A nuclear power unit transient abnormality diagnosis method based on load-driven frequency domain decoupling operator

Abstract

The application discloses a nuclear power unit transient abnormality diagnosis method based on a load-driven frequency domain decoupling operator and belongs to the field of nuclear power station operation fault diagnosis. In view of the problem that the prior art is difficult to detect extremely low amplitude hidden abnormalities under variable load conditions, a nominal model is obtained by identifying a reactor dynamic mechanism model under different loads; three characteristic sub-bands of low frequency, medium frequency and high frequency are adaptively divided according to a real-time load, and load-dependent compensation coefficients of the sub-bands are calculated by using a nonlinear scheduling function; then, the difference between the pulse frequency responses of a current operation model and the nominal model in the sub-bands is weighted and summed by using the compensation coefficients to construct a frequency domain decoupling operator; and a dynamic envelope threshold is generated according to the real-time load, and when the frequency domain decoupling operator exceeds the envelope threshold, a transient abnormality is determined and an alarm is given. The application can penetrate normal variable load transient fluctuations, sensitively capture extremely low amplitude abnormalities, and realize the organic unification of flexible containment of normal maneuver and rigid interception of hidden abnormalities.

Description

Technical Field

[0001] This invention belongs to the field of nuclear power plant operation fault diagnosis, and in particular relates to a method for diagnosing transient anomalies in nuclear power units based on load-driven frequency domain decoupling operators. Background Technology

[0002] Traditional nuclear power plants typically operate at full power and stably as baseload power sources. However, under the new power system dispatch requirements, nuclear power units must possess deep load tracking and transient maneuvering capabilities. This necessitates that the various process control systems within the nuclear power plant have a high degree of dynamic coordinated response and operational flexibility.

[0003] However, nuclear power units are highly complex nonlinear physical systems with strong coupling and large time delays. Under transient maneuvering conditions such as continuous variable load, the neutron dynamics and thermo-hydraulic parameters within the nuclear reactor, such as coolant temperature, flow rate, and pressure, undergo drastic nonlinear dynamic drift with power levels. This drastic time-varying characteristic of the system's global dynamics presents significant technical barriers to anomaly diagnosis and condition monitoring in the underlying control loop. Currently, most mainstream anomaly detection methods in industry rely on traditional time-domain residual analysis. For example, invention patent CN121541551A discloses a nuclear power plant load tracking method and system based on hybrid time-domain model predictive control. The method includes the following steps: dividing the entire time domain of the nuclear power plant load tracking operation scenario into a finite time domain and an infinite time domain, and constructing a hybrid time-domain predictive model that includes both the finite and infinite time domains; setting safety boundary constraints for state variables and operational constraints for control variables as operating constraints based on the safety regulations followed by the load tracking operation and the control requirements of the nuclear power plant; constructing a hybrid time-domain optimization objective function for load tracking; obtaining real-time state observations and steady-state setpoints through the distributed control module and the upper-level scheduling module of the nuclear power plant, respectively, and constructing two-point boundary conditions based on the real-time state observations and steady-state setpoints to constrain the predictive control optimization process; integrating the hybrid time-domain predictive model, operating constraints, hybrid time-domain optimization objective function, and two-point boundary conditions to construct and solve a rolling optimization problem, obtaining the optimal control sequence at the current moment, implementing closed-loop update iteration, and completing the nuclear power plant load tracking. However, during variable load transient processes, the normal transient peaks caused by power command switching can easily be confused with real early minor equipment faults, such as actuator jamming or performance degradation. If a fixed time-domain detection threshold is used, a large number of false alarms will be triggered during normal maneuvers; if the threshold is artificially relaxed to adapt to the variable operating condition envelope, the detection sensitivity of the system will be severely dulled, leading to high-risk missed detections.

[0004] On the other hand, the early degradation of critical control equipment in nuclear power plants under complex operating conditions is often highly concealed. These subtle, anomalous disturbances do not appear uniformly across the entire frequency band, but rather are highly concentrated in specific characteristic sub-bands. For example, aging and wear of mechanical components often manifest as noise distortion in the high-frequency band, while the slow drift of thermodynamic sensors is disguised within the low-frequency dominant dynamics. Traditional time-domain analysis methods perform global aliasing processing on broadband signals, failing to effectively isolate specific frequency band characteristics, resulting in a severely inadequate ability to identify early, concealed faults hidden in local frequency bands.

[0005] In summary, existing diagnostic methods for dealing with transient operational conditions of nuclear power units generally suffer from inherent limitations, such as poor threshold adaptability due to dynamic drift of the system's background and low detection rate of hidden anomalies caused by broadband signal aliasing. Therefore, there is an urgent need in this field for a novel diagnostic method that can deeply decouple the system's frequency domain characteristic manifold and whose detection mechanism is driven by real-time load dynamics, in order to accurately separate normal transients under varying operating conditions from real and weak anomalies, thus building a solid defense for the safe and stable operation of nuclear power units under complex power grid scheduling. Summary of the Invention

[0006] To address the technical challenges faced by nuclear power units participating in transient grid operations, such as deep load tracking, where the inherent high nonlinear dynamic drift of nuclear power units leads to high false alarm rates with traditional fixed threshold detection methods, and where conventional broadband time-domain residual analysis struggles to identify and isolate early, subtle faults and hidden anomalies within specific frequency bands, this invention provides a transient anomaly diagnosis method for nuclear power units based on a load-driven frequency-domain decoupling operator. This method identifies the reactor dynamic mechanism model under different loads to obtain the nominal model; then, it adaptively divides the real-time load into three characteristic sub-bands: low-frequency, medium-frequency, and high-frequency, and calculates the load-dependent compensation coefficient for each sub-band using a nonlinear scheduling function; next, it constructs a frequency-domain decoupling operator, which uses the compensation coefficients to weighted sum the pulse frequency response differences between the current operating model and the nominal model in each sub-band; finally, it generates a dynamic envelope threshold based on the real-time load, and when the frequency-domain decoupling operator exceeds the threshold, a transient anomaly is identified and an alarm is triggered.

[0007] Based on the above technical objectives, this embodiment provides a method for diagnosing transient anomalies in nuclear power units based on a load-driven frequency domain decoupling operator, including the following steps: Step 1: Establish a mechanism model based on the physical topology of the nuclear power unit in advance, transform it into a set of time-domain differential equations describing the dynamic characteristics of the nuclear power unit, and identify and obtain the nominal model for the transient maneuvering operation characteristics of the nuclear power unit. Step 2: Based on the current load, the full frequency domain space is adaptively decoupled into three characteristic sub-bands: low frequency, medium frequency, and high frequency, according to the nominal model. The sub-band boundaries dynamically expand and contract with the load. Step 3: Using the real-time acquired load command signal, calculate the compensation coefficients of the three characteristic sub-bands through the nonlinear gain scheduling function, and use the compensation coefficients to weight and sum the pulse frequency response differences between the current operating model and the nominal model in each sub-band to construct a frequency domain decoupling operator; Step 4: Dynamically generate a dynamic envelope threshold based on the current load. If the frequency domain decoupling operator is greater than the dynamic envelope threshold, it is determined that there is a transient anomaly in the nuclear power unit and an alarm is triggered.

[0008] Preferably, the nuclear power unit in step 1 is one or more of the following: nuclear power plant reactor power control system, coolant temperature control system, pressurizer pressure and water level control system, and steam generator water level control system.

[0009] Preferably, step 1 specifically includes: Based on the physical topology of the reactor core of a nuclear power plant, a dynamic mechanism model is established that includes neutron dynamics of the reactor, fuel temperature effect, coolant temperature effect and the reactivity value of control rods. The dynamic mechanism model is transformed into a set of time-domain differential equations describing the dynamic characteristics of nuclear power units. After linearization, a state-space expression is obtained, and time-varying parameters that drift nonlinearly with the reactor power level are included to obtain the mechanism model. Under multiple stable loads, test excitation signals are injected into the mechanism model, dynamic input and output data are collected, and the discrete state-space model and transfer function corresponding to each load point are identified, which serve as the nominal model under this load point.

[0010] This invention takes into account that in actual variable load operation, some key physical parameters are not constant, but drift nonlinearly with the reactor power level. Therefore, in order to make the simulation accurately reflect the real complex operating conditions of nuclear power plants, the time-varying effect of parameters is fully taken into account in the established mechanism model.

[0011] More preferably, the state variables in the state-space expression include: relative neutron density, fuel temperature deviation, and coolant temperature deviation, and the input variables include control rod displacement; The time-varying parameters include: control rod value gain, fuel reactivity temperature coefficient, and coolant reactivity temperature coefficient.

[0012] More preferably, the multiple stable loads include three operating points: low load, half load, and full load, which correspond to 10%, 50%, and 90% of the rated power, respectively.

[0013] Furthermore, the nominal model includes the full power range as well as the normal dynamic characteristics of the real-time load.

[0014] Preferably, step 2, which involves adaptively decoupling the full-frequency domain space into three characteristic sub-bands—low-frequency, mid-frequency, and high-frequency—based on the nominal model, includes: The sensitivity function and open-loop transfer function are calculated based on the transfer function in the nominal model. The sub-band boundary is determined based on the amplitude cross-frequency and phase cross-frequency of the sensitivity function and open-loop transfer function. Specifically, the frequency band where the absolute value of the sensitivity function exceeds the first threshold and includes the sensitivity peak is defined as the intermediate frequency (IF) feature segment; the low frequency (LFM) feature segment is the frequency range below the lower limit of the IF feature segment; and the high frequency (HF) feature segment is the frequency range above the IF feature segment.

[0015] This invention adaptively decouples the entire frequency domain into three characteristic sub-bands: low frequency, medium frequency, and high frequency. These sub-bands can respectively reflect the dominant transient response of the nuclear power unit, the slow thermodynamic drift, and the sensor measurement noise and early degradation distortion characteristics of the equipment. Through this adaptive division method, it is ensured that the boundaries of each frequency band can dynamically expand and contract with the variable load operation of the nuclear power unit.

[0016] Furthermore, the first threshold is .

[0017] Preferably, the nonlinear gain scheduling function in step 3 adopts the following formula: , in, As the reference load power, , and The real-time compensation coefficients satisfy normalization constraints and are preset scheduling parameters for the corresponding feature sub-bands. During transient maneuvering, the nuclear power unit updates its compensation coefficients based on real-time load command signals. This is to adaptively absorb the frequency domain characteristic shift caused by the nonlinearity of nuclear power units.

[0018] Preferably, the calculation formula for the frequency domain decoupling operator is as follows: , , in, This represents the pulse frequency response of the currently running model. The pulse frequency response of the nominal model. Represents the imaginary unit and satisfies , Represents the normalized frequency. Indicates frequency, It is the data sampling period. For the set of characteristic subbands, These are the compensation coefficients for each characteristic sub-band under the corresponding load.

[0019] Preferably, the step of dynamically generating a dynamic envelope threshold based on the current load includes: pre-setting threshold parameters at multiple discrete load points, wherein the threshold parameters include the mean and standard deviation and are positively correlated with the load level; and then generating a dynamic envelope threshold that changes continuously and smoothly with the load through interpolation.

[0020] On the other hand, the present invention also provides a nuclear power unit transient anomaly diagnosis device based on a load-driven frequency domain decoupling operator, including a memory and a processor. The memory is used to store a computer program, and the processor is used to implement the nuclear power unit transient anomaly diagnosis method based on a load-driven frequency domain decoupling operator when the computer program is executed.

[0021] Compared with the prior art, the present invention has the following beneficial effects: This invention breaks through the limitations of traditional broadband time-domain residual global aliasing, and finely decouples the slow thermodynamic sensor drift (low frequency), dominant dynamic anomalies (mid frequency), and early mechanical structure degradation distortion (high frequency) in a multi-frequency band space, which greatly improves the detection sensitivity of early weak faults and hidden anomalies.

[0022] This invention constructs a nonlinear gain scheduling function continuously driven by real-time load, adaptively assigning optimal compensation coefficients to each characteristic subband. This mechanism accurately suppresses the background nonlinear drift caused by transient power maneuvers in nuclear power units and effectively eliminates false alarms caused by frequent switching of operating conditions.

[0023] The frequency domain decoupling projection operator designed in this invention can sensitively capture the minute fluctuations of the characteristic manifold surface of nuclear power units, and its mathematical calculation structure is simplified, effectively reducing the online computing load of the industrial underlying control system and meeting the stringent real-time requirements of nuclear power plants.

[0024] This invention abandons the rigid, static, fixed threshold and constructs an envelope threshold mechanism that dynamically expands and contracts with real-time load. This allows the safety alarm threshold to automatically absorb normal transient fluctuations generated during operation, perfectly balancing the accuracy of abnormal alarms with the robustness of nuclear power units. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating the transient anomaly diagnosis method for nuclear power units based on load-driven frequency domain decoupling operators provided in an embodiment of the present invention.

[0026] Figure 2 This is a comparative verification diagram of time-domain output deviation caused by early weak and hidden anomalies under transient motor loads involved in the embodiment.

[0027] Figure 3A comparison diagram of the frequency domain decoupling operator and the dynamic envelope threshold provided in the embodiment. Detailed Implementation

[0028] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be noted that the embodiments described below are intended to facilitate the understanding of the present invention and do not constitute any limitation thereof.

[0029] The technical concept of this invention is as follows: Addressing the technical challenges faced by nuclear power units participating in transient power grid operations, such as deep load tracking, where the inherently high nonlinear dynamic drift of nuclear power units leads to a high false alarm rate with traditional fixed threshold detection methods, and where conventional broadband time-domain residual analysis struggles to identify and isolate early, subtle faults and hidden anomalies within specific frequency bands, this invention provides a transient anomaly diagnosis method for nuclear power units based on a load-driven frequency-domain decoupling operator. This method involves identifying the reactor dynamic mechanism model under different loads to obtain the nominal model; then, adaptively dividing the real-time load into three characteristic sub-bands (low-frequency, medium-frequency, and high-frequency), and calculating the load-dependent compensation coefficient for each sub-band using a nonlinear scheduling function; constructing a frequency-domain decoupling operator, which uses the compensation coefficients to weighted sum the pulse frequency response differences between the current operating model and the nominal model in each sub-band; and finally, generating a dynamic envelope threshold based on the real-time load. When the decoupling operator exceeds the threshold, a transient anomaly is identified and an alarm is triggered.

[0030] The invention will be further described in detail below with reference to a simulation example of the reactor power control system of a pressurized water reactor nuclear power plant, and in conjunction with the accompanying drawings. Figure 1 The flowchart shown is a method for diagnosing transient anomalies in nuclear power units based on load-driven frequency domain decoupling operators. The specific implementation steps are as follows: S1. Establish a mechanism model based on the physical topology of the nuclear power unit in advance, transform it into a set of time-domain differential equations describing the dynamic characteristics of the nuclear power unit, and identify and obtain the nominal model for the transient maneuvering characteristics of the nuclear power unit.

[0031] In this embodiment, a dynamic mechanism model of the reactor core is obtained by combining point reactor neutron dynamics analysis, core fuel and coolant temperature effects, and reactivity factors. The calculation formula is as follows: , in, The average neutron generation time; For equivalent single-group delayed neutron fraction; For time The corresponding relative neutron density; For time The corresponding relative equivalent single-group delayed neutron precursor nuclear density; The equivalent single-group delayed neutron precursor nuclear decay coefficient; The heat transfer coefficient between fuel and coolant; The mass flow rate and heat capacity of the coolant; Fuel temperature; The percentage of fuel power in the total reactor power; The heat capacity of the fuel; The heat capacity of the coolant; To control the induced reactivity of the control rod; The value gain (lifetime) of the control rod represents the responsiveness value per unit length of the control rod. It is a time-varying parameter that is related to the power level and position. To control rod displacement; The reactivity temperature coefficient of the fuel; The reactivity temperature coefficient of the coolant; Reactor power can be expressed as: , In the formula, The rated power of the reactor, therefore, can also be used. This is used to represent the relative power of the reactor.

[0032] Assuming in At any given moment, the reactor is in a steady state, at which point: , , , , When a nuclear power unit is subjected to minor external disturbances, it causes small changes in reactivity, resulting in the corresponding generation of... , , , , ,have: , , , , And the coolant inlet temperature remains constant.

[0033] After derivation and linearization, the system of differential equations is summarized as follows: .

[0034] After derivation, the linearized model at a specific power steady-state point can be described by the following state-space expression: , The time derivative of the state vector is represented; take the state vector respectively. Input vector Output vector for: ; The corresponding state matrix Input matrix Output matrix Feedforward matrix They are respectively: , , , .

[0035] The main constant parameters of the reactor core are as follows: Table 1. Main constant parameters of pressurized water reactor core

[0036] To ensure the simulation model accurately reflects the complex operating conditions of a real nuclear power plant, this embodiment fully incorporates the time-varying effects of parameters. In actual variable load operation, some key physical parameters are not constant but rather drift nonlinearly with the reactor power level. The time-varying parameters that vary with power in this embodiment are given by the following formula: , , , , .

[0037] Based on the established mechanism model, a wideband test excitation signal is injected under specific variable load conditions. According to the dynamic input and output data matrix of the reactor power control system, the state sequence of the reactor power control system is extracted using a data-driven subspace identification algorithm. Then, the discrete state-space model clusters and corresponding transfer functions under each power load point are directly identified, and the nominal model is obtained as the basis for subsequent frequency domain decoupling analysis.

[0038] S2. Based on the current load, the full frequency domain space is adaptively decoupled into three characteristic sub-bands: low frequency, medium frequency, and high frequency, based on the nominal model. The sub-band boundaries dynamically expand and contract with the load.

[0039] In this embodiment, to comprehensively capture hidden anomalies, the frequency range containing the dominant pole of the reactor power control system is decoupled into a mid-frequency characteristic sub-band reflecting the dominant transient response of the reactor power control system; the range below the dominant pole frequency and significantly affected by control law regulation is decoupled into a low-frequency characteristic sub-band reflecting slow thermodynamic drift; and the range above the cutoff frequency of the reactor power control system is decoupled into a high-frequency characteristic sub-band containing sensor measurement noise and early equipment degradation distortion characteristics. The division of each characteristic sub-band is not based on fixed static physical frequency values, but rather on the current power load of the reactor power control system. Adaptive decoupling is performed on the dynamic feature indicators below. Specifically: Calculate the sensitivity function based on the transfer function in the nominal model identified by S1: ,in, It is the identity matrix. The frequency response of the nominal model identified for S1. This is the controller frequency response of the reactor power control system, used to provide feedback on the dynamic characteristics of the reactor power control system.

[0040] Simultaneously, calculate the open-loop transfer function of the reactor power control system. Amplitude Crossover Frequency and phase crossover frequency The sub-bands are adaptively divided based on the sensitivity function, amplitude crossover frequency, and phase crossover frequency. Intermediate frequency characteristic sub-band: ( And includes sensitivity peaks. frequency band Divided into mid-frequency characteristic sub-bands This reflects the dominant transient response of the reactor power control system; The range below the mid-frequency characteristic subband, which is strongly suppressed by the control law, is decoupled into a low-frequency characteristic subband that reflects the slow thermodynamic drift of the reactor power control system: .

[0041] The range above the mid-frequency characteristic subband and the range of reactor power control system gain attenuation are decoupled into a high-frequency characteristic subband that includes high-frequency degradation distortion of actuators and measurement noise: .

[0042] This adaptive partitioning method ensures that the boundaries of each frequency band can dynamically expand and contract in response to the variable load operation of the reactor power control system.

[0043] S3. Using the load command signal acquired in real time, the compensation coefficients of the three characteristic sub-bands are calculated through the nonlinear gain scheduling function. The difference in pulse frequency response between the current operating model and the nominal model in each sub-band is weighted and summed using the compensation coefficients to construct a frequency domain decoupling operator.

[0044] In this embodiment, the load command signal of the nuclear power plant is acquired in real time, and the compensation coefficient of each sub-band under the current operating condition is dynamically calculated using a preset nonlinear gain scheduling function. This embodiment uses the following nonlinear scheduling formula: , in, As the reference load power, , and For the preset scheduling parameters corresponding to the feature sub-bands, the real-time compensation coefficients satisfy the normalization constraint: .

[0045] During variable load (transient maneuvering operation), the reactor power control system dynamically updates the compensation coefficients of each sub-band based on real-time power data. The compensation coefficient distribution dynamically calculated under three typical reference load points of 0.1, 0.5, and 0.9 in this embodiment is shown in Table 2. Table 2. Characteristic Subband Adaptive Compensation Coefficient Allocation Table under Typical Reference Load Point

[0046] Under varying load conditions, the normal maneuvering transients of the reactor power control system are often highly coupled with early degradation characteristics, making abnormal conditions extremely difficult to handle and distinguish. To verify the diagnostic sensitivity of our proposed method in addressing this challenge, simulations were conducted at load points of 0.1, 0.5, and 0.9, introducing three extremely low-amplitude, concealed abnormal conditions that caused steady-state power deviations of only 0.1, 0.01, and 0.001, respectively.

[0047] To visually demonstrate the concealment of extremely low amplitude anomalies in the system's time-domain signal. Figure 2 The system at the reference load point is given. Below are the transient response curves of the three core physical quantities (neutron density response, fuel temperature response, and coolant temperature response) over time. The figure compares these curves with the baseline normal operating condition without anomalies. (black solid line), and introduce three different amplitude minor anomalies respectively ( solid blue line; Orange dashed line; The system's time-domain output (with red dots) is shown. From the time-domain response trajectories of these three physical quantities, the following phenomena can be clearly observed: Visibility of significant anomalies: In a variable load transient process with an overall upward trend, only when the anomaly amplitude reaches 0.100 (red dotted line in the figure) does the output of various physical quantities show a significant deviation from the baseline normal trajectory that is visible to the naked eye.

[0048] Extreme concealment of weak anomalies: When the anomaly amplitude is further reduced to 0.010 (orange dashed line) or 0.001 (blue solid line), its corresponding output curve closely matches the baseline operating condition curve without anomalies (black solid line), almost completely overlapping.

[0049] Based on the above chart data, the following conclusions can be drawn: Figure 2 As shown, conventional time-domain monitoring methods (which rely on the time-domain residuals of the observed output signal) can only identify significant anomalies with an amplitude of 0.1. When the anomaly amplitude drops to 0.01 or 0.001, the anomaly signal is completely submerged in the normal transient fluctuations caused by the variable load, and conventional methods are difficult to distinguish, thus failing.

[0050] The load-driven frequency domain decoupling operator proposed in this invention is employed. This method detects anomalies by calculating the frequency domain distance between the current operating model and the nominal model of the reactor power control system. The calculation formula is as follows: , , in, This represents the pulse frequency response of the currently running model. The pulse frequency response of the nominal model. Represents the normalized frequency. Indicates frequency, It is the data sampling period. These are the compensation coefficients for each characteristic sub-band under the corresponding load. For the set of characteristic subbands, These correspond to the low-frequency, mid-frequency, and high-frequency characteristic sub-bands decomposed in S2, respectively. This represents the sub-band number, and here it is set to 3. Frequency domain decoupling operators under different operating conditions. The calculation results are shown in Table 3: Table 3. Calculation results of frequency domain decoupling operators under different load points and abnormal amplitudes.

[0051] As shown in Table 3, through frequency domain decoupling and spatial projection, the operator effectively penetrates the normal transient fluctuations of variable load and extremely sensitively captures the very low amplitude shift of the dynamic characteristics of nuclear power units caused by the hidden anomaly at the 0.001 level.

[0052] S4. Dynamically generate a dynamic envelope threshold based on the current load. If the frequency domain decoupling operator is greater than the dynamic envelope threshold, then determine that there is a transient anomaly in the nuclear power unit and issue an alarm.

[0053] In this embodiment, the envelope threshold is continuously and dynamically adjusted based on the real-time load. Historical distance data under normal variable load conditions is collected, and the mean and standard deviation within the sliding window are calculated. Principles of building based on load The dynamic confidence interval of the independent variable is used as the upper limit of the alarm threshold, and its threshold function is expressed as: , in, For load Click on the statistical mean of the frequency domain decoupling operator. Standard deviation, It is the confidence factor and has a value of 3.

[0054] In this embodiment, the dynamic envelope threshold characteristic parameters under different load points are shown in Table 4: Table 4. Mapping table of dynamic envelope threshold characteristic parameters under typical load points

[0055] As shown in Table 4, the dynamic envelope threshold setting is only positively correlated with the current absolute load level of the reactor. The physical mechanism is that as the load level increases, the primary loop thermal-hydraulic disturbance and nonlinear dynamics increase, thus the distance statistical mean and standard deviation both increase monotonically.

[0056] In actual industrial distributed control system (DCS) online monitoring, in order to avoid false alarms caused by step abrupt changes in the dynamic envelope threshold between discrete load nodes, this embodiment uses a one-dimensional cubic spline interpolation algorithm to smooth and continuous the discrete data points in Table 4, generating a safe envelope threshold that smoothly expands and contracts with the load.

[0057] Frequency domain decoupling operator calculated based on Table 3 Compared with the dynamically generated dynamic envelope threshold based on Table 4, such as Figure 3 It can be seen that even when faced with a concealment anomaly of 0.001, the frequency domain decoupling operator calculated in this embodiment... It still significantly exceeded the upper limit of the dynamic envelope threshold, successfully triggering an abnormal alarm signal. Furthermore, this frequency domain decoupling operator effectively penetrated normal variable load transient fluctuations and extremely sensitively captured the very low amplitude deviation of the reactor power control system dynamic characteristics caused by a 0.001-level hidden anomaly. This achieves a perfect balance between flexibly containing normal transient fluctuations and rigidly intercepting very low amplitude hidden anomalies under variable load conditions.

[0058] On the other hand, the embodiment also provides a nuclear power unit transient anomaly diagnosis device based on a load-driven frequency domain decoupling operator, including a memory and a processor. The memory is used to store a computer program, and the processor is used to implement the nuclear power unit transient anomaly diagnosis method based on a load-driven frequency domain decoupling operator when the computer program is executed.

[0059] It should be noted that the nuclear power unit transient anomaly diagnosis method and equipment based on load-driven frequency domain decoupling operators provided in the above embodiments should be illustrated using the above-described functional module division as an example when performing transient anomaly diagnosis of nuclear power units. The functions described above can be assigned to different functional modules as needed, that is, the internal structure of the terminal or server can be divided into different functional modules to complete all or part of the functions described above. Furthermore, the nuclear power unit transient anomaly diagnosis method based on load-driven frequency domain decoupling operators and the nuclear power unit transient anomaly diagnosis equipment embodiment based on load-driven frequency domain decoupling operators provided in the above embodiments belong to the same concept. For details of their specific implementation process, please refer to the embodiment of the nuclear power unit transient anomaly diagnosis method based on load-driven frequency domain decoupling operators, which will not be repeated here.

[0060] The embodiments described above provide a detailed explanation of the technical solutions and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, additions, and equivalent substitutions made within the scope of the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for diagnosing transient anomalies in nuclear power units based on load-driven frequency domain decoupling operators, characterized in that, Includes the following steps: Step 1: Establish a mechanism model based on the physical topology of the nuclear power unit in advance, transform it into a set of time-domain differential equations describing the dynamic characteristics of the nuclear power unit, and identify and obtain the nominal model for the transient maneuvering operation characteristics of the nuclear power unit. Step 2: Based on the current load, the full frequency domain space is adaptively decoupled into three characteristic sub-bands: low frequency, medium frequency, and high frequency, according to the nominal model. The sub-band boundaries dynamically expand and contract with the load. Step 3: Using the real-time acquired load command signal, calculate the compensation coefficients of the three characteristic sub-bands through the nonlinear gain scheduling function, and use the compensation coefficients to weight and sum the pulse frequency response differences between the current operating model and the nominal model in each sub-band to construct a frequency domain decoupling operator; Step 4: Dynamically generate a dynamic envelope threshold based on the current load. If the frequency domain decoupling operator is greater than the dynamic envelope threshold, it is determined that there is a transient anomaly in the nuclear power unit and an alarm is triggered.

2. The method for diagnosing transient anomalies in nuclear power units based on load-driven frequency domain decoupling operators according to claim 1, characterized in that, The nuclear power unit in step 1 is one or more of the following: nuclear power plant reactor power control system, coolant temperature control system, pressurizer pressure and water level control system, and steam generator water level control system.

3. The method for diagnosing transient anomalies in nuclear power units based on load-driven frequency domain decoupling operators according to claim 2, characterized in that, Step 1 specifically includes: Based on the physical topology of the reactor core of a nuclear power plant, a dynamic mechanism model is established that includes neutron dynamics of the reactor, fuel temperature effect, coolant temperature effect and the reactivity value of control rods. The dynamic mechanism model is transformed into a set of time-domain differential equations describing the dynamic characteristics of nuclear power units. After linearization, a state-space expression is obtained, and time-varying parameters that drift nonlinearly with the reactor power level are included to obtain the mechanism model. Under multiple stable loads, test excitation signals are injected into the mechanism model, dynamic input and output data are collected, and the discrete state-space model and transfer function corresponding to each load point are identified, which serve as the nominal model under this load point.

4. The method for diagnosing transient anomalies in nuclear power units based on load-driven frequency domain decoupling operators according to claim 3, characterized in that, The state variables in the state-space expression include: relative neutron density, fuel temperature deviation, and coolant temperature deviation; the input variables include control rod displacement. The time-varying parameters include: control rod value gain, fuel reactivity temperature coefficient, and coolant reactivity temperature coefficient.

5. The method for diagnosing transient anomalies in nuclear power units based on load-driven frequency domain decoupling operators according to claim 4, characterized in that, Step 2, which describes the adaptive decoupling of the full-frequency domain space into three characteristic sub-bands—low-frequency, mid-frequency, and high-frequency—based on the nominal model, includes: The sensitivity function and open-loop transfer function are calculated based on the transfer function in the nominal model. The sub-band boundary is determined based on the amplitude cross-frequency and phase cross-frequency of the sensitivity function and open-loop transfer function. Specifically, the frequency band where the absolute value of the sensitivity function exceeds the first threshold and includes the sensitivity peak is defined as the intermediate frequency (IF) feature segment; the low frequency (LFM) feature segment is the frequency range below the lower limit of the IF feature segment; and the high frequency (HF) feature segment is the frequency range above the IF feature segment.

6. The method for diagnosing transient anomalies in nuclear power units based on load-driven frequency domain decoupling operators according to claim 1, characterized in that, The nonlinear gain scheduling function in step 3 uses the following formula: ， in, As the reference load power, , and The real-time compensation coefficients satisfy normalization constraints and are preset scheduling parameters for the corresponding feature sub-bands. During transient maneuvering, the nuclear power unit updates its compensation coefficients based on real-time load command signals. This is to adaptively absorb the frequency domain characteristic shift caused by the nonlinearity of nuclear power units.

7. The method for diagnosing transient anomalies in nuclear power units based on load-driven frequency domain decoupling operators according to claim 1, characterized in that, The calculation formula for the frequency domain decoupling operator is as follows: ， ， in, This represents the pulse frequency response of the currently running model. The pulse frequency response of the nominal model. Represents the imaginary unit and satisfies , Represents the normalized frequency. Indicates frequency, It is the data sampling period. For the set of characteristic subbands, These are the compensation coefficients for each characteristic sub-band under the corresponding load.

8. The method for diagnosing transient anomalies in nuclear power units based on load-driven frequency domain decoupling operators according to claim 1, characterized in that, The method of dynamically generating a dynamic envelope threshold based on the current load includes: pre-setting threshold parameters at multiple discrete load points, wherein the threshold parameters include the mean and standard deviation and are positively correlated with the load level; and then generating a dynamic envelope threshold that changes continuously and smoothly with the load through interpolation.

9. A transient anomaly diagnosis device for nuclear power units based on a load-driven frequency domain decoupling operator, comprising a memory and a processor, wherein the memory is used to store a computer program, characterized in that, The processor is used to implement the nuclear power unit transient anomaly diagnosis method based on load-driven frequency domain decoupling operator as described in any one of claims 1-8 when executing the computer program.

Images