts local switching thresholds and command transmission strategies, enabling the system to adapt to actual operating conditions. This ensures grid frequency stability while optimizing resource allocation, improving overall system performance and stability, and mitigating operational risks associated with inaccurate data or untimely system responses.

[0034] Example 4 In step S3, an improved PID algorithm is used to track the frequency change, which introduces a feedforward compensation term To respond to frequency mutations in advance and combine with anti-saturation mechanism Suppress the integral accumulation when the deviation is large, and use the incomplete differential filtering formula Smoothed differential output; Use particle swarm optimization algorithm to adjust PID parameters online, and update the speed formula through And the position update formula Minimize the comprehensive cost function ; Dynamically adjust the primary frequency modulation dead zone threshold through the formula Update the adjustment coefficient by combining the least squares method ; Using the hidden Markov model to predict communication failures, the forward probability formula Calculate state probability based on risk index Trigger local mode smooth switching in advance, and use the weight function when switching Realize linear transition of control signal.

[0035] When tracking frequency changes, an improved PID algorithm is used. Feedforward compensation is introduced ,in for The feedforward control quantity at time , is the feedforward gain, for The frequency difference change rate at time, It is the advance time constant, which responds to frequency mutation in advance; combined with the anti-integral saturation mechanism , for The integration coefficient at time, is the basic integral coefficient, is the frequency deviation, is the dead zone threshold of the primary frequency modulation, which suppresses the integral accumulation when the deviation is large; through the incomplete differential filtering formula , for The differential control quantity at the moment, is the differential coefficient, is the filtering time constant, is the rate of change of deviation, is the sampling period, smoothing the differential output.

[0036] Use particle swarm optimization algorithm to adjust PID parameters online, and update the speed formula through And the position update formula , For the The particle in The speed of generation, is the inertia weight, For the The particle in The speed of generation, 、 is the learning factor, 、 is a random number between $[0,1]$, For the The individual historical optimal position of each particle, For the The particle in The position of the generation, $gbest$ is the global optimal position, minimizing the comprehensive cost function , is the cost function, is the exponential forgetting factor, To control the energy penalty coefficient, To control the output.

[0037] Dynamically adjust the primary frequency modulation dead zone threshold through the formula , for The dead zone threshold of a frequency modulation at the moment, is the baseline threshold, 、 is the adjustment coefficient, for The grid load rate at the moment, for Standard deviation of the time-frequency difference signal; update the adjustment coefficient in combination with the least squares method , for The adjustment coefficient vector at time , is the covariance matrix, is the input feature vector, This is the actual assessment deviation.

[0038] Using the hidden Markov model to predict communication failures, the forward probability formula , For Always in the state And the observation sequence is The probability of From the state Transfer to state The probability of In state Observed The probability of calculating the state probability; based on the risk index , for The communication failure risk index at the moment triggers the smooth switching of local mode in advance, and the weight function is used during the switching , is the switching trigger moment, is the transition time to achieve linear transition of the control signal.

[0039] This solution significantly improves frequency tracking accuracy and system stability by combining an improved PID algorithm with multiple optimization strategies. The feedforward compensation term enables proactive response to frequency changes, significantly shortening system adjustment time. The anti-integral windup mechanism and incomplete differential filtering effectively suppress overshoot and high-frequency noise interference, avoiding drastic fluctuations in the controlled variable and enhancing the system's robustness under complex operating conditions. A particle swarm optimization algorithm adjusts PID parameters in real time to minimize the overall cost function, enabling the controller to adapt to changes in grid operating conditions and optimize control performance. The primary frequency regulation deadband threshold is dynamically adjusted, optimized in real time based on grid load factor and frequency difference fluctuations, reducing unnecessary frequency regulation actions and balancing system stability and economic efficiency. A hidden Markov model predicts communication failures and, combined with a smooth switching strategy, triggers local mode in advance, effectively avoiding control failures caused by communication failures and ensuring continuous and reliable system operation. Overall, this improves the efficiency and reliability of grid frequency control and reduces operational risks.

[0040] Example 5 In step S4, the master, hybrid and local states are switched according to the communication state and load changes. By defining a three-dimensional state space ( , , ) Determine the state transition condition, when 50ms≤ <100ms and 1%≤ <5% triggers the master mode to switch to hybrid mode. ≥100ms or ≥5% or >0.9 triggers the switch from hybrid mode to local mode; Using weighted fusion formula in hybrid mode Calculate the action signal, where the weight , ; The dynamic threshold adaptive algorithm is based on the formula Automatically adjust the frequency modulation dead zone, every 100ms according to the latest collected data through the formula Update threshold; Handle switching conflicts through priority decision rules when a communication interruption or substation hardware failure is detected or When the value is >0.95, local mode is triggered directly. In non-emergency switching, state transfer is performed through a 500ms buffer period. Using the formula Calculate the switching delay by the formula Assess load shock and use the exponential moving average formula Store and analyze historical switching data.

[0041] Usage: First define the three-dimensional state space with communication delay ( , in ms), packet loss rate ( , in %) and grid load rate ( , normalized to [0,1]) as the state variable, and the mode switching condition is determined accordingly. and When the system switches from master mode to hybrid mode; when or or , the system switches from hybrid mode to local mode.

[0042] In hybrid mode, a weighted fusion formula is used Calculate the action signal, where for Action signal in moment mixed mode, for The action signal of the master mode at all times, for The action signal of the local mode at the moment. Weight , , is the signal weight of the master mode, is the local mode signal weight, 、 is the adjustment coefficient, 、 Toggle midpoint parameter for weights.

[0043] The dynamic threshold adaptive algorithm is based on the formula Automatically adjust the frequency dead zone, where for The FM dead zone threshold at the moment, is the baseline threshold, 、 、 is the adjustment coefficient, for Frequency difference at each moment. According to the latest collected data, the formula Update threshold, is the updated threshold, is the threshold at the previous moment.

[0044] Handle switching conflicts through priority decision rules when a communication interruption or substation hardware failure is detected or Directly trigger the local mode when switching is not urgent; The state transfer is executed during the buffer period. The formula is used Calculate the switching delay, where is the switching delay time, is the switching completion moment, is the switching start time; through the formula Evaluate load shock, where is the load impact degree, is the power at the time of switching completion, is the power at the start of switching, is the rated power; using the exponential moving average formula Store and analyze historical switching data, including For the The exponential moving average of the data, is the smoothing factor, For the The original data.

[0045] This solution significantly improves the system's adaptability and stability under diverse operating conditions by establishing a three-state switching mechanism based on communication status and load fluctuations. Clear state transition conditions and priority determination rules ensure the system can quickly and accurately switch between master, hybrid, and local modes. In emergency situations such as communication failures or high loads, local mode can be directly triggered to ensure continuous system operation. A weighted fusion strategy in hybrid mode dynamically assigns weights to master and local signals based on communication quality and load factor, effectively preventing control failures caused by communication delays or instability. A dynamic threshold adaptive algorithm adjusts the frequency regulation deadband in real time, enabling the system to better adapt to changing grid conditions, reducing unnecessary frequency regulation operations and improving energy efficiency. Switching delay calculation, load impact assessment, and exponential moving average analysis of historical data help quantify system performance, optimize switching strategies, and minimize the impact of mode switching on grid operations. This improves overall system reliability and cost-effectiveness, ensuring stable grid operation while reducing operating costs for power generation companies.

[0046] Example 6 In step S5, the control effect is verified using digital twin technology, a reinforcement learning model is constructed to continuously optimize the control strategy, and the control effect and system safety are verified based on digital twin technology; The digital twin model includes a grid-side model based on the power system power flow equation and frequency response characteristics, a communication network model based on queuing theory and network delay distribution, and a replicated controller digital twin. A Bayesian optimization algorithm is used to minimize the deviation between the twin model and the physical system to calibrate the model parameters. The reinforcement learning model adopts Markov decision process modeling. The state space includes frequency, frequency deviation, communication delay, packet loss rate, load rate and operation mode. The action space includes PID parameters, frequency modulation dead zone threshold and mixed mode weight. The reward function comprehensively considers frequency tracking error, power fluctuation, communication risk and control smoothness. A proximal strategy optimization algorithm is used to train the control strategy, and an entropy regularization term is added to the objective function to enhance exploration. A control effect verification framework is constructed, including frequency quality indicators, assessment compliance rate, and switching performance indicators. A closed-loop optimization process is formed through offline training, online deployment and adaptive update mechanism; a risk assessment model is used to calculate the system operation risk, and a safety penalty term is added to the reinforcement learning objective function to ensure that the optimization process does not violate the safety boundary.

[0047] In a preferred embodiment, the grid-side model includes a set of differential algebraic equations: ; in, is the generator power angle, is the angular velocity, is the inertia time constant, is the mechanical power, is the electromagnetic power, is the damping coefficient; the queuing delay in the communication network model is expressed by the formula Calculate, where is the arrival rate, is the service rate; the Bayesian optimization algorithm is implemented by formula Calibrate model parameters; the reward function is expressed as Calculation: The objective function of the proximal strategy optimization algorithm is calculated by the formula ; Calculation: The root mean square error in the frequency quality index is calculated by the formula Calculation; the risk assessment model is calculated by the formula Calculate system operation risks.

[0048] First, a digital twin model is constructed based on the power system flow equation and frequency response characteristics, using a set of differential algebraic equations. Construct a grid-side model, where is the generator power angle, is the angular velocity, is the inertia time constant, is the mechanical power, is the electromagnetic power, is the damping coefficient, is the synchronous angular velocity, is the rated angular velocity; based on queuing theory and network delay distribution, through the formula Compute queuing delay in a communication network model where is the arrival rate, is the service rate; at the same time, the controller digital twin is reproduced. Using the Bayesian optimization algorithm, according to the formula Calibrate the model parameters, where is the optimal parameter, is the parameter vector, is the weight of each data point, For physical systems Output at any moment, For digital twin models Moment based on parameters output to ensure the consistency between the digital twin model and the physical system.

[0049] Then, a reinforcement learning model is constructed, using Markov decision process modeling, defining the state space as frequency, frequency difference, communication delay, packet loss rate, load rate and operation mode; the action space includes PID parameters, frequency modulation dead zone threshold and mixed mode weight; through the formula Calculate the reward function, where In state Execute an action Rewards, is the weight coefficient, is the current frequency, is the reference frequency, is the power fluctuation, is the communication risk index, and are the actions at the current and previous moments respectively. The control strategy is trained using the proximal strategy optimization algorithm, and the objective function Optimization parameters, where is the objective function value, For expectations, is the probability ratio, For advantage estimation, To trim parameters, an entropy regularization term is added to the objective function to enhance exploration capabilities.

[0050] Construct a frequency quality index (by formula Calculate the root mean square error, where is the frequency root mean square error, is the number of data points, is the actual frequency, As the reference frequency), assessment compliance rate and control effect verification framework of switching performance indicators. Through offline training, online deployment and adaptive update mechanism, a closed-loop optimization process is formed, and a risk assessment model is adopted. According to the formula Calculate system operation risk, including for System operation risks at all times, is the weight, For the The severity of the risk, For the The risk The probability of occurrence at each moment, adding a safety penalty term to the reinforcement learning objective function ensures that the optimization process does not violate the safety boundary.

[0051] This solution achieves high-precision verification and continuous optimization of control strategies through the deep integration of digital twin technology and reinforcement learning. The digital twin model is built based on precise physical equations and mathematical models. Its parameters are calibrated using a Bayesian optimization algorithm, enabling highly realistic simulations of physical system operation. This provides a near-realistic testing environment for control strategies, avoiding the costs and risks of high-risk experiments in actual systems. By scientifically defining the state space, action space, and reward function, and combining proximal policy optimization algorithms with entropy regularization, the reinforcement learning model can automatically explore and optimize control strategies under complex and changing grid conditions, effectively improving frequency tracking accuracy and reducing power fluctuations while balancing communication risks and control smoothness.

[0052] The establishment of a control effectiveness verification framework and risk assessment model quantitatively evaluates system performance and safety from multiple dimensions, ensuring that the optimized strategy is both efficient and reliable. This closed-loop optimization process enables the system to continuously iteratively update the control strategy based on actual operating data, enhancing the system's adaptability and robustness. This overall solution effectively improves the intelligence level of grid frequency control, the safety and stability of system operation, reduces the cost of manual intervention, and provides a strong guarantee for the stable operation of the grid in complex scenarios such as high penetration of renewable energy.

[0053] Example 7 In step S6, quantum encryption communication and active-active redundant architecture are used to ensure system security and reliability; the quantum encryption communication is implemented based on the decoy state BB84 protocol, and the decoy state strength formula is optimized. Improve the single photon detection rate and use the formula The final key rate is calculated to quantify the security, according to the formula Dynamically update keys; The active-active redundant architecture calculates synchronization delay through the active-standby synchronization protocol , using conflict resolution algorithms Determine switching priority based on dynamic load distribution formula Achieve traffic balance.

[0054] In the preferred solution, in step S6, the quantum encryption and active-active architecture synergy mechanism includes: Assign encryption paths to data flows, using the formula Select the optimal encrypted transmission path; the reliability of the active-active system is determined by the Markov model state transition rate matrix and system availability formula Evaluation, combined with the key cache synchronization formula Ensure key synchronization security through key consumption prediction formula and performance consistency evaluation formula Implement anomaly detection and security auditing.

[0055] Usage: To ensure system security and reliability, quantum encryption communication and active-active redundant architecture are adopted, and a coordination mechanism between the two is established.

[0056] In quantum encryption communication, based on the decoy state BB84 protocol, the formula Optimize the strength of the decoy state, where For optimal signal strength, is the single photon yield, is the single photon bit error rate, thereby improving the single photon detection rate; using the formula Calculate the final key rate, where is the final key rate, is the basis vector matching probability, is the total gain, is the error correction efficiency function, is the total bit error rate, is the binary entropy function, is the single photon bit error rate, which is used to quantify encryption security; according to the formula Dynamically update the key, where is the key update period, is the key buffer size, is the key consumption rate, It is a safe time window.

[0057] In terms of active-active redundant architecture, synchronization delay is calculated through the master-slave synchronization protocol ,in For synchronization delay, is the propagation delay, To handle delays, Delayed confirmation; using conflict resolution algorithms Determine the switching priority, where For the system The switching priority, is the weight coefficient, For the system health status, For the system The load, For the system Data freshness; based on dynamic load distribution formula Achieve traffic balance, For the system The load distribution ratio, For the system capacity, For the system The current load.

[0058] In the collaborative mechanism of quantum encryption and active-active architecture, the security level evaluation formula is used For data flow Allocate an encryption path, where For data flow The security level, is the weight coefficient, For data flow The sensitivity of For data flow The integrity requirement of Select the optimal encrypted transmission path, where For data flow The optimal path, For the path, is a set of paths, For path The quantum key distribution rate, For path The reliability of the active-active system is calculated by the Markov model state transition rate matrix. and system availability formula Assessment, including for System availability at all times, is the probability of normal operation of the dual systems, The probability of a single system operating normally; combined with the key cache synchronization formula Ensure key synchronization security, including is the key cache synchronization interval, is the minimum security key size, is the quantum key distribution rate, is the redundancy coefficient; through the key consumption prediction formula and performance consistency evaluation formula Implement anomaly detection and security auditing, including for The predicted key consumption at each moment, for Actual key consumption at the moment, for The predicted key consumption at each moment, is the coefficient, for Key consumption trends at each moment; for Performance consistency indicators of active-active systems at all times, is the number of performance indicators, They are the first performance indicator value.

[0059] This solution enhances system security and reliability across multiple dimensions through quantum cryptography, an active-active redundant architecture, and their collaborative mechanisms. Quantum cryptography, based on the decoyed BB84 protocol, optimizes key generation efficiency and security by optimizing key parameters. A dynamic key update mechanism ensures continuous communication security and prevents information leakage and tampering from a data encryption perspective. The active-active redundant architecture ensures uninterrupted system operation despite hardware failures or load fluctuations by precisely calculating synchronization delays, scientifically determining switching priorities, and dynamically allocating load, improving system availability and data integrity.

[0060] The synergistic mechanism further enhances system performance, intelligently selecting encryption paths based on the data stream security level, balancing security and transmission efficiency. System reliability is assessed using a Markov model, combined with key cache synchronization and anomaly detection formulas, enabling comprehensive monitoring of system operating status and risk prevention. The overall solution effectively mitigates risks such as external network attacks and internal hardware failures, ensuring stable and secure system operation in complex environments. It also provides a solid foundation for identifying and controlling grid frequency signal assessment standards, reducing economic losses and operational risks caused by security issues and system failures.

[0061] Example 8 Actual deployment steps: 1. System Architecture Construction and Initialization (S1) (1) Building the system architecture 1. Master Station Deployment: A high-performance server was deployed in the master station computer room as the frequency acquisition server. The CentOS operating system was installed, and Python, combined with the pyserial module, was used to implement serial communication with the frequency transmitter and complete grid frequency signal acquisition. The control algorithm model for the instruction generation unit was built using MATLAB / Simulink, compiled into an executable file, and deployed on the server. Huawei CloudEngine S12700 series switches were selected as the core switches for the dispatching data network. The OSPF dynamic routing protocol and TCP / IP transmission protocol were configured through the command line to establish communication links between the master station and the substations.

[0062] 2. Substation Deployment: The substation uses Advantech's UNO-3083G industrial computer as the edge computing gateway, pre-installed with Windows Embedded Standard 7, a data processing program written in C language, and an integrated `Newtonsoft.Json` module for communication data processing. Weidmüller terminal blocks are used as hard-wired interface units to achieve electrical isolation and connection of signals. Beidou Starlink BDS-2000 and other Beidou / GPS dual-mode clock servers are deployed as synchronization terminals to achieve accurate time synchronization of substation equipment through the NTP protocol. Phoenix QUINT-PS power supplies are installed as AC / DC redundant power modules to ensure stable power supply for substation equipment.

[0063] 3. Communication Network Deployment: A dedicated dispatching data network with a dual-ring structure was constructed using optical fiber. The network topology design and configuration were simulated using the Huawei eNSP simulator. In actual deployment, OSPF was configured through the Huawei switch command line to divide VLANs for network isolation. A data transmission program was developed using the TCP socket module to establish stable and reliable communication between the master station and slave stations.

[0064] (2) Device initialization Use Stateflow in MATLAB / Simulink to design the finite state machine model of the substation equipment, generate C code from the model, and then burn it to the edge computing gateway. In the hardware self-test phase, use Python to write scripts and combine the `psutil` module to monitor the hardware status of the equipment in real time. According to the exponential distribution failure rate model formula ( is the component passing probability, is the component failure rate, (Time-consuming for detection) Calculate the probability of component passing; security authentication uses the OpenSSL library to implement digital certificate verification and encrypted communication, based on the queuing theory M / M / 1 model formula ( It takes time for security authentication. is the authentication processing rate, The time consumption of security authentication is calculated by using the arrival rate of authentication requests. The communication delay prediction uses Python's `statsmodels` library to implement the first-order autoregressive model AR(1), and the formula ( is the communication delay predicted for the next moment, is the communication delay at the current moment, is the autoregressive coefficient, is Gaussian white noise) for analysis and prediction; use C to write a program based on multi-dimensional logical expressions and fault judgment formulas Realize fault detection and status switching, and complete the status verification and conversion of substation equipment from cold start to standby operation.

[0065] 2. Data Collection and Processing (S2) Deploy a custom-developed data acquisition program on the data acquisition device to collect real-time grid frequency, power generation side frequency difference, and communication status data. Use Python to write a median filter program to establish a length of The sliding window of The grid frequency signal is denoised and the standard deviation of the three-way filtered signal is calculated. When the standard deviation exceeds the set threshold, the signal redundancy check is triggered.

[0066] Use Python to write a sliding window algorithm program with a length of A quadratic polynomial is fitted to the frequency difference within the sliding window of , solve the coefficient by the least square method to get the rate of change , and introduce the forgetting factor , through the formula Perform exponentially weighted averaging.

[0067] Use the EWMA algorithm based on the Kalman filter framework to predict the communication status and use Python to implement the state transition equation and the observation equation The recursive calculation of for The state vector at the moment, is the state transition matrix, is the process noise, for The moment observation vector, is the observation matrix, is the observation noise.

[0068] Fuzzy logic is introduced to evaluate the system status. The fuzzy logic toolbox of MATLAB is used to define the fuzzy sets and membership functions of the power grid frequency difference, frequency difference change rate, and communication delay. The rule base is constructed and the Mamdani inference algorithm and the center of gravity method are used. Perform reasoning and defuzzification, and use the formula according to the fuzzy output results Dynamically adjust local switching thresholds and instruction transmission strategies.

[0069] 3. Frequency Tracking and Control (S3) Deploy the control program developed in C language on the edge computing gateway and the main station server to implement the improved PID algorithm. Introduce the feedforward compensation term , combined with the anti-saturation mechanism , and through the incomplete differential filtering formula Improve the PID algorithm.

[0070] Use Python to write a particle swarm optimization algorithm program and update the formula through speed And the position update formula Minimize the comprehensive cost function , adjust PID parameters online.

[0071] Dynamically adjust the primary frequency modulation dead zone threshold through the formula Update the adjustment coefficient by combining the least squares method .

[0072] Use Python to implement hidden Markov model to predict communication failures, through the forward probability formula Calculate state probability based on risk index Trigger local mode smooth switching in advance, and use the weight function when switching Realize linear transition of control signal.

[0073] 4. Mode Switching and Threshold Adjustment (S4) Embed mode switching and threshold adjustment logic in the master and slave control programs. Define the three-dimensional state space ( , , ),when and When the system switches from master mode to hybrid mode; when or or , the system switches from hybrid mode to local mode.

[0074] In hybrid mode, a weighted fusion formula is used Calculate the action signal, where the weight , .

[0075] The dynamic threshold adaptive algorithm is based on the formula Automatically adjust the frequency dead zone, every According to the latest collected data through the formula Update the threshold.

[0076] Handle switching conflicts through priority decision rules when a communication interruption or substation hardware failure is detected or Directly trigger the local mode when switching is not urgent; The state transfer is executed during the buffer period. The formula is used Calculate the switching delay by the formula Assess load shock and use the exponential moving average formula Store and analyze historical switching data.

[0077] 5. Control Effect Verification and Strategy Optimization (S5) Use MATLAB and Python to build a digital twin model based on the differential algebraic equations of the power system flow equation and frequency response characteristics Construct a grid-side model; based on queuing theory and network delay distribution, use the formula Calculate the queuing delay in the communication network model; and use C to reproduce the controller digital twin. Use Python to write the Bayesian optimization algorithm program, according to the formula Calibrate model parameters.

[0078] Build a reinforcement learning model, use Python's `stable_baselines3` library to implement Markov decision process modeling, define the state space and action space, and use the formula Calculate the reward function. Use the proximal policy optimization algorithm to train the control strategy, through the objective function Optimize parameters and add entropy regularization terms to the objective function to enhance exploration capabilities.

[0079] Construct a frequency quality index (by formula Calculate the root mean square error), assessment compliance rate and control effect verification framework of switching performance indicators. Through offline training, online deployment and adaptive update mechanism, a closed-loop optimization process is formed, and the risk assessment model is implemented in Python. According to the formula Calculate the system operation risk and add a safety penalty term to the reinforcement learning objective function to ensure that the optimization process does not violate the safety boundary.

[0080] VI. Safety and Reliability Assurance (S6) Deploy quantum encryption and active-active redundancy related programs on communication equipment and system servers. Quantum encryption communication is based on the decoy state BB84 protocol, and the program is written in Python. Optimize the strength of the decoy state and use the formula Calculate the final key rate according to the formula Dynamically update the key.

[0081] Active-active redundant architecture calculates synchronization delay through active-standby synchronization protocol , using conflict resolution algorithms Determine switching priority based on dynamic load distribution formula Achieve traffic balance.

[0082] In the collaborative mechanism of quantum encryption and active-active architecture, the security level evaluation formula is used Assign encryption paths to data flows, using the formula Select the optimal encrypted transmission path. The reliability of the active-active system is calculated by the Markov model state transition rate matrix. and system availability formula Evaluation, combined with the key cache synchronization formula Ensure key synchronization security through key consumption prediction formula and performance consistency evaluation formula Implement anomaly detection and security auditing.

[0083] The above embodiments are merely preferred technical solutions of the present invention and should not be construed as limiting the present invention. The scope of protection of the present invention shall be the technical solutions set forth in the claims, including equivalent alternatives to the technical features of the technical solutions set forth in the claims. In other words, equivalent alternatives and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method for identifying grid frequency signal assessment standards, characterized by: The method includes: S1. Build a system architecture consisting of a master station, substations, and communication network, and implement device initialization through a finite state machine; S2, collects multi-dimensional data on grid frequency, power generation side frequency difference, and communication status, applies median filtering, sliding window, and EWMA algorithms to process the signals, introduces fuzzy logic to evaluate system status, and dynamically adjusts the response strategy; S3: Use an improved PID algorithm to track frequency changes, combine feedforward compensation with PSO parameter optimization, and use a hidden Markov model to predict communication failures and trigger local mode switching in advance; S4: Switch between the master, hybrid, and local states according to the communication status and load changes. The dynamic threshold adaptive algorithm automatically adjusts the frequency regulation dead zone according to the grid load. S5. Use digital twin technology to verify the control effect; S6, optional quantum encryption and active-active redundancy design ensure system security and reliability.

2. The method for identifying a grid frequency signal assessment standard according to claim 1, wherein: In step S1, a system architecture consisting of a master station, a substation, and a communication network is constructed. The master station deploys a frequency acquisition server, an instruction generation unit, and a dispatching data network core switch. The substation includes an edge computing gateway, a hard-wired interface unit, a synchronous timing terminal, and an AC / DC redundant power supply module. The communication network adopts a dedicated dispatching data network with a dual-ring network structure and is configured with the OSPF dynamic routing protocol and the TCP transmission protocol. Use a finite state machine to initialize the substation equipment, define a state set including cold start state, hardware self-test state, safety authentication state, communication handshake state, standby operation state, and fault lock state, and calculate the probability of hardware self-test components passing using an exponential distribution failure rate model; The M / M / 1 model of queuing theory is used to calculate the time consumption of security authentication, the first-order autoregressive model AR(1) is used to predict the communication link delay, and fault detection and state switching are realized based on multi-dimensional logical expressions. Among them, the hardware self-test component is calculated through the probability formula: , the time-consuming formula for security authentication is , the communication delay prediction formula is ; The fault judgment formula is: , through the model and formula, the state verification and conversion of the substation equipment from cold start to standby operation are realized to ensure the normal operation of the system components.

3. The method for identifying the grid frequency signal assessment standard according to claim 1, characterized in that: In step S2, the grid frequency, the power generation side frequency difference, and the communication status data are collected, and the median filter algorithm is applied to denoise the grid frequency signal. The filtered signal is obtained by establishing a sliding window of length N and taking the median value of the sampled values ​​in the window. The formula is: ； And determine whether to trigger signal redundancy check by calculating the standard deviation of the three-way filtered signal; The frequency difference change rate on the power generation side is calculated using the sliding window algorithm, and the frequency difference is fitted with a quadratic polynomial within a sliding window of length M. , solve the coefficient to get the rate of change , and introduce the forgetting factor α to perform exponential weighted averaging to obtain ; The EWMA algorithm based on the Kalman filter framework is used to predict the communication state, and the state transition equation and the observation equation Perform recursive calculations to dynamically predict the round-trip delay and packet loss rate of TCP links; Fuzzy logic is introduced to evaluate the system status. By defining the fuzzy sets and membership functions of the power grid frequency difference, frequency difference change rate, and communication delay, a rule base is constructed and the Mamdani reasoning algorithm and center of gravity method are used. Perform reasoning and defuzzification, and use the formula according to the fuzzy output results Dynamically adjust local switching thresholds and instruction transmission strategies.

4. The method for identifying the grid frequency signal assessment standard according to claim 1, characterized in that: In step S3, an improved PID algorithm is used to track the frequency change, which introduces a feedforward compensation term To respond to frequency mutations in advance and combine with anti-saturation mechanism Suppress the integral accumulation when the deviation is large, and use the incomplete differential filtering formula Smoothed differential output; Use particle swarm optimization algorithm to adjust PID parameters online, and update the speed formula through And the position update formula Minimize the comprehensive cost function ; Dynamically adjust the primary frequency modulation dead zone threshold through the formula Update the adjustment coefficient by combining the least squares method ; Using the hidden Markov model to predict communication failures, the forward probability formula Calculate state probability based on risk index Trigger local mode smooth switching in advance, and use the weight function when switching Realize linear transition of control signal.

5. The method for identifying grid frequency signal assessment standards according to claim 1, characterized in that: In step S4, the master, hybrid and local states are switched according to the communication state and load changes. By defining a three-dimensional state space ( , , ) Determine the state transition condition, when 50ms≤ <100ms and 1%≤ <5% triggers the master mode to switch to hybrid mode. ≥100ms or ≥5% or >0.9 triggers the switch from hybrid mode to local mode; Using weighted fusion formula in hybrid mode Calculate the action signal, where the weight , ; The dynamic threshold adaptive algorithm is based on the formula Automatically adjust the frequency modulation dead zone, every 100ms according to the latest collected data through the formula Update threshold; Handle switching conflicts by using priority decision rules. When a communication interruption or substation hardware failure is detected, When the value is >0.95, local mode is triggered directly. In non-emergency switching, state transfer is performed through a 500ms buffer period. Using the formula Calculate the switching delay by the formula Assess load shock and use the exponential moving average formula Store and analyze historical switching data.

6. The method for identifying the grid frequency signal assessment standard according to claim 1, characterized in that: In step S5, the control effect is verified using digital twin technology, a reinforcement learning model is constructed to continuously optimize the control strategy, and the control effect and system safety are verified based on digital twin technology; The digital twin model includes a grid-side model based on the power system power flow equation and frequency response characteristics, a communication network model based on queuing theory and network delay distribution, and a replicated controller digital twin. A Bayesian optimization algorithm is used to minimize the deviation between the twin model and the physical system to calibrate the model parameters. The reinforcement learning model adopts Markov decision process modeling. The state space includes frequency, frequency deviation, communication delay, packet loss rate, load rate and operation mode. The action space includes PID parameters, frequency modulation dead zone threshold and mixed mode weight. The reward function comprehensively considers frequency tracking error, power fluctuation, communication risk and control smoothness. The control strategy is trained using a proximal policy optimization algorithm, and an entropy regularization term is added to the objective function to enhance exploration. Build a control effect verification framework, including frequency quality indicators, assessment compliance rate, and switching performance indicators; A closed-loop optimization process is formed through offline training, online deployment, and adaptive update mechanisms; A risk assessment model is used to calculate the system operation risk, and a safety penalty term is added to the reinforcement learning objective function to ensure that the optimization process does not violate the safety boundary.

7. The method for identifying the grid frequency signal assessment standard according to claim 6, characterized in that: The grid-side model includes a set of differential algebraic equations: ； in, is the generator power angle, is the angular velocity, is the inertia time constant, is the mechanical power, is the electromagnetic power, is the damping coefficient; the queuing delay in the communication network model is expressed by the formula Calculate, where is the arrival rate, is the service rate; the Bayesian optimization algorithm is implemented by formula Calibrate model parameters; the reward function is expressed as Calculation: The objective function of the proximal strategy optimization algorithm is calculated by the formula ； Calculation: The root mean square error in the frequency quality index is calculated by the formula Calculation; the risk assessment model is calculated by the formula Calculate system operation risks.

8. The method for identifying the grid frequency signal assessment standard according to claim 1, wherein: step S6 In the process, quantum encryption communication and active-active redundant architecture are used to ensure the security and reliability of the system; the quantum encryption communication is based on the decoy state BB84 protocol, and the decoy state strength formula is optimized. Improve the single photon detection rate and use the formula The final key rate is calculated to quantify the security, according to the formula Dynamically update keys; The active-active redundant architecture calculates synchronization delay through the active-standby synchronization protocol , using conflict resolution algorithms Determine switching priority based on dynamic load distribution formula Achieve traffic balance.

9. The method for identifying the grid frequency signal assessment standard according to claim 8, characterized in that: In step S6, the quantum encryption and active-active architecture synergy mechanism includes: Assign encryption paths to data flows, using the formula Select the optimal encrypted transmission path; the reliability of the active-active system is determined by the Markov model state transition rate matrix and system availability formula Evaluation, combined with the key cache synchronization formula Ensure key synchronization security through key consumption prediction formula and performance consistency evaluation formula Implement anomaly detection and security auditing.

Images