An adaptive intelligent control method for high-voltage fuses

Abstract

The application relates to the technical field of direct-current power distribution protection, and discloses a high-voltage fuse self-adaptive intelligent control method, which comprises the following steps: collecting main branch operation current and thermistor measurement temperature, setting an observation reference and calculating a current change rate; substituting into an extended Kalman filter for error convergence, and calculating optimal equivalent temperature of a fusible component; calculating a dynamic action integral threshold according to the optimal equivalent temperature and the current change rate, accumulating a short-time Joule integral, and determining that a thermal collapse critical area is entered when the integral is out of limit; performing logical operation on the first-order and second-order derivatives of the optimal equivalent temperature, and identifying a phase change latent heat starting moment; calculating a synchronous triggering moment according to the starting moment, a physical melting enthalpy value and a system advance angle, and controlling a separation device to be tripped and an IGBT module of a commutation branch to be synchronously conducted for commutation when the moment is reached. The application effectively reduces the misoperation and refusal rates under complex working conditions, and realizes flexible breaking without electric arc.

Description

Technical Field

[0001] This invention relates to the field of DC power distribution protection technology, specifically to an adaptive intelligent control method for high-voltage fuses. Background Technology

[0002] With the development of high-voltage direct current (HVDC) distribution networks, higher requirements have been placed on the rapid and reliable interruption of DC fault currents. Existing high-voltage protection devices typically employ a hybrid topology consisting of a main branch composed of fusible components and mechanical disconnect devices, and a converter branch composed of power electronic devices connected in parallel. In terms of protection control logic, traditional control methods often use fixed current amplitudes or static Joule integral values ​​as the action threshold for judging system faults.

[0003] This control method, which uses a single static parameter, has significant adaptability limitations. In actual operation, power distribution networks face complex load disturbances. For example, the starting of large motors or the connection of capacitive loads generates conventional transient surge currents. Fixed operating thresholds are prone to integral over-limit under such normal operating conditions, leading to malfunctions of protection devices. Furthermore, in the event of a high-impedance grounding fault, the fault current rises slowly. Fusible components, due to prolonged current flow, are already in a state of thermal fatigue or extremely high temperature criticality. However, because the static integral threshold cannot detect the heat accumulation at the material's base, the control system response is lagging, easily leading to protection failure or equipment overheating and damage. Limited by high-voltage insulation spacing and strong electromagnetic interference, the control system finds it difficult to directly obtain the true internal temperature of the high-voltage components using contact sensors, resulting in a lack of data support for dynamic correction of the operating threshold.

[0004] Furthermore, during fault disconnection, existing control logic often relies on fixed timer programs to coordinate the timing of the main branch's mechanical tripping and the converter branch's conduction. Because the melting phase transition time of fusible components varies significantly under different short-circuit currents, fixed delay triggering mechanisms cannot accurately match the actual moment when the material enters the latent heat period of the phase transition and generates high impedance. This timing deviation leads to the operating current not smoothly transferring to the converter branch, causing the mechanical disconnect device to generate an arc when tripped under load, reducing the electrical life of the contacts. Moreover, after the converter branch completes the current transfer, existing controllers typically use hard shutdown commands to directly disconnect the power electronic devices. The residual inductance energy in the line, under extremely high current change rates, can trigger severe line overvoltages, threatening the system's insulation safety. Summary of the Invention

[0005] The technical problem solved by this invention is that existing high-voltage DC protection devices usually use a fixed static Joule integral threshold as the action criterion, which fails to dynamically adjust based on ambient temperature and internal heat accumulation. This makes them prone to false tripping under transient surge conditions and failure to trip under high impedance fault conditions. At the same time, the high current breaking process relies on a single mechanical contact, which is prone to arcing and has low breaking reliability.

[0006] To address the above problems, the present invention provides the following technical solution: This invention provides an adaptive intelligent control method for high-voltage fuses, applied to a control system. The control system includes a processor and a main branch and a converter branch connected in parallel. The main branch includes a fusible component and a separation device connected in series. The converter branch includes an IGBT module and a varistor. The IGBT module contains a thermistor. The method is executed by the processor and includes: The operating current of the main branch and the measured temperature of the thermistor are collected. The measured temperature is used as a common heat sink observation benchmark, and the current change rate is calculated based on the operating current. The heat power generated by the operating current is used as the input variable, and the common heat sink observation reference is used as the observation variable. The extended Kalman filter is substituted to perform error convergence and the optimal equivalent temperature of the fusible component is calculated. The dynamic action integral threshold is calculated based on the optimal equivalent temperature and the current change rate. The discrete short-time Joule integral is accumulated. When it is greater than or equal to the dynamic action integral threshold, the main branch is determined to have entered the thermal collapse critical region. After entering the thermal collapse critical region, the first and second derivatives of the optimal equivalent temperature are obtained. Logical operations are performed in combination with the set critical temperature threshold and the derivative states of the two to identify the start time of the latent heat of phase change of the fusible component. The synchronization trigger time is calculated based on the start time, physical melting enthalpy, and system advance angle. When the synchronization trigger time is reached, the separation device is controlled to trip, and a conduction command is simultaneously output to the IGBT module for switching.

[0007] Further, the process of using the heat power generated by the operating current as an input variable, substituting the common heat sink observation reference as an observation variable into the extended Kalman filter for error convergence, and calculating the optimal equivalent temperature of the fusible component includes: The resistance-temperature polynomial coefficient table stored in the register is called to map the equivalent temperature of the previous discrete moment to the corresponding dynamic equivalent resistance value. The square of the operating current of the previous discrete moment is multiplied by the dynamic equivalent resistance value to obtain the input heat power injected into the fusible component at the previous discrete moment. By using the state transition matrix, the input control matrix, the input thermal power, and the optimal posterior estimated state vector of the previous discrete time step, advance prediction is performed to obtain the prior state vector and the prior error covariance matrix. Construct an observation matrix, and calculate the Kalman gain matrix based on the prior error covariance matrix, the observation matrix, and the preset observation noise covariance matrix; The innovation deviation between the actual collected common heat sink observation benchmark and the theoretical observation value is calculated. The Kalman gain matrix is ​​used to weight the innovation deviation and superimpose it onto the prior state vector. The first element in the corrected optimal posterior state vector is extracted and assigned as the optimal equivalent temperature.

[0008] Further, the process of calculating the dynamic action integral threshold based on the optimal equivalent temperature and the current change rate includes: Call the static Joule integral threshold, physical melting point, environmental reference temperature, and thermal fatigue compensation coefficient pre-stored in the internal non-volatile memory; Calculate the difference between the optimal equivalent temperature and the environmental reference temperature, and calculate the inherent difference between the physical melting point and the environmental reference temperature. Divide these two differences to obtain the normalized temperature evolution rate of the current physical state. Multiply the normalized temperature evolution rate by the thermal fatigue compensation coefficient, and subtract the product from the constant 1 to obtain the thermal boundary compensation term; Execute the transient current suppression function and generate a dynamic adjustment coefficient based on the current change rate; The static Joule integral threshold, the thermal boundary compensation term, and the dynamic adjustment coefficient are multiplied and superimposed. When the calculation result is negative or has an unreasonable minimum value, the lower limit threshold is applied to force the output of the preset safety action baseline value as the dynamic action integral threshold at the current discrete moment.

[0009] Furthermore, the process of executing the transient current suppression function and generating a dynamic adjustment coefficient based on the current change rate includes: The real-time current change rate is compared with the set surge lower limit slope value and short-circuit fault upper limit slope value. When the rate of change of current is within the range of the lower limit slope value of surge and the upper limit slope value of short circuit fault, it indicates that the main branch is experiencing a non-faulty transient large load pulling process. The value of the transient current suppression function is assigned a preset gain coefficient, and the resulting calculation result constitutes the dynamic adjustment coefficient for the current period. When the rate of change of current is outside the above range, the value of the function is fixed as a constant of 1, and the resulting calculation result constitutes the dynamic adjustment coefficient for the current period.

[0010] Furthermore, the process of accumulating discrete short-time Joule integrals includes: The square of the discrete sampled value of the running current is calculated, and the square is multiplied by the step size of the control cycle to quantify the infinitesimal energy injected into the main branch in the current single control cycle. The infinitesimal energy is added to the integral history value stored in the register at the previous discrete moment to obtain the discrete short-time Joule integral at the current discrete moment. When the discrete sampled value of the operating current is detected to fall below the system's rated operating current safety threshold, and this state is maintained for a period of time exceeding the preset reset time window, the discrete short-time Joule integral register is cleared.

[0011] Furthermore, the process of obtaining the first and second derivatives of the optimal equivalent temperature includes: A buffer area is allocated in the internal memory, a historical temperature sliding data window based on a first-in-first-out queue is constructed, and the latest acquired optimal equivalent temperature is written to the head of the queue. Call the first-order convolution coefficient matrix and the second-order convolution coefficient matrix that are pre-stored in the read-only memory area; Each discrete optimal equivalent temperature within the historical temperature sliding data window is multiplied one by one with its corresponding first-order and second-order convolution coefficients in the data window, and the results are summed to obtain the first-order derivative and the second-order derivative that contain the physical evolution trend.

[0012] Furthermore, the process of performing logical operations based on the set critical temperature threshold and the derivative states of both to identify the onset time of the latent heat of phase change in the fusible component includes: Extract the optimal equivalent temperature calculated at the current discrete moment and compare it with the set critical temperature threshold. When the optimal equivalent temperature is greater than or equal to the critical temperature threshold, it is confirmed that the current state satisfies the spatial absolute value condition. Take the absolute value of the obtained first derivative. When the absolute value is less than or equal to the set dead zone where the first derivative approaches zero, it is confirmed that the time-domain stationarity condition is satisfied. By comparing the second derivative states of two adjacent discrete time points, if the second derivative of the previous discrete time point is strictly greater than zero and the second derivative of the current discrete time point is less than or equal to zero, the inflection point evolution condition is confirmed to be satisfied. The spatial absolute value condition, the temporal stationarity condition, and the inflection point evolution condition are subjected to a logical AND operation. When all three conditions are met, the current discrete time stamp inside the control system is locked as the start time of the latent heat of phase transition, and the timing lock flag in the system memory is set synchronously.

[0013] Furthermore, the process of calculating the synchronization triggering time based on the start time, the physical enthalpy of melting, and the system advance angle includes: By substituting the physical melting point into the polynomial calculation model, the fixed equivalent resistance value of the melting point corresponding to the critical state of phase transformation of the material can be obtained. The square of the discrete sampled value of the operating current is calculated, and the square value is multiplied by the equivalent resistance value of the melting point to calculate the instantaneous active heat power of the phase change. When the instantaneous active heat power of the phase change is lower than the set small bias constant, the instantaneous active heat power of the phase change is forcibly assigned to the bias constant. The theoretical latent heat time window is estimated by dividing the physical enthalpy of melting, which includes effective phase change mass parameters, by the instantaneous active heat power of the phase change. The thermal collapse limit time is obtained by adding the start time of the phase change latent heat to the theoretical latent heat time window. The synchronization trigger time is calculated by subtracting the system advance angle from the thermal collapse limit time.

[0014] Furthermore, the system advance angle is composed of the superposition of the processor's instruction transmission bus delay time, the IGBT module's drive circuit response time, and the dead time of the mechanical mechanism of the separation device; When the calculated synchronization trigger time is less than or equal to the current absolute running time inside the control system, the current absolute running time is forcibly assigned to the synchronization trigger time.

[0015] Furthermore, after the separation device is tripped and a turn-on command is simultaneously output to the IGBT module for commutation, the method further includes: Start the internal insulation build-up delay timer; When the insulation establishment delay timer reaches the preset duration, it indicates that the main branch has the isolation capability to withstand the system recovery voltage. It outputs a stepped-down signal to the drive terminal of the IGBT module or controls the drive circuit to connect to the preset turn-off resistor, thereby controlling the IGBT module to execute the flexible turn-off logic. When the voltage reaches the clamping voltage threshold of the varistor, the parallel varistor becomes conductive, converting the residual electromagnetic energy of the system into heat energy for dissipation.

[0016] This invention provides an adaptive intelligent control method for high-voltage fuses. It has the following beneficial effects: 1. This invention calculates the optimal equivalent temperature of the fusible component by using an extended Kalman filter and reconstructs the dynamic action integral threshold by combining the rate of change of the operating current. This method allows the protection action threshold to be adjusted according to the actual operating conditions of the system. When a transient large load surge occurs, the threshold is increased to avoid false tripping, and when a high impedance fault causes internal heat accumulation, the threshold is decreased to avoid failure to trip, thereby improving the reliability of the system protection action.

[0017] 2. This invention extracts the first and second derivatives of the optimal equivalent temperature to identify the starting moment of the latent heat of phase change in the fusible component, and calculates the synchronous triggering time by combining the physical melting enthalpy and the system advance angle. This control logic transforms the triggering basis into physical phase change state monitoring, realizing the timing coordination of the tripping of the main branch separation device and the conduction of the IGBT in the converter branch. It utilizes the impedance rise characteristic in the early stage of the phase change of the fusible component to complete the current transfer and suppress the arc generated during the disconnection tripping.

[0018] 3. After the commutation process, the present invention uses an insulation delay timer to confirm the mechanical isolation state of the separation device. By outputting a stepped voltage reduction signal to the drive end or connecting a turn-off resistor, the IGBT module is controlled to perform flexible turn-off. This step reduces the transient rate of change of current and, combined with the conduction clamping characteristics of the varistor, dissipates residual electromagnetic energy, suppressing the parasitic overvoltage of the line caused by the sudden interruption of current in the high voltage DC network, thus ensuring the safety of the system during the disconnection phase. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the high-voltage DC protection topology of the present invention; Figure 2 The flowchart is as follows: This invention provides an adaptive intelligent control method for high-voltage fuses. Figure 3 This is a comparison diagram of the adaptive integration threshold and actual energy evolution of the present invention; Figure 4 This is a diagram showing the spatiotemporal feature-based latent heat identification and feedforward converter calibration of the present invention. Detailed Implementation

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] See attached document Figure 1 The present invention provides an adaptive intelligent control method for high-voltage fuses, wherein the hardware basis for executing the method may include a control system.

[0022] The control system includes a main branch and a converter branch connected in parallel with the main branch; The main branch and the converter branch are connected via a busbar; The main branch includes a series-connected fusible component and a separation device; The commutation branch includes parallel IGBT modules and varistors; The IGBT module has an internal thermistor; The control system also includes a processor and a converter connected to the processor; The processor is connected to both the control terminal of the separation device and the drive terminal of the IGBT module.

[0023] See attached document Figure 2 This method is executed by the processor and specifically includes the following steps: S100 synchronously acquires the operating current of the main branch and the measured temperature of the thermistor through the converter, sets the measured temperature as the common heat sink observation benchmark, and calculates the current change rate based on the operating current.

[0024] S200: Establish a discretized thermal network model and construct an extended Kalman filter. Use the heat power generated by the operating current as the input variable and the common heat sink observation benchmark as the observation variable to enter the extended Kalman filter for error convergence, and calculate the optimal equivalent temperature of the fusible component.

[0025] S300 calculates the action integral threshold based on the optimal equivalent temperature and current change rate. The processor accumulates the short-time Joule integral, and determines that the main branch has entered the critical region when the short-time Joule integral is greater than or equal to the action integral threshold.

[0026] S400, after determining that the system has entered the critical region, establishes a historical temperature sliding data window. A smoothing polynomial differential operator is used to perform convolution operations on the data within the historical temperature sliding data window to obtain the first and second derivatives of the optimal equivalent temperature. Based on the logical operation results of the critical temperature threshold, the first derivative state, and the second derivative state, the starting moment of the latent heat of phase change in the fusible component is identified.

[0027] S500 calculates the synchronization trigger time based on the start time of the latent heat of phase change, the physical enthalpy of fusion of the fusible component, and the system advance angle. When the running time reaches the synchronization trigger time, the processor outputs a command to control the separation device to trip, and synchronously outputs a gate turn-on command to the IGBT module to perform commutation.

[0028] In this embodiment, the sampling process of electrical parameters in step S100 may specifically include the following steps: S101, the processor outputs a trigger signal according to the set control cycle, which controls the converter to discretize the operating current of the main branch.

[0029] As an optional implementation, the acquisition of the operating current relies on the underlying sensing hardware of the control system. The control system also includes current sensors installed on the main branch. Specifically, the current sensors are closed-loop Hall effect current sensors or alloy shunts. The converter is specifically an analog-to-digital converter. The converter receives the analog voltage signal output from the current sensor and converts it into a digital discrete current sequence, which is then input to the processor. This hardware configuration can be used to effectively monitor the operating current of the high-voltage direct current system.

[0030] For the design of the pre-amplifier signal conditioning and sample-and-hold circuit for the analog-to-digital converter, those skilled in the art can set up the corresponding anti-aliasing filter circuit according to the requirements of sampling accuracy. The hardware circuit design is a well-known technology in this field and will not be described in detail here.

[0031] To map continuous-time physical signals to digital systems, the step size of the control cycle is denoted as in the digital signal processing logic. The control cycle step size The value of is determined in principle based on the Nyquist sampling theorem and the highest switching frequency of the insulated-gate bipolar transistor inside the control system, and is usually set in the microsecond range to ensure that transient fault characteristics can be captured. The current discrete time is denoted as . The processor at the current discrete moment The discrete sampled values ​​of the operating current of the main branch are denoted as follows: .

[0032] S102, the processor obtains the current sample value at the current discrete moment and retrieves the current sample value at the previous discrete moment from the internal register. Based on this, the processor obtains the current change rate at the current discrete moment through first-order difference operation.

[0033] The rate of change of current is used to characterize the instantaneous fluctuation characteristics of the load current in the main branch. From a physical perspective, although the current amplitude in the early stages of a high-impedance fault has not yet reached the absolute action threshold, its rate of increase is already different from normal operating conditions. By tracking this rate of change in real time, a reference for the state evolution can be provided before the fault worsens. In the specific calculation and execution process, the processor divides the difference between the current sampled value at the current discrete moment and the previous discrete moment's current sampled value by the step size of the control cycle to obtain the current rate of change at the current discrete moment. Specifically, when the system starts and enters the initial sampling cycle, since the internal registers have not yet been written with valid historical sampling data, the processor pre-assigns the reference sampling data corresponding to that historical moment to zero or the current steady-state bias current value, thereby avoiding logical dead zones such as data overflow or calculation anomalies in the early stages of the digital differential algorithm.

[0034] The processor stores the calculated rate of change of current in its internal memory. This value provides the basis for subsequent algorithms to determine whether the system is under normal load fluctuation conditions or short-circuit fault conditions. Through the above-mentioned discrete sampling and differential operation mechanism, the processor can acquire electrical parameters that characterize the dynamic evolution trend of the system in real time.

[0035] In this embodiment, the extraction process of the common reference thermal boundary conditions in step S100 may specifically include the following steps: S103, the processor acquires the measurement signal of the thermistor through the converter and converts the measurement signal into corresponding discrete temperature data.

[0036] In one possible implementation, the thermistor is integrated into the internal base plate area of ​​the IGBT module. In high-voltage DC operating environments, due to insulation spacing limitations and electromagnetic interference, directly installing a temperature sensor on the fusible component of the main branch presents engineering challenges. This embodiment achieves multiplexing of system temperature parameters by reading the signals from the original sensing elements of the IGBT module in the converter branch.

[0037] For the design of the bias conditioning circuit of the thermistor and the nonlinear function mapping of resistance to temperature, those skilled in the art can perform table lookup calculations or polynomial fitting based on the material parameters of the selected thermistor. The hardware conditioning and digital conversion logic are well-known technologies in the field and will not be described in detail here.

[0038] To prevent subsequent thermal network model calculation errors caused by thermistor disconnection or short circuit, the processor performs a reasonable range check on the acquired discrete temperature data after conversion. If the current discrete temperature data is detected to exceed the preset physical ambient temperature or operating limit range, the processor will intercept the abnormal data and automatically call the fixed safe ambient temperature value pre-stored in the register as a replacement, thereby preventing the control algorithm from falling into a logic dead zone due to a single point of failure of the sensor.

[0039] S104, the processor establishes a spatial heat conduction mapping relationship and sets the converted discrete temperature data as the common heat sink observation reference for the control system.

[0040] At the hardware physical topology level, the main branch and the commutator branch are electrically connected through a bus. The bus not only carries current but also forms a heat conduction path. The Joule heat generated by the fusible components and the heat loss generated by the IGBT modules are both conducted to the bus. Based on this physical structural characteristic, the bus can be considered, thermodynamically, as a common heat sink shared by related heat-generating components within the control system.

[0041] Because of the large size and heat capacity of the busbar, its condition has a significant impact on the heat dissipation process of the fusible component to the outside. The thermistor is placed close to the connection interface between the IGBT module and the busbar, and the temperature value it reflects can reflect the ambient background temperature of the area.

[0042] Considering the spatial distance between the physical mounting location of the thermistor and the heat dissipation end of the fusible component, the heat conduction path inevitably includes structural thermal resistance generated by the assembly contact surface and the metal material. When mapping discrete temperature data to a common heat sink observation benchmark, the processor introduces an additive correction based on a static deviation compensation value obtained from calibration experiments to offset the measurement error caused by the spatial temperature gradient.

[0043] In the control program, the processor will use the current discrete time... The acquired discrete temperature data, after undergoing the aforementioned deviation compensation correction, is assigned as the common heat sink observation benchmark, denoted as... This parameter serves as the foundational input for subsequent iterations of the discretized thermal network model. By constructing the aforementioned cross-branch parameter mapping mechanism, the processor can acquire the boundary temperature conditions of the fusible component without relying on additional external high-pressure isolation sensors, providing a data basis for overcoming control parameter drift caused by environmental fluctuations.

[0044] In this embodiment, the process of electrothermal coupling network model and prior deduction in step S200 may specifically include the following steps: S201, the processor constructs a discretized thermal network model based on a multi-order RC network topology in its internal memory and defines the system's state vector.

[0045] Under high-voltage operating conditions, the internal temperature of fusible components is difficult to measure physically using contact sensors. This technical solution employs equivalent thermal resistance and heat capacity parameters to construct a multi-order system of thermal conduction differential equations. To obtain the intrinsic thermal resistance and heat capacity coefficients required for the equations, during the system calibration phase before shipment, the thermal conductivity and specific heat capacity parameters of each physical level of the fusible component are extracted by combining a finite element thermal field simulation platform with laboratory infrared thermal imaging experiments. These parameters are then stored as matrix coefficients in the processor's non-volatile memory.

[0046] To perform iterative execution of continuous thermodynamic physical processes within a digital control system, the processor, based on the discrete step size set by the control system, transforms the continuous-time heat conduction equations into a discretized heat network model. Let the state vector of the discretized system be... , The spatial partitioning order of this discretized thermal network model is defined. The state vector contains the calculated temperature values ​​corresponding to each discrete node from the core region inside the fusible component to the external heat dissipation boundary. To facilitate the subsequent calculation and determination of the protection threshold, the processor defines the first term of the state vector as the core equivalent temperature of the fusible component, denoted as . That is, satisfying mathematical relations .

[0047] S202, the processor calculates the thermal power input variable for the current discrete cycle based on historical state data.

[0048] The temperature rise of fusible components under current-carrying conditions mainly originates from the Joule heating effect. Considering that the resistivity of metallic materials is not a constant but increases non-linearly with increasing temperature, the processor incorporates temperature correction compensation logic when calculating the heat generation power.

[0049] At the initial moment of program execution The processor pre-sets the temperature of each node in the state vector to the currently measured ambient bias temperature, thus avoiding logic crashes caused by missing register data in the early stages of computation. During a normal loop control cycle, the processor calls the pre-stored resistance-temperature polynomial coefficient table in the registers. The processor retrieves the previous discrete time step. Operating current and the known core equivalent temperature During the actual calculation execution process, the processor calculates the operating current. The square value, and the core equivalent temperature according to the polynomial coefficient table. This is mapped to the corresponding dynamic equivalent resistance value. Subsequently, the processor multiplies the squared current value by the dynamic equivalent resistance value to obtain the input heat power injected into the fusible component at the previous discrete time step. This calculation step can reduce the power estimation error caused by using a fixed resistor.

[0050] The S203 processor, based on the constructed discretized thermal network model and the extended Kalman filter algorithm, calculates the prior state vector and prior error covariance matrix at the current discrete moment.

[0051] The processor uses the optimal posterior estimate of the state vector from the previous time step. Input thermal power In addition, a one-step advance prediction is performed using a pre-set system matrix. Specifically, the processor extracts the state transition matrix to characterize the heat dissipation coupling relationship of each node in the thermal network. and the input control matrix used to characterize the thermal power excitation effect. The processor will use the state transition matrix. With the optimal posterior estimated state vector Perform matrix multiplication while inputting the control matrix. With input thermal power Multiply. By adding the two products together, the processor obtains the current discrete time. Prior state vector .

[0052] To assess the uncertainty of the advance prediction results, the processor synchronously deduces and updates the prior error covariance matrix. Mathematically, the processor performs state transition matrix... The posterior error covariance matrix of the previous discrete time step and the transpose of the state transition matrix The three components are then subjected to matrix multiplication in sequence, and the result is compared with the preset process noise covariance matrix. Add them together to obtain the current discrete time. Prior error covariance matrix .

[0053] When the control system is first started, the processor initializes the initial error covariance matrix as a square matrix with non-zero diagonal elements. The process noise covariance matrix is ​​set to compensate for inherent calculation biases caused by model simplification and unmodeled disturbances. The above derivation and calculation provide a benchmark for subsequent fusion of observation data.

[0054] In this embodiment, the process of optimal convergence of the posterior state based on the observation benchmark in step S200 may specifically include the following steps: S204, the processor introduces the acquired common heat sink observation benchmark into the filtering algorithm and establishes a mapping matrix between the observed variable and the internal nodes of the discretized thermal network model.

[0055] After completing the prior prediction simulation, the control system faces the problem of accumulated errors easily generated by the open-loop integration of the model. To suppress this parameter drift, the processor calls the common heat sink observation reference obtained in step S100. This is used as the observation input for the extended Kalman filter algorithm.

[0056] In order for this external physical observation to participate in internal matrix operations, the processor needs to construct the observation matrix. At the physical structure level, the heat generated inside the fusible component is ultimately conducted to the busbar, which corresponds to the outermost heat dissipation boundary node of the discretized thermal network model. Based on this, the processor maps the elements representing the temperature of this boundary node in the state vector to theoretical observation terms, and then applies these observations to the observation matrix. The corresponding element position is set to a non-zero correlation coefficient. By constructing this mapping mechanism, the processor establishes a mathematical correlation between internal node temperature parameters and external measurable parameters.

[0057] S205, the processor calculates the Kalman gain at the current discrete time based on the prior error covariance matrix, the observation matrix, and the set observation noise variance.

[0058] Kalman gain is used to adjust the weighting of model prediction data and sensor measurement data in the final result. In the specific computational execution logic, the processor extracts the prior error covariance matrix obtained from the previous calculation step. The processor will observe the matrix. Prior error covariance matrix and the transpose of the observation matrix The three are multiplied together.

[0059] To accommodate the inherent measurement errors in the underlying sensing hardware, the processor incorporates a pre-calibrated observation noise covariance matrix into the product result. As an optional implementation, the observation noise covariance matrix The constant value can be determined based on the measurement accuracy tolerance provided in the thermistor's manufacturer's datasheet, or by continuously sampling the thermistor under steady-state temperature conditions and calculating the variance. The processor inverts the summed matrix. Subsequently, the processor calculates the prior error covariance matrix. The transpose of the observation matrix Multiplying the resulting inverse matrix in turn yields the Kalman gain matrix for the current discrete period. .

[0060] S206, the processor uses Kalman gain to perform closed-loop feedback correction on the prior state vector, and obtains the optimal posterior state vector with error convergence.

[0061] After obtaining the Kalman gain matrix, the processor proceeds to the error correction stage. The processor calculates the theoretical observations, i.e., the observation matrix. With prior state vector Multiplication. Further, the processor calculates the observational information bias, using the actual acquired common heat sink observation benchmark. Subtract the above theoretical observations.

[0062] After obtaining the innovation deviation, the processor utilizes the Kalman gain matrix. The deviation is weighted by matrix multiplication. The processor then adds the weighted correction to the prior state vector. Above. This additive operation introduces actually measured physical parameters to compensate for deviations caused by model extrapolation, thereby outputting the corrected optimal posterior state vector. .

[0063] S207, the processor extracts the corresponding eigenvalues ​​from the optimal posterior state vector as the optimal equivalent temperature, and simultaneously updates the posterior error covariance matrix.

[0064] After completing the closed-loop correction calculation, the processor extracts the optimal posterior state vector according to the state vector definition rules set in S201. The first element in the equation. This element, in a physical sense, corresponds to the estimated temperature of the core heating zone of the fusible component at the current discrete moment. The processor assigns it the optimal equivalent temperature of the fusible component. The result is stored in a register for subsequent action determination modules to use.

[0065] As the closing step of the current algorithm cycle, the processor updates the error covariance matrix for use in the next control cycle iteration. The specific calculation logic is as follows: the processor subtracts the Kalman gain matrix from the identity matrix. With observation matrix The product of these is then combined with the resulting difference matrix and the prior error covariance matrix. Multiplying them yields the posterior error covariance matrix at the current discrete time step. .

[0066] By performing the above steps, the processor achieves dynamic tracking and data output of the internal thermal state of the fusible component.

[0067] In this embodiment, the dynamic reconstruction process of the action threshold with dual-parameter constraints in step S300 may specifically include the following steps: S301, the processor calls the static Joule integral threshold, physical melting point, ambient reference temperature, and thermal fatigue compensation coefficient pre-stored in the internal non-volatile memory.

[0068] In traditional protection logic, the circuit breaker criterion typically relies on a single static parameter, which is difficult to adapt to the characteristic drift caused by long-term equipment operation. As an alternative implementation, this solution introduces multi-dimensional compensation parameters into the processor's control architecture. The static Joule integral threshold refers to the rated operating data of the fusible component calibrated under standard laboratory ambient temperature and in a brand-new, unaged state. The ambient reference temperature corresponds to the initial background temperature of the aforementioned calibration experiment. The physical melting point is determined by the crystalline properties of the metallic material of the fusible component. To quantify the degree of material aging, the processor records the historical overload counts and cumulative current-carrying operation time of the control system and obtains the corresponding thermal fatigue compensation coefficient by consulting a preset lifespan conversion experience table. In actual engineering settings, to prevent overcompensation from causing protection logic failure, the value range of this thermal fatigue compensation coefficient is usually calibrated and constrained to be between 0 and 0.2. The above basic parameters provide underlying constant support for building a dynamic adaptive model.

[0069] S302, the processor combines the obtained optimal equivalent temperature with the above-mentioned parameters to calculate the thermal boundary compensation term characterizing the temperature evolution margin.

[0070] In each control cycle, the processor receives the optimal equivalent temperature output from step S200. In the specific mathematical processing logic, the processor calculates the difference between the optimal equivalent temperature and the ambient reference temperature, and also calculates the inherent difference between the physical melting point and the ambient reference temperature. The processor divides these two differences to obtain the normalized temperature evolution rate of the current physical state. Based on this, the processor multiplies the normalized temperature evolution rate by the thermal fatigue compensation coefficient, and subtracts this product from the constant 1 to obtain the thermal boundary compensation term. This compensation term, at the physical level, reflects the remaining heat storage space between the current actual temperature of the fusible component and its physical melting point.

[0071] S303, the processor generates a dynamic adjustment coefficient by executing a transient current suppression function based on the acquired current change rate.

[0072] In the actual operation of power distribution systems, operations such as starting a motor or connecting a large capacitive load will generate conventional transient surge currents. To prevent the protection device from tripping erroneously under such normal load fluctuation conditions, the processor extracts the current change rate calculated in step S100 and inputs it into a preset transient current suppression function.

[0073] In the specific program implementation, the transient current suppression function is set to a specific piecewise logic structure. The processor compares the real-time current change rate with the set surge lower limit slope value and short-circuit fault upper limit slope value. These two slope boundary values ​​are calibrated based on the maximum starting inrush current characteristics of the controlled load under normal operating conditions. When the current change rate is within the range of the surge lower limit slope value and the short-circuit fault upper limit slope value, it indicates that the main branch is experiencing a non-fault-related transient large load pulling process. Under this condition, the processor assigns a preset gain coefficient to the value of the transient current suppression function, which is usually set between 1.1 and 1.3, to moderately relax the energy integration threshold for a short period of time. When the current change rate is outside the above range, i.e., when the system is operating smoothly or exhibits severe short-circuit fault characteristics, the processor fixes the value of the function to a constant 1. The resulting calculation results constitute the dynamic adjustment coefficient for the current cycle.

[0074] S304 The processor jointly calculates the static Joule integral threshold, the thermal boundary compensation term, and the dynamic adjustment coefficient to generate the dynamic action integral threshold for the current discrete moment.

[0075] As the core control rule for threshold adaptive generation in this embodiment, the processor performs multiplication and superposition calculations on the three parameters obtained from the aforementioned calculations. The specific mathematical expression is as follows: ; In the formula, For the current discrete time The dynamic action integral threshold; The static Joule integral threshold; The optimal equivalent temperature; The ambient reference temperature; Physical melting point; This is the thermal fatigue compensation coefficient; As for the rate of change of current The dynamic adjustment coefficient is the output of the transient current suppression function, which is the independent variable.

[0076] To ensure the rigor of the underlying control logic and prevent algebraic calculation overflow, the processor performs a lower threshold limiting before outputting the above product result. Under extreme high-temperature overload conditions, if the optimal equivalent temperature... Approaching or even temporarily exceeding the physical melting point If the product calculation result is negative or an unreasonable minimum value, the processor will force the output of the preset safety action baseline value as the dynamic action integration threshold at the current discrete moment to avoid the comparator getting stuck in the calculation dead zone.

[0077] Through the above logical processing, the generation of the protection device's operating threshold is no longer dependent on a single static parameter. Its calculation output can be dynamically corrected based on the accumulated aging state inside the fusible component and the transient evolution trend of the external load, providing a reference for the subsequent system to identify real faults under complex operating conditions.

[0078] In this embodiment, the energy accumulation and triggering process for the critical thermal collapse state in step S300 may specifically include the following steps: S305: The processor acquires the discrete sampled value of the operating current at the current discrete moment, and calculates the discrete short-time Joule integral at the current discrete moment by iteratively accumulating the step size of the control cycle.

[0079] After obtaining the dynamic action integral threshold, the control system needs to quantitatively evaluate the actual energy impact borne by the main branch. The processor acquires the discrete sampled value of the operating current collected in step S100. In the specific calculation execution, the processor calculates the square of the discrete sampled value of the operating current and multiplies the squared value by the step size of the control cycle to quantify the infinitesimal energy injected into the main branch in the current single control cycle. Subsequently, the processor adds this infinitesimal energy to the integral history value stored in the register at the previous discrete moment to obtain the discrete short-time Joule integral at the current discrete moment.

[0080] As an optional implementation, to prevent register data overflow due to the continuous accumulation of minute amounts of energy under long-term normal load conditions, an integral clearing mechanism is incorporated into the processor's control logic. When the processor detects that the discrete sampled value of the operating current falls below the system's rated operating current safety threshold, and this state is maintained for more than a preset reset time window, the processor clears the discrete short-time Joule integral register. In the specific parameter tuning, the aforementioned system rated operating current safety threshold is calibrated based on the maximum steady-state operating current of the controlled load in the high-voltage DC distribution network; the duration of the reset time window is set with reference to the thermal time constant of the fusible component under natural cooling conditions, typically taking a value of 1 to 3 times this thermal time constant. This reset operation ensures that the discrete short-time Joule integral is used only to evaluate the continuous overload or short-circuit energy accumulation process.

[0081] S306, the processor calls the digital comparator in the control logic to compare the discrete short-time Joule integral at the current discrete moment with the dynamic action integral threshold at the corresponding moment.

[0082] At the end of the control cycle, the processor executes the aforementioned comparison instruction. When the discrete short-time Joule integral is less than the dynamic action integral threshold, it indicates that the energy absorbed by the main branch has not yet reached the set limit for triggering protection action. In this state, the processor does not output an intervention instruction, the system maintains a closed-loop state, and proceeds to the next control cycle.

[0083] When the discrete short-time Joule integral is greater than or equal to the dynamic action integral threshold, the processor determines that the energy absorbed by the main branch has reached the system's adaptive tolerance limit. Based on model deduction, the internal temperature of the fusible component is approaching the melting critical point. At this time, the processor determines that the system has entered the thermal collapse critical region.

[0084] S307, after determining that it has entered the thermal collapse critical region, the processor generates an internal state switching flag and controls the program to jump into the phase change latent heat time window monitoring mechanism.

[0085] In existing conventional high-voltage fuse control logic, once the integral value exceeds the limit, the controller usually issues a direct tripping command. Considering that directly cutting off high voltage and high current can easily cause arcing between mechanical contacts, this embodiment adjusts the triggering timing with feedforward.

[0086] After establishing the critical thermal collapse state, the processor does not immediately output hardware drive signals to the separation device or the insulated gate bipolar transistor module. Instead, the processor sets a monitoring flag in its internal register, suspends the current Joule integral accumulation main loop, and activates the high-frequency feature recognition program corresponding to step S400. Through this state transition mechanism, the control system shifts the judgment dimension of protection execution from macroscopic energy accumulation to the microscopic material phase transition feature monitoring stage, creating control conditions for subsequent commutation assistance using the high resistance characteristics in the early stage of phase transition.

[0087] In this embodiment, step S400, which involves establishing a historical temperature sliding data window and performing convolution operations using a smoothing polynomial differential operator, may specifically include the following steps: S401: After the program jumps to the phase change latent heat time window monitoring mechanism, the processor allocates an independent buffer area in its internal memory to construct a historical temperature sliding data window based on the first-in-first-out queue.

[0088] In terms of the specific execution mechanism, the processor receives the optimal equivalent temperature output by step S200 in each control cycle. The processor writes the newly acquired optimal equivalent temperature to the head of the queue and simultaneously removes the oldest historical data from the tail of the queue. Through this dynamic data push-in and remove operation, the control system continuously maintains a discrete sample sequence of a fixed length in memory.

[0089] As an optional implementation, the length of this historical temperature sliding data window is denoted as... To meet the symmetry requirements of subsequent smooth polynomial operations, the length... It is usually set to an odd number greater than 3. In specific engineering calibrations, The specific value is determined based on the typical phase transition duration of the fusible component under the expected short-circuit current and the step size of the underlying control cycle. This sliding data window provides the necessary time-domain data foundation for subsequently capturing the short-lived microscopic physical phase transition characteristics.

[0090] Considering that sufficient historical data has not yet accumulated in the buffer area when the program first jumps into the monitoring mechanism, the processor is set to execute the initial filling logic. The processor will continuously fill the queue with newly acquired optimal equivalent temperatures, but when the number of data points in the queue reaches... Previously, subsequent difference operations were temporarily bypassed. This preventative mechanism avoids logical dead zones in the underlying control program caused by array out-of-bounds errors.

[0091] S402, the processor calls the smoothed polynomial differential operator to perform discrete convolution operation on the data sequence within the historical temperature sliding data window to obtain the first and second derivatives of the optimal equivalent temperature.

[0092] In digital signal control environments, extracting the slope and curvature of physical quantities typically relies on differential algorithms. Directly performing a second finite difference operation on a discrete temperature sequence can amplify high-frequency noise, potentially causing false triggering of subsequent identification logic. While conventional infinite impulse response low-pass filters possess noise reduction capabilities, their inherent group delay introduces phase lag, causing the control system to miss the brief physical commutation time window.

[0093] To overcome the engineering conflict between numerical computation and real-time performance, the processor introduces a smooth polynomial differential operator based on the Savitzky-Golay algorithm. This operator utilizes the principle of local polynomial least squares fitting to reconstruct the time domain of sample points within the data window, and directly parses the derivatives of each order during the fitting process.

[0094] In the underlying implementation of the program, the processor extracts all discrete temperature sequences within the historical temperature sliding data window and calls the first-order and second-order convolution coefficient matrices pre-stored in the read-only memory area. During the calculation, the processor multiplies each discrete optimal equivalent temperature within the data window with its corresponding first-order convolution coefficient within the data window, and sums all the products to calculate the first derivative of the optimal equivalent temperature at the current discrete moment. Based on the same logic, the processor multiplies each discrete optimal equivalent temperature with its corresponding second-order convolution coefficient and sums them to obtain the second derivative of the optimal equivalent temperature at the current discrete moment.

[0095] For generating the convolution coefficient matrix, those skilled in the art can do so based on a preset polynomial fitting order and a determined window length. The corresponding coefficient values ​​are obtained offline using the standard least squares method. In practical applications, to balance the smoothing effect with the retention of phase transition characteristics, the order of this polynomial fitting is usually set to second or third order. Its mathematical derivation process is well-known in the field and will not be elaborated upon here.

[0096] Through the above operational logic, the processor only needs to execute basic constant multiplication and accumulation instructions during online operation, avoiding complex online matrix inversion operations. This data processing architecture, while filtering out broadband sampling noise, obtains the first and second derivatives containing the physical evolution trend, providing basic state variables for subsequent analysis of the latent heat state of phase transition.

[0097] In this embodiment, step S400, which involves identifying the initiation time of the latent heat of phase transition based on the logical operation results of the critical temperature threshold, the first derivative state, and the second derivative state, may specifically include the following steps: S403, the processor performs a spatial dimension state comparison based on the optimal equivalent temperature and the set critical temperature threshold.

[0098] During the phase transition from solid to liquid state of the internal metallic material of a fusible component, its temperature approaches and stagnates near the physical melting point. In the specific discrimination mechanism, the processor extracts the optimal equivalent temperature calculated at the current discrete moment and compares it numerically with the set critical temperature threshold.

[0099] As an optional implementation, considering potential computational biases in the discrete thermal network model derivation and approximation errors during polynomial fitting, the critical temperature threshold is not equivalent to the absolute physical melting point of the fusible component, but rather equal to the physical melting point minus a preset temperature tolerance boundary. The specific value of this temperature tolerance boundary is set based on the error statistical interval during the factory calibration phase of the control system. When the processor determines that the optimal equivalent temperature at the current discrete moment is greater than or equal to this critical temperature threshold, it confirms that the current state satisfies the spatial absolute value condition for phase transition identification.

[0100] S404, the processor performs time-domain stationarity verification based on the first derivative of the optimal equivalent temperature and obtains the first derivative state.

[0101] According to the physical evolution of metal phase transitions, during the absorption of latent heat of phase transition, the injected Joule thermal energy is used to break the chemical bonds in the metal lattice, causing the macroscopic temperature rise to stagnate. To quantify this trend at the digital level, the processor obtains the first derivative calculated in step S402.

[0102] Due to underlying floating-point truncation errors and residual noise, the calculated first derivative is difficult to maintain strictly zero during the temperature rise stagnation period. To accommodate this engineering limitation, the processor internally presets a first-order differential zero-reaching dead zone. The processor takes the absolute value of the acquired first derivative and compares this absolute value with the first-order differential zero-reaching dead zone. In actual parameter tuning, the value of this first-order differential zero-reaching dead zone is set based on the maximum ambient temperature drift rate during steady-state system operation, typically controlled within a minimum range to filter out normal fluctuations. When this absolute value is less than or equal to the first-order differential zero-reaching dead zone, the processor determines that the upward trend of the optimal equivalent temperature has slowed to within the tolerance range, thus confirming that the time-domain stationarity condition for phase transition identification is met.

[0103] S405, the processor extracts the inflection point features of the physical state evolution based on the second derivative of the optimal equivalent temperature, and obtains the second derivative state.

[0104] In the early stages approaching the physical melting point, the resistance of the fusible component increases with rising temperature, the Joule heating power increases in a positive feedback manner, and the temperature rise curve shows an accelerated upward trend, at which point the corresponding second derivative is positive. When the material structure enters the latent heat stage of phase transition, the slope of the temperature rise curve begins to decline, changing from an accelerated upward trend to a smooth extension, forming a mathematical inflection point.

[0105] The processor captures this physical characteristic by comparing the states of the second derivatives at two adjacent discrete time points. The processor retrieves the second derivatives from the previous discrete time point and the current discrete time point. When the processor determines that the second derivative at the previous discrete time point is strictly greater than zero, and the second derivative at the current discrete time point is less than or equal to zero, it indicates that the second derivative has crossed zero from positive to negative. Under this condition, the processor confirms that the inflection point evolution condition for phase transition identification is met.

[0106] S406, the processor performs a logical AND operation on the above spatial absolute value condition, temporal stationarity condition and inflection point evolution condition, and establishes the starting moment of the latent heat of phase transition.

[0107] To prevent misjudgments caused by sudden changes in data from a single dimension, the processor calls a software logic gate at the end of the control cycle to perform a logical AND operation on the three conditional states output by stages S403 to S405. Only when the spatial absolute value condition, the temporal stationarity condition, and the inflection point evolution condition are all met simultaneously, the processor determines that the fusible component has entered the latent heat evolution stage of phase transition. During this stage, the equivalent resistance of the fusible component exhibits a nonlinear increase, objectively creating a commutation back voltage in the main branch that hinders the continuous flow of current.

[0108] At the trigger moment when the logic operation result is true, the processor locks the current discrete time stamp within the control system as the start moment of the phase transition latent heat. To prevent the control program from repeatedly triggering the above judgment logic due to slight data fluctuations during the phase transition plateau, thus causing the timestamp to oscillate and refresh the dead zone, the processor simultaneously sets the timing lock flag in the system memory after recording the start moment of the phase transition latent heat. This flag causes the processor to bypass and disable the phase transition identification mechanism in subsequent control loops before the end of this protection process. By constructing the above multi-dimensional feature judgment and locking mechanism, the control system transforms continuous physical state changes into anchor reference points that can be used for subsequent feedforward commutation timing.

[0109] In this embodiment, the feedforward prediction process regarding the latent heat evolution time window in step S500 may specifically include the following steps: S501, the processor calls the physical melting enthalpy value pre-stored in the internal non-volatile memory and obtains the equivalent resistance value of the melting point corresponding to the physical melting point.

[0110] After determining the onset of the latent heat of phase change, the control system needs to assess the safe commutation time that the fusible component can maintain before structural disintegration. The length of this time window is constrained by the thermodynamic properties of the material and the real-time energy injection rate.

[0111] As an optional implementation, the processor reads the physical enthalpy of melting from a register. To ensure mathematical consistency of energy calculation dimensions, the physical enthalpy of melting in this embodiment pre-integrates the effective phase change mass parameter of the fusible component in practical applications, characterizing the absolute total heat energy that the component of a specific specification must absorb to melt from a solid to a liquid state. This parameter is calibrated based on material properties and fixed in the read-only memory area during the factory setting stage of the control system. Based on the obtained physical enthalpy of melting, the processor calls a pre-set table of resistance and temperature polynomial coefficients, substitutes the physical melting point into the polynomial calculation model, and obtains the equivalent resistance value of the melting point corresponding to the critical phase change state of the material.

[0112] S502, the processor combines the discrete sampled values ​​of the operating current at the current discrete moment with the equivalent resistance value of the melting point to calculate the instantaneous active heat power of the phase transition.

[0113] After a fusible component enters the latent heat stage of phase change, its macroscopic temperature essentially stagnates near its physical melting point. Based on this physical characteristic, the processor uses the aforementioned fixed equivalent resistance value of the melting point for extrapolation when calculating the heat-causing energy.

[0114] In the specific computational logic, the processor extracts the discrete sample value of the operating current at the current discrete moment, calculates the square of the discrete sample value of the operating current, and multiplies the square value by the equivalent resistance value of the melting point. The product result physically represents the instantaneous active heat power of phase change injected by the power distribution network into the fusible component in the melting plateau period under the current operating condition.

[0115] To prevent division-to-zero errors caused by sudden drops in external operating current during computation, the processor executes a lower limit clamping check procedure after obtaining the instantaneous active thermal power of the phase transition. If the calculated instantaneous active thermal power of the phase transition is lower than a set small bias constant, the processor will force the instantaneous active thermal power of the phase transition to be assigned the value of the bias constant, thereby preventing the subsequent division calculation logic from falling into the dead zone of data overflow.

[0116] The S503 processor uses the physical enthalpy of melting and the instantaneous active heat power of the phase transition to estimate the theoretical latent heat time window through division operations.

[0117] After completing the above parameter preparation and fault tolerance verification, to prevent the fusible components from absorbing excessive energy and generating destructive metal vapor arcs, the control system planned a feedforward commutation mechanism. As the core principle for generating the feedforward time in this embodiment, the processor executes estimation logic, dividing the physical enthalpy of fusion by the instantaneous active heat power of the phase transition. Its specific mathematical model is as follows: ; In the formula, For the current discrete time The theoretical latent heat time window; It is the physical enthalpy of melting; For the current discrete time Discrete sampled values ​​of the operating current; This is the equivalent resistance value at the melting point; This represents the instantaneous active heat power during phase transition.

[0118] By performing a division operation, the processor obtains the specific duration of the theoretical latent heat time window. This calculation process, without relying on external high-pressure temperature sensors, integrates macroscopic electrical parameters with the underlying phase change thermodynamic mechanism of the material, outputting a safe predictive reference value characterizing the remaining duration of liquefaction evolution.

[0119] In this embodiment, the process of multi-branch timing interlocking control and flexible disconnection in step S500 may specifically include the following steps: The S504 processor calculates the synchronization trigger time based on the start time of the phase change latent heat, the theoretical latent heat time window, and the system advance angle.

[0120] After obtaining the theoretical latent heat time window, the control system enters the calibration phase of feedforward execution. The processor invokes the pre-tuned system advance angle. As an optional implementation, the system advance angle is composed of the superposition of the inherent hardware delay parameters of the control system, typically covering the processor's instruction transmission bus delay time, the IGBT module's drive circuit response time, and the dead time of the mechanical mechanism of the separation device.

[0121] The processor will compare the start time of the latent heat of phase transition locked in step S400 with the current discrete time. The theoretical latent heat time windows are added together to obtain the theoretical thermal collapse limit time of the fusible component. The processor uses this thermal collapse limit time to subtract the system advance angle to calculate the synchronization trigger time. As the core basis for the execution of the feedforward action in this embodiment, its specific mathematical expression is as follows: ; In the formula, For synchronous triggering time; This is the starting moment of the latent heat of phase transition; For the current discrete time The theoretical latent heat time window; This is the system's advance angle.

[0122] Under extreme short-circuit high-current conditions, the theoretical latent heat time window will be significantly compressed. If the calculated synchronization trigger time is less than or equal to the current absolute running time within the control system, it means the system has lost its timing margin for feedforward waiting. For this condition, the processor has over-limit action logic. The processor will forcibly assign the current absolute running time to the synchronization trigger time, thereby preventing the underlying timing comparator from falling into a logical dead zone of continuous waiting.

[0123] The S505 processor monitors the absolute runtime of the system and outputs an interlocking switching instruction when the absolute runtime reaches the synchronization trigger moment.

[0124] The processor continuously compares the system's internal absolute runtime with the synchronization trigger time. When the absolute runtime reaches the synchronization trigger time, the processor executes a multi-branch interlocking trigger sequence.

[0125] As part of the coordinated operation of multiple branches, the processor outputs a trip command to the control terminal of the main branch's disconnect device and simultaneously outputs a gate turn-on command to the drive terminal of the IGBT module in the commutation branch. During this execution phase, the fusible component is in the final stage of its latent heat of phase change, and its internal liquefaction results in a high impedance characteristic in its equivalent resistance, creating a voltage drop in the main branch that limits current flow. Simultaneously, the IGBT module turns on upon receiving the gate turn-on command, providing a low-impedance shunt path. Relying on the impedance limitation of the main branch and the low-impedance shunt effect of the commutation branch, the operating current is transferred from the main branch to the commutation branch. This process is used to suppress the generation of a large-energy arc during the trip operation of the disconnect device.

[0126] The S506 processor controls the IGBT module to perform a flexible shutdown based on the insulation establishment state and uses a varistor to dissipate residual energy.

[0127] As the commutation process progresses, the operating current is transferred to the IGBT module. Upon receiving a trip command, the internal mechanical contacts of the disconnector open. Simultaneously, the processor starts an internal insulation establishment delay timer while outputting the trip command. The preset duration of this timer is calibrated based on the time required for the mechanical contacts of the disconnector to move to the safe insulation distance.

[0128] When the insulation setup delay timer reaches its preset duration, it indicates that the main branch has the isolation capability to withstand the system recovery voltage. At this time, the processor outputs a gate turn-off command to the IGBT module. To prevent the line parasitic inductance overvoltage caused by a large current interruption from breaking down the insulation boundary, the processor executes flexible turn-off logic. In the specific control implementation, the processor outputs a stepped-down signal to the driver terminal of the IGBT module to reduce the gate drive voltage in stages, or controls the drive circuit to connect a preset turn-off resistor, thereby slowing down the rate of decrease of the collector current.

[0129] As the IGBT module is turned off, the residual inductive energy in the power distribution network causes the terminal voltage at both ends of the converter branch to rise. When the voltage reaches the clamping voltage threshold of the varistor, the parallel varistor becomes conductive, converting the residual electromagnetic energy of the system into heat energy for dissipation, thus completing the disconnection process of the high voltage DC protection system.

[0130] For the selection of nonlinear clamping characteristic parameters and the setting of energy dissipation capacity of varistors, those skilled in the art can perform conventional matching based on the rated voltage of the high-voltage DC bus and the maximum expected inductance of the system. The selection calculation is a well-known technology in the field and will not be elaborated here.

[0131] In this embodiment, the present invention provides an adaptive intelligent control device for high-voltage fuses. As a specific system hierarchy division method, this adaptive intelligent control device for high-voltage fuses is constructed based on the same inventive concept as the aforementioned adaptive intelligent control method for high-voltage fuses. The computer program instructions set within the device are stored in the processor's non-volatile storage medium and loaded into memory during processor operation to be executed as virtual functional modules. Specifically, the adaptive intelligent control device for high-voltage fuses may include a parameter acquisition module, a state estimation module, a threshold reconstruction and determination module, a feature recognition module, and a feedforward triggering module.

[0132] The parameter acquisition module is configured to synchronously acquire the operating current of the main branch and the measured temperature of the thermistor via a converter. After acquiring the underlying physical parameters, the parameter acquisition module sets the measured temperature as the common heat sink observation benchmark and obtains the current change rate based on differential calculation of the operating current. In the specific data flow relationship, the parameter acquisition module buffers the digital discrete signal transmitted by the converter, and its output establishes communication connections with the inputs of the state estimation module and the threshold reconstruction and judgment module, respectively, to transmit the current change rate and the common heat sink observation benchmark.

[0133] The state estimation module is configured to establish a discretized thermal network model and construct an extended Kalman filter. It receives data from the parameter acquisition module, uses the heat power generated by the operating current as an input variable, and substitutes the common heat sink observation reference as an observation variable into the extended Kalman filter for calculation. The state estimation module uses Kalman gain for error convergence and outputs the optimal equivalent temperature of the fusible component. Internally, the state estimation module incorporates matrix multiplication, addition, and inversion logic, relying on observed boundary conditions to perform closed-loop feedback correction of parameter drift caused by the pure thermal network model derivation. The optimal equivalent temperature generated by the state estimation module is simultaneously pushed to the threshold reconstruction and judgment module and the feature recognition module.

[0134] The threshold reconstruction and determination module is configured to calculate the dynamic action integral threshold based on the optimal equivalent temperature and current change rate. Simultaneously with threshold generation, the integration register within the threshold reconstruction and determination module synchronously accumulates the discrete short-time Joule integral. When the threshold reconstruction and determination module detects that the discrete short-time Joule integral is greater than or equal to the dynamic action integral threshold, it determines that the main branch has entered the thermal collapse critical region. In the actual architecture, the threshold reconstruction and determination module internally includes a threshold generation unit for parameter multiplication and superposition, and a digital discrimination unit for energy accumulation comparison. When the system state meets the out-of-limit conditions, the threshold reconstruction and determination module sends a timing activation flag to the feature recognition module.

[0135] The feature recognition module is configured to establish a historical temperature sliding data window upon receiving an activation flag indicating entry into the thermal collapse critical region. The module utilizes a smoothed polynomial differential operator to perform convolution and summation operations on the data within the historical temperature sliding data window, extracting the first and second derivatives of the optimal equivalent temperature. Combined with preset tolerance parameters, the feature recognition module performs a digital logic AND operation based on the critical temperature threshold, the first derivative state, and the second derivative state to identify the initiation time of the latent heat of phase transition in the fusible component. This feature recognition module converts the state parameters of the physical phase transition into discrete-time variables and passes these time variables to the feedforward trigger module.

[0136] The feedforward trigger module is configured to calculate the synchronization trigger time based on the start time of the latent heat of phase change, the physical enthalpy of fusion of the fusible component, and the system advance angle. Internally, the feedforward trigger module includes a comparator that monitors the absolute running time of the system. When the absolute running time reaches the synchronization trigger time, the feedforward trigger module outputs an action execution command. In the specific module execution logic, the feedforward trigger module controls the separation device to perform a tripping operation and simultaneously outputs a gate turn-on command to the IGBT module for commutation. Furthermore, the feedforward trigger module also includes insulation delay control logic. After confirming that the separation device has established a safe insulation distance, the feedforward trigger module outputs a stepped-down signal to the IGBT module to perform a flexible turn-off operation, transferring the residual inductive electromagnetic energy in the distribution network to the varistor for dissipation. As the final execution interface of the system, the feedforward trigger module realizes the conversion and output of control algorithm commands to the underlying hardware drive signals.

[0137] Regarding the storage media partitioning, program pointer invocation mechanism, and process scheduling scheme for multi-threaded tasks within the physical processor or microcontroller for the aforementioned virtual functional modules, those skilled in the art can develop the underlying code architecture based on conventional embedded operating systems and memory management technologies. The register addressing and scheduling execution methods of the programs are well-known technologies in the field and will not be elaborated upon here.

[0138] Specific application examples and experimental verification analysis: Example scenario setup: This embodiment relies on a high-voltage DC distribution network with a rated voltage of 10kV and a rated operating current of 500A, and the controlled load is a large DC motor. The fusible components in the main branch circuit are made of silver-based materials (physical melting point...). The static Joule integral threshold is calibrated under standard conditions as follows: .

[0139] During system simulation, two typical operating conditions were injected to verify the reliability of the solution: Normal surge conditions: during operation When the motor starts, it generates a short-term starting surge current with a peak value of 1200A, and then... It will then fall back to the rated range.

[0140] High impedance fault condition: during operation When a high-impedance ground fault occurs in the power distribution network, the operating current continues to rise at a relatively small slope, causing heat to accumulate continuously in the fusible components.

[0141] Combined with appendix Figure 3 Comparative analysis of dynamic threshold and energy integral evolution: See attached document Figure 3 , Figure 3 This is a comparison graph of the adaptive integration threshold and the actual energy evolution according to an embodiment of the present invention. The horizontal axis of the graph represents the system running time (ms), and the vertical axis represents the Joule integral energy value. .

[0142] The maloperational defects of traditional static protection: such as Figure 3 As shown by the dotted lines, the action threshold of traditional protection methods is fixed as the static Joule integral threshold. During the surge condition ranging from 5ms to 15ms, Figure 3 The dashed line with forks represents the actual injected energy (i.e., the discrete short-time Joule integral). The current rises rapidly. Because traditional methods cannot identify the rate of change of current, the crossed dashed line... The area is marked with fine dots, which in actual engineering will directly cause the protection device to trip erroneously, resulting in the motor failing to start.

[0143] This invention provides surge suppression and accurate fault identification: Utilizing the adaptive method of this invention, the parameter acquisition module obtains the rate of change of current during a surge. Based on the threshold reconstruction formula Transient current suppression function It identifies non-fault surges and outputs a gain coefficient greater than 1. For example... Figure 3 The thick solid line in the middle (dynamic action integral threshold) As shown, the thick solid line was briefly lifted upwards during the surge interval (5-15ms), successfully encompassing the crossed dashed line (actual energy) below, effectively preventing false movements.

[0144] As time progresses to the high-impedance fault evolution region after 20ms, the optimal equivalent temperature... As the temperature continues to rise, the thermal boundary compensation term makes the dynamic adjustment coefficient in the formula less than 1. At this point, Figure 3 The thick solid line (dynamic action integral threshold) begins to decline smoothly, reflecting the reduction in the actual remaining heat capacity of the fusible component. Finally, the dashed line with the fork merges with the declining thick solid line. The timing of the intersection is precise, allowing the control system to accurately determine that the main branch has entered the thermal collapse critical zone and activate subsequent monitoring programs.

[0145] Combined with appendix Figure 4 Phase transition latent heat identification and feedforward triggering analysis: See attached document Figure 4 , Figure 4 This is a spatiotemporal characteristic joint latent heat identification and feedforward commutation calibration diagram according to an embodiment of the present invention. The diagram presents the microscopic physical characteristics after entering the thermal collapse critical region (20ms-35ms interval).

[0146] The horizontal axis represents the microscopic time window (ms) for phase change monitoring, the left vertical axis represents the optimal equivalent temperature (°C), and the right vertical axis represents the normalized rate of change.

[0147] Spatiotemporal feature extraction: such as Figure 4 The thick solid line in the figure shows the optimal equivalent temperature. As it approaches its physical melting point of 960℃, it exhibits a distinct and gradual phase transition plateau. During this period, the short dashed line (first derivative state) extracted by the smoothing polynomial differential operator rapidly drops and approaches the zero mark, satisfying the time-domain stationarity condition. Meanwhile, Figure 4 The dotted line (second derivative state) in the figure shows a clear mathematical inflection point.

[0148] Time anchor locking and commutation triggering: when Figure 4 When the dotted line (second derivative) in the graph crosses the zero mark from the positive interval (satisfying the zero-crossing characteristic), the control system performs a logical AND operation. At this moment... Figure 4 The moment when the latent heat of phase change begins is indicated by a vertical solid line and a circular anchor point. (At approximately 28ms).

[0149] In lock Then, the processor follows the feedforward formula. Calculate the theoretical latent heat time window and use the formula Subtracting the system advance angle, the final calculated synchronization trigger time is obtained. exist Figure 4 The center is marked by a vertical solid line on the right and a cross anchor point. When the system reaches this cross anchor point, the processor outputs an interlocking switching command, achieving arc-free flexible disconnection.

Claims

1. A high-voltage fuse adaptive intelligent control method, applied to a control system, the control system comprising a processor and a main branch and a converter branch connected in parallel, the main branch comprising a fusible component and a separation device connected in series, the converter branch comprising an IGBT module and a varistor, the IGBT module containing a thermistor, characterized in that, The method is executed by the processor and includes: The operating current of the main branch and the measured temperature of the thermistor are collected. The measured temperature is used as a common heat sink observation benchmark, and the current change rate is calculated based on the operating current. The heat power generated by the operating current is used as the input variable, and the common heat sink observation reference is used as the observation variable. The extended Kalman filter is substituted to perform error convergence and the optimal equivalent temperature of the fusible component is calculated. The dynamic action integral threshold is calculated based on the optimal equivalent temperature and the current change rate. The discrete short-time Joule integral is accumulated. When it is greater than or equal to the dynamic action integral threshold, the main branch is determined to have entered the thermal collapse critical region. After entering the thermal collapse critical region, the first and second derivatives of the optimal equivalent temperature are obtained. Logical operations are performed in combination with the set critical temperature threshold and the derivative states of the two to identify the start time of the latent heat of phase change of the fusible component. The synchronization trigger time is calculated based on the start time, physical melting enthalpy, and system advance angle. When the synchronization trigger time is reached, the separation device is controlled to trip, and a gate turn-on command is simultaneously output to the IGBT module for commutation.

2. The method according to claim 1, characterized in that, The process of using the heat power generated by the operating current as an input variable, substituting the common heat sink observation reference as an observation variable into the extended Kalman filter for error convergence, and calculating the optimal equivalent temperature of the fusible component includes: The resistance-temperature polynomial coefficient table stored in the register is called to map the core equivalent temperature of the previous discrete moment to the corresponding dynamic equivalent resistance value. The square of the operating current of the previous discrete moment is multiplied by the dynamic equivalent resistance value to obtain the input heat power injected into the fusible component at the previous discrete moment. By using the state transition matrix, the input control matrix, the input thermal power, and the optimal posterior estimated state vector of the previous discrete time step, advance prediction is performed to obtain the prior state vector and the prior error covariance matrix. Construct an observation matrix, and calculate the Kalman gain matrix based on the prior error covariance matrix, the observation matrix, and the preset observation noise covariance matrix; The innovation deviation between the actual collected common heat sink observation benchmark and the theoretical observation value is calculated. The Kalman gain matrix is ​​used to weight the innovation deviation and superimpose it onto the prior state vector. The first element in the corrected optimal posterior state vector is extracted and assigned as the optimal equivalent temperature.

3. The method according to claim 1, characterized in that, The process of calculating the dynamic action integral threshold based on the optimal equivalent temperature and the current change rate includes: Call the static Joule integral threshold, physical melting point, environmental reference temperature, and thermal fatigue compensation coefficient pre-stored in the internal non-volatile memory; Calculate the difference between the optimal equivalent temperature and the environmental reference temperature, and calculate the inherent difference between the physical melting point and the environmental reference temperature. Divide these two differences to obtain the normalized temperature evolution rate of the current physical state. Multiply the normalized temperature evolution rate by the thermal fatigue compensation coefficient, and subtract the product from the constant 1 to obtain the thermal boundary compensation term; Execute the transient current suppression function and generate a dynamic adjustment coefficient based on the current change rate; The static Joule integral threshold, the thermal boundary compensation term, and the dynamic adjustment coefficient are multiplied and superimposed. When the calculation result is negative or has an unreasonable minimum value, the lower limit threshold is applied to force the output of the preset safety action baseline value as the dynamic action integral threshold at the current discrete moment.

4. The method according to claim 3, characterized in that, The process of executing the transient current suppression function and generating a dynamic adjustment coefficient based on the current change rate includes: The real-time current change rate is compared with the set surge lower limit slope value and short-circuit fault upper limit slope value. When the rate of change of current is within the range of the lower limit slope value of surge and the upper limit slope value of short circuit fault, it indicates that the main branch is experiencing a non-faulty transient large load pulling process. The value of the transient current suppression function is assigned a preset gain coefficient, and the resulting calculation result constitutes the dynamic adjustment coefficient for the current period. When the rate of change of current is outside the above range, the value of the function is fixed as a constant of 1, and the resulting calculation result constitutes the dynamic adjustment coefficient for the current period.

5. The method according to claim 1, characterized in that, The process of accumulating discrete short-time Joule integrals includes: The square of the discrete sampled value of the running current is calculated, and the square is multiplied by the step size of the control cycle to quantify the infinitesimal energy injected into the main branch in the current single control cycle. The infinitesimal energy is added to the integral history value stored in the register at the previous discrete moment to obtain the discrete short-time Joule integral at the current discrete moment. When the discrete sampled value of the operating current is detected to fall below the system's rated operating current safety threshold, and this state is maintained for a period of time exceeding the preset reset time window, the discrete short-time Joule integral register is cleared.

6. The method according to claim 1, characterized in that, The process of obtaining the first and second derivatives of the optimal equivalent temperature includes: A buffer area is allocated in the internal memory, a historical temperature sliding data window based on a first-in-first-out queue is constructed, and the latest acquired optimal equivalent temperature is written to the head of the queue. Call the first-order convolution coefficient matrix and the second-order convolution coefficient matrix that are pre-stored in the read-only memory area; Each discrete optimal equivalent temperature within the historical temperature sliding data window is multiplied one by one with its corresponding first-order and second-order convolution coefficients in the data window, and the results are summed to obtain the first-order derivative and the second-order derivative that contain the physical evolution trend.

7. The method according to claim 1, characterized in that, The process of performing logical operations based on a set critical temperature threshold and the derivative states of both to identify the onset time of the latent heat of phase change in the fusible component includes: Extract the optimal equivalent temperature calculated at the current discrete moment and compare it with the set critical temperature threshold. When the optimal equivalent temperature is greater than or equal to the critical temperature threshold, it is confirmed that the current state satisfies the spatial absolute value condition. Take the absolute value of the obtained first derivative. When the absolute value is less than or equal to the set dead zone where the first derivative approaches zero, it is confirmed that the time-domain stationarity condition is satisfied. By comparing the second derivative states of two adjacent discrete time points, if the second derivative of the previous discrete time point is strictly greater than zero and the second derivative of the current discrete time point is less than or equal to zero, the inflection point evolution condition is confirmed to be satisfied. The spatial absolute value condition, the temporal stationarity condition, and the inflection point evolution condition are subjected to a logical AND operation. When all three conditions are met, the current discrete time stamp inside the control system is locked as the start time of the latent heat of phase transition, and the timing lock flag in the system memory is set synchronously.

8. The method according to claim 1, characterized in that, The process of calculating the synchronization trigger time based on the start time, physical melting enthalpy, and system advance angle includes: By substituting the physical melting point into the polynomial calculation model, the fixed equivalent resistance value of the melting point corresponding to the critical state of phase transformation of the material can be obtained. The square of the discrete sampled value of the operating current is calculated, and the square value is multiplied by the equivalent resistance value of the melting point to calculate the instantaneous active heat power of the phase change. When the instantaneous active heat power of the phase change is lower than the set small bias constant, the instantaneous active heat power of the phase change is forcibly assigned to the bias constant. The theoretical latent heat time window is estimated by dividing the physical enthalpy of melting, which includes effective phase change mass parameters, by the instantaneous active heat power of the phase change. The thermal collapse limit time is obtained by adding the start time of the phase change latent heat to the theoretical latent heat time window. The synchronization trigger time is calculated by subtracting the system advance angle from the thermal collapse limit time.

9. The method according to claim 8, characterized in that, The system advance angle is composed of the processor's instruction transmission bus delay time, the IGBT module's drive circuit response time, and the dead time of the mechanical mechanism of the separation device. When the calculated synchronization trigger time is less than or equal to the current absolute running time inside the control system, the current absolute running time is forcibly assigned to the synchronization trigger time.

10. The method according to claim 1, characterized in that, After the separation device trips and a gate-on command is simultaneously output to the IGBT module for commutation, the method further includes: Start the internal insulation build-up delay timer; When the insulation establishment delay timer reaches the preset duration, it indicates that the main branch has the isolation capability to withstand the system recovery voltage. The IGBT module is then controlled to execute the flexible turn-off logic, specifically by outputting a stepped voltage reduction signal to the drive terminal of the IGBT module to reduce the gate drive voltage in stages, or by controlling the drive circuit to connect a preset turn-off resistor, thereby slowing down the rate of decrease of the collector current. When the voltage reaches the clamping voltage threshold of the varistor, the parallel varistor becomes conductive, converting the residual electromagnetic energy of the system into heat energy for dissipation.

Images