Busbar temperature rise prediction and dynamic current regulation method and device for column head tank

Abstract

The present application relates to a kind of column head cabinet busbar temperature rise prediction and dynamic current regulation method and device, belong to the field of power distribution safety control, method includes: by obtaining busbar row surface temperature sequence and corresponding time's current-carrying capacity data, generate sequence data;Based on the sequence data, the temperature rise rate in unit time is calculated;The temperature rise rate is input into pre-trained temperature prediction model, and the busbar row temperature prediction value under future set length output by the temperature prediction model is obtained;When the temperature prediction value exceeds the temperature rise value of busbar row material permission, generate current-carrying capacity correction instruction, the threshold of the temperature rise value is preset in advance;According to the current-carrying capacity correction instruction, reduce feeder loop current, until the predicted temperature value is lower than the allowable temperature rise value;Realize the effect of starting regulation before overheating occurs by temperature rise trend prediction, avoid the thermal deformation of busbar row material.

Description

Technical Field

[0001] This invention relates to the field of power distribution safety control technology, and in particular to a method and device for predicting the temperature rise of busbars in a power distribution unit and for dynamic current carrying capacity adjustment. Background Technology

[0002] Existing busbar temperature rise control in switchgear mainly relies on a threshold alarm mechanism. This involves installing fixed temperature sensors on the busbar surface; when the measured temperature exceeds the material's safety threshold, an alarm is triggered or the power is cut off. This solution has three major drawbacks:

[0003] (1) Delay defect: When the alarm is triggered, the busbar is already in an overheated state, and the copper busbar material undergoes irreversible deformation;

[0004] (2) Waste of current carrying capacity: In order to avoid risks, the operators usually set the maximum current carrying capacity to 60%-70% of the theoretical value, resulting in low utilization of busbar resources;

[0005] (3) Local overheating is missed: Traditional single-point temperature measurement cannot capture the gradient temperature rise of key nodes such as connection ends / bends. Summary of the Invention

[0006] The main objective of this invention is to provide a method and device for predicting the temperature rise and dynamically adjusting the current carrying capacity of busbars in a cabinet, thereby enabling adjustment to be initiated before overheating occurs based on temperature rise trend prediction, thus avoiding thermal deformation of the busbar material.

[0007] To achieve the above objectives, the present invention provides a method for predicting the temperature rise and dynamically adjusting the current carrying capacity of the busbar in a row-head cabinet, comprising the following steps:

[0008] Obtain the surface temperature sequence of the busbar and the corresponding current carrying capacity data at the corresponding time, and generate sequence data;

[0009] The rate of temperature rise per unit time is calculated based on the sequence data;

[0010] The temperature rise rate is input into a pre-trained temperature prediction model to obtain the busbar temperature prediction value for a future set time period output by the temperature prediction model.

[0011] When the predicted temperature exceeds the allowable temperature rise value of the busbar material, a current carrying capacity correction command is generated, and the threshold value of the temperature rise value is preset in advance.

[0012] The feeder loop current is reduced according to the current carrying capacity correction command until the predicted temperature value is lower than the allowable temperature rise value.

[0013] Further, the step of obtaining the busbar surface temperature sequence includes:

[0014] Temperature values ​​of nodes on the inner wall of the busbar are collected through distributed contacts of pre-set temperature sensors inside the busbar.

[0015] Record the temperature measurements at each node at fixed time intervals;

[0016] The temperature values ​​of each node at the same timestamp are integrated into the first temperature sequence data. The second temperature sequence data corresponding to the previous period timestamp of the first temperature sequence data is retained, while the temperature sequence content of other timestamps is forgotten.

[0017] Further, the steps of obtaining the current carrying capacity data of the busbar at the corresponding time and generating sequence data include:

[0018] The current value of each circuit is measured in real time by pre-installing current transformers in the feeder circuits inside the cabinet busbar.

[0019] The current values ​​of each circuit are converted into digital signals of current carrying capacity, and the mapping relationship between the conversion timestamp and the current carrying capacity data is recorded;

[0020] According to the timestamp of temperature acquisition, the first temperature sequence data is bound to the current carrying capacity digital signal to obtain the sequence data.

[0021] Further, the step of calculating the temperature rise rate per unit time based on the sequence data includes:

[0022] Extract the first temperature sequence data with the current timestamp and its associated current carrying capacity data from the sequence data;

[0023] Obtain the second temperature sequence data from the previous period's timestamp;

[0024] Calculate the temperature difference between all nodes of the busbar in the first temperature sequence data and the second temperature sequence data;

[0025] Divide the temperature difference by a fixed time interval to obtain the temperature rise rate of all nodes of the busbar.

[0026] Further, the step of inputting the temperature rise rate into a pre-trained temperature prediction model and obtaining the predicted busbar temperature for a future set time period output by the temperature prediction model includes:

[0027] Combine the temperature rise rate of all nodes at the current timestamp with the bound carrying capacity data into the model input vector;

[0028] The input vector is fed into a pre-trained temperature prediction model;

[0029] Obtain the temperature prediction values ​​of all nodes output by the temperature prediction model for a future set time span. The prediction process of the temperature prediction model is as follows: generate a feature matrix by linearly weighting the temperature rise rate and current carrying capacity data in the input vector; perform multiplication and accumulation operations on the feature matrix and the pre-stored model parameter matrix; and output the temperature prediction values ​​of all nodes for a future set time span through a linear mapping function.

[0030] Furthermore, the step of generating a current-carrying capacity correction command when the predicted temperature exceeds the allowable temperature rise value of the busbar material includes:

[0031] Compare the predicted temperature values ​​of all nodes with the allowable temperature rise values ​​of the busbar material;

[0032] When the predicted temperature of any target node exceeds the allowable temperature rise value, a current carrying capacity correction instruction containing the identifier of the feeder loop associated with that node is generated, wherein the correction instruction contains the current amplitude value to be reduced.

[0033] Further, the step of reducing the feeder loop current according to the current-carrying correction command until the predicted temperature value is lower than the allowable temperature rise value includes:

[0034] Match the target thyristor regulator according to the feeder loop identifier in the correction instruction;

[0035] The thyristor regulator is controlled to reduce the output current by the required reduction value.

[0036] Further, the step of matching the target thyristor regulator according to the feeder loop identifier in the correction instruction includes:

[0037] Based on the determination of any target node location exceeding the temperature rise value, read the preset mapping table between the feeder loop identifier and the connection port of the thyristor regulator;

[0038] The physical control port of the target thyristor regulator is located through the mapping table, and the physical control port is mapped to the node location.

[0039] Furthermore, the pre-training method for the temperature prediction model includes:

[0040] Under normal operating conditions, the temperature values ​​of key nodes of the busbar are continuously collected by distributed contact temperature sensors, and the current carrying capacity data is synchronously measured by the current transformer of the feeder circuit. A complete sequence data package containing the first temperature sequence data with the current timestamp and its bound current carrying capacity data, as well as the second temperature sequence data corresponding to the previous cycle timestamp, is generated at fixed time intervals.

[0041] The sequence data packets are stored in the local cache area of ​​the column head cabinet in real time. When the amount of cached data reaches the preset storage capacity, the model training enable signal is triggered.

[0042] In response to the enable signal, all sequence data packets are extracted from the buffer, and the second temperature sequence data in each data packet is extracted as the input feature set, and the first temperature sequence data corresponding to the timestamp is used as the output tag set.

[0043] The initial model parameter matrix is ​​constructed and loaded into the matrix operation processor built into the column head cabinet. The input feature set is input into the processor in chronological order to perform matrix multiplication and addition operations, generating a preliminary predicted temperature sequence.

[0044] The absolute error between the preliminary predicted temperature sequence and the output label set is calculated. When the average absolute error exceeds the allowable error threshold, the parameter correction amount is generated through the error backpropagation controller.

[0045] Update the model parameter matrix using the parameter correction amount and repeat the matrix operation and error calculation process until the mean absolute error stabilizes within the allowable error threshold.

[0046] The final optimized model parameter matrix is ​​burned into the read-only memory of the column head cabinet, the training enable signal is turned off, and the buffer is cleared.

[0047] This invention also proposes a device for predicting the temperature rise of the busbar in a row-head cabinet and for dynamic current carrying regulation, comprising:

[0048] The acquisition unit is used to acquire the surface temperature sequence of the busbar and the corresponding current carrying capacity data at the time, and generate sequence data;

[0049] A calculation unit is used to calculate the temperature rise rate per unit time based on the sequence data;

[0050] The model unit is used to input the temperature rise rate into a pre-trained temperature prediction model and obtain the predicted busbar temperature value under a future set time period output by the temperature prediction model.

[0051] The instruction unit is used to generate a current carrying capacity correction instruction when the predicted temperature value exceeds the allowable temperature rise value of the busbar material, wherein the threshold value of the temperature rise value is preset in advance;

[0052] The action unit is configured to reduce the feeder loop current according to the current carrying capacity correction command until the predicted temperature value is lower than the allowable temperature rise value.

[0053] The method and apparatus for predicting the temperature rise and dynamically adjusting the current carrying capacity of the busbar in the head unit provided by this invention have the following beneficial effects:

[0054] (1) Adjustment is initiated before overheating occurs by predicting the temperature rise trend, so as to avoid thermal deformation of the busbar material.

[0055] (2) Use existing monitoring and execution devices to achieve closed-loop control and avoid additional hardware investment. Attached Figure Description

[0056] Figure 1 This is a flowchart illustrating the method for predicting the temperature rise of the busbar in the head unit and adjusting the dynamic current carrying capacity in one embodiment of the present invention.

[0057] Figure 2 This is a structural block diagram of the busbar temperature rise prediction and dynamic current carrying adjustment device in one embodiment of the present invention;

[0058] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0060] Reference Figure 1 This is a flowchart illustrating a method for predicting busbar temperature rise and dynamically adjusting current in a busbar control cabinet proposed in this invention. The method employs a BMS system deployed in the busbar control cabinet to process signals and execute the aforementioned method. Alternatively, the method can be executed by deploying a computer device on the busbar control cabinet. The executed method includes the following steps:

[0061] S1, acquire the surface temperature sequence of the busbar and the corresponding current carrying capacity data at the time, and generate sequence data;

[0062] S2, calculate the temperature rise rate per unit time based on the sequence data;

[0063] S3, input the temperature rise rate into the pre-trained temperature prediction model, and obtain the busbar temperature prediction value under the future set time period output by the temperature prediction model.

[0064] S4, when the predicted temperature exceeds the allowable temperature rise value of the busbar material, a current carrying capacity correction command is generated, and the threshold value of the temperature rise value is preset in advance;

[0065] S4, reduce the feeder loop current according to the current carrying capacity correction command until the predicted temperature value is lower than the allowable temperature rise value.

[0066] In one embodiment of S1, the step of obtaining the busbar surface temperature sequence includes:

[0067] Temperature values ​​of nodes on the inner wall of the busbar are collected through distributed contacts of pre-set temperature sensors inside the busbar.

[0068] Record the temperature measurements at each node at fixed time intervals;

[0069] The temperature values ​​of each node at the same timestamp are integrated into the first temperature sequence data. The second temperature sequence data corresponding to the previous period timestamp of the first temperature sequence data is retained, while the temperature sequence content of other timestamps is forgotten.

[0070] In this embodiment, during the manufacturing stage of the busbar of the cabinet, distributed contact temperature sensor contacts are pre-embedded in the inner wall of the busbar. The contact layout is determined based on the busbar thermal field simulation results. Specifically, main monitoring nodes are set in the edge area where the current skin effect is significant, at the connection terminal crimping point, and in the right-angle bend area. The distance between adjacent nodes does not exceed 1 / 10 of the busbar length. The contacts adopt a spring-loaded pin structure. After the thermal expansion of the busbar when energized, its thermally conductive ceramic probe forms a continuous pressure contact with the inner wall of the copper busbar, eliminating the measurement gap caused by thermal expansion and contraction of traditional patch sensors. Each contact is connected to the data acquisition module on the top of the cabinet through a high-temperature resistant insulated wire. The module polls and reads the node temperature value at a fixed interval of 5 seconds. During the acquisition process, outliers are filtered in two stages: first, invalid data exceeding the physical temperature limit of the copper busbar (e.g., -20℃ to 150℃) are discarded; second, abrupt changes in temperature exceeding 30℃ between adjacent nodes are removed (judged as wire interference). The valid temperature values ​​are integrated into a structured sequence by timestamp. The data collected at the current moment is marked as the first temperature sequence, and the data at the same time point in the previous period is marked as the second temperature sequence. Historical redundant data is cleared in real time to save storage resources.

[0071] In another embodiment of S1, the step of obtaining the current carrying capacity data of the busbar at the corresponding time and generating sequence data includes:

[0072] The current value of each circuit is measured in real time by pre-installing current transformers in the feeder circuits inside the cabinet busbar.

[0073] The current values ​​of each circuit are converted into digital signals of current carrying capacity, and the mapping relationship between the conversion timestamp and the current carrying capacity data is recorded;

[0074] According to the timestamp of temperature acquisition, the first temperature sequence data is bound to the current carrying capacity digital signal to obtain the sequence data.

[0075] In this embodiment, a switchable current transformer is integrated at the output terminal of each feeder circuit of the cabinet. Its magnetic core uses a nanocrystalline ribbon winding structure to suppress magnetic saturation caused by high-order harmonics. The secondary output is connected to a high-precision analog-to-digital converter (ADC) module via a shielded cable. The high-precision ADC module performs synchronous sampling with a 5-second cycle. Specifically, it captures the instantaneous current value triggered by the falling edge of the temperature acquisition clock pulse and converts it into a 24-bit current-carrying digital signal via a Σ-Δ ADC. Immediately after signal conversion, a timestamp is added, with the tag format aligned to the temperature acquisition system at the millisecond level. Data binding is completed in the cabinet's edge computing unit: when both temperature sequence data and current-carrying data at the same timestamp are ready, a data packaging thread is triggered to encapsulate the first temperature sequence data packet and the corresponding current-carrying digital signal into a structured data frame. The frame header contains metadata such as timestamp, node count, and checksum. Specifically, when current-carrying data at a certain timestamp is missing, it is automatically filled with data from the previous valid cycle and marked as complete, ensuring the continuity of the sequence data.

[0076] To further explain, each feeder circuit of the cabinet integrates a switchable current transformer. The magnetic core of the current transformer is made of nanocrystalline material, which is vacuum annealed and then wound into a closed loop structure. The high permeability and low coercivity of the nanocrystalline material are used to suppress magnetic saturation caused by high-order harmonics. The secondary output winding is led out with triple-shielded stranded wire. The outer layer is grounded with a braided copper mesh to eliminate electromagnetic interference. The inner layer is wrapped with aluminum foil to isolate electrostatic coupling. The middle ferrite magnetic ring absorbs common-mode noise. The shielded cable terminal is connected to the differential input port of the high-precision analog-to-digital converter module. Then, the high-precision analog-to-digital converter module performs synchronous sampling with a fixed period of 5 seconds. It uses the master clock crystal oscillator to output a 10MHz square wave pulse, which is divided by a frequency divider to generate a 5-second temperature acquisition clock pulse. After a 200ns delay, the falling edge of this pulse triggers the current sample-and-hold circuit, instantly freezing the analog signal on the secondary side of the current transformer. The frozen signal is input to the differential input of the Σ-Δ ADC, and the input voltage is oversampled 256 times through a switched capacitor network to convert the analog signal into a high-speed bit stream. After being downsampled by the Sinc³ digital filter, a 24-bit current carrying capacity digital signal is generated. The rising edge of the signal conversion immediately triggers the timestamp unit, which synchronously receives the IRIG-B time code signal from the temperature acquisition system, unifying the absolute time of the current carrying capacity data and temperature data to the same millisecond-level time coordinate system. The timestamp is written into the data frame header verification area. If the time code signal is abnormal, it automatically switches to local crystal oscillator timing and records the status flag bit, thus obtaining the sequence data.

[0077] In one embodiment of step S2, the step of calculating the temperature rise rate per unit time based on the sequence data includes:

[0078] Extract the first temperature sequence data with the current timestamp and its associated current carrying capacity data from the sequence data;

[0079] Obtain the second temperature sequence data from the previous period's timestamp;

[0080] Calculate the temperature difference between all nodes of the busbar in the first temperature sequence data and the second temperature sequence data;

[0081] Divide the temperature difference by a fixed time interval to obtain the temperature rise rate of all nodes of the busbar.

[0082] In this embodiment, a dual-port RAM storage area is established within the edge computing unit. Port A receives the current sequence data frame from the data bus in real time, while port B cyclically stores historical sequence data corresponding to the timestamp of the previous cycle. When the rising edge of the 5-second clock pulse is triggered, the first temperature sequence data and its bound current carrying capacity data for the current timestamp are first extracted from port A, and the second temperature sequence data at the same time point of the previous cycle are simultaneously read from port B. The two sets of temperature sequence data are input in parallel to a comparator array, which includes voltage comparison circuits equal to the number of busbar nodes. Each circuit performs analog differential amplification on the temperature value of the same node number, and the output is converted into a temperature difference voltage signal by a precision instrumentation amplifier. This voltage signal is connected to a 24-bit Σ-Δ ADC to be converted into a digital quantity, and then calibrated by a scaling factor memory to finally generate a digital quantity of the temperature difference for each node. This digital quantity is input to a hardware divider, whose divisor is preset with a fixed time interval value. The binary division operation is performed through a shift register and an accumulator, and the output is the digital quantity of the temperature rise rate per unit time. The calculation result is stored in a FIFO buffer queue to await model calls.

[0083] In one embodiment of S3, the step of inputting the temperature rise rate into a pre-trained temperature prediction model and obtaining the predicted busbar temperature for a future set duration output by the temperature prediction model includes:

[0084] Combine the temperature rise rate of all nodes at the current timestamp with the bound carrying capacity data into the model input vector;

[0085] The input vector is fed into a pre-trained temperature prediction model;

[0086] Obtain the temperature prediction values ​​of all nodes output by the temperature prediction model for a future set time span. The prediction process of the temperature prediction model is as follows: generate a feature matrix by linearly weighting the temperature rise rate and current carrying capacity data in the input vector; perform multiplication and accumulation operations on the feature matrix and the pre-stored model parameter matrix; and output the temperature prediction values ​​of all nodes for a future set time span through a linear mapping function.

[0087] In this embodiment, a dedicated prediction channel is established in the matrix operation coprocessor of the edge computing unit. When a temperature rise rate data packet is received, the temperature rise rate values ​​of all nodes at the current timestamp (such as the dT / dt values ​​of nodes K1 to Kn) are first concatenated with the bound current carrying capacity digital signal in a preset order to form an input vector. The vector is stored at the starting address of the DDR3 cache. The coprocessor's direct memory access controller (DMA) loads the vector into the input register file of the floating-point unit (FPU) and simultaneously reads the pre-trained model parameter matrix from the read-only memory (ROM). The FPU performs a linear weighted combination operation: combining the input vector elements with the parameters... The elements in the first column of the matrix are multiplied in parallel by a hardware multiplier array. The product is accumulated by a pipelined addition tree to generate the first element of the feature matrix. This process is repeated until the feature matrix is ​​filled. The feature matrix is ​​then multiplied and accumulated (MAC) with the remaining submatrices of the parameter matrix. The matrix is ​​computed in blocks by a pulsating array architecture, and one row is multiplied by one column in each clock cycle. The final output vector is processed by a linear mapping function (implemented by a Taylor series approximation circuit). The mapping result is stored at a specified address in the output buffer. This data corresponds to the predicted temperature values ​​of all nodes for a future set time span, preferably 200s to 300s.

[0088] To further explain, a parallel computing pipeline is constructed within the computational core unit of the matrix operation coprocessor: the input vector elements (n-dimensional) and the first column elements of the parameter matrix (n-dimensional) undergo synchronous dot product operations through a hardware multiplier array. This array consists of n groups of 18×18-bit fixed-point Booth multiplier ASICs. Each multiplier independently receives the input vector elements and the first column elements of the parameter matrix, triggering the multiplication operation on the rising edge of clock CLK1, and outputting the product after 3 clock cycles. All products are input to a Kogge-Stone carry-lookahead adder tree, which employs a four-level structure. The pipeline architecture consists of the following stages: the first stage performs carry-preserving addition (CSA) on pairs of n products to generate sum and carry vectors; the second stage merges the sum / carry vectors into intermediate results using a 3:2 compressor; the third stage achieves cross-bit propagation via a Manchester carry chain; the fourth stage finally outputs the first element of the feature matrix; the computation controller then updates the parameter matrix column address pointers, repeats the dot multiplication and addition tree accumulation process with the input vector and the second column element of the parameter matrix, iterating until a complete m×1 dimensional feature matrix is ​​generated, with the number of iterations preset in the status register by the number of columns m of the parameter matrix.

[0089] In one embodiment of S4, the step of generating a current-carrying capacity correction command when the predicted temperature value exceeds the allowable temperature rise value of the busbar material includes:

[0090] The predicted temperature T of all nodes predThe allowable temperature rise value T of the busbar material allow Compare;

[0091] When the predicted temperature of any target node exceeds the allowable temperature rise value, a current carrying capacity correction instruction containing the identifier of the feeder loop associated with that node is generated, wherein the correction instruction contains the current amplitude value to be reduced.

[0092] In this embodiment, a multi-channel digital comparator array is deployed in the safety control module of the busbar cabinet. This array contains independent comparison units equal to the number of busbar nodes. Each unit receives two inputs: one from the predicted future temperature value of the target node in the prediction buffer, and the other connected to the allowable temperature rise value of the busbar material pre-programmed in a read-only memory. The comparator performs a binary amplitude comparison operation; when T... pred > T allow A high-level over-limit signal is output; all over-limit signals at nodes are connected to a priority encoder, which generates the address of the highest-risk node based on the thermal hazard level of the node's physical location (connection terminal > bending area > straight section); this address is input into the feeder loop mapping table (stored in EPROM) to retrieve the associated feeder loop identifier; simultaneously, the temperature rise deviation value ΔT = T of the over-limit node is recorded. pred - T allow The input current correction calculation unit calculates the required reduction in current based on a simplified formula derived from Ohm's law and the Joule heating effect: ΔI = k×ΔT (where k is the reciprocal of the temperature coefficient of resistance of the busbar, preset through the material parameter table). Finally, the loop identifier and the current amplitude value are encapsulated into a current carrying capacity correction instruction frame. The frame structure includes a start symbol 0xAA, a loop identifier code (2 bytes), a current amplitude value (IEEE 754 floating-point number), and a CRC checksum, and is sent to the adjustment execution unit via the RS-485 bus.

[0093] In one embodiment of S5, the step of reducing the feeder loop current according to the current-carrying correction command until the predicted temperature value is lower than the allowable temperature rise value includes:

[0094] Match the target thyristor regulator according to the feeder loop identifier in the correction instruction;

[0095] The thyristor regulator is controlled to reduce the output current by the required reduction value.

[0096] In this embodiment, a cluster of intelligent thyristor regulators is deployed in the power distribution unit of the row head cabinet. Each regulator is connected to a specific feeder circuit via hardwire. When a current carrying capacity correction command is received, the command parsing module first extracts the circuit identification code, converts it into a physical slot selection signal through a hardware address decoder, and activates the control port of the target thyristor regulator. The current regulation process adopts phase angle control technology: the zero-crossing detection circuit inside the regulator monitors the AC voltage waveform in real time, and starts a delay counter at the same time as detecting the zero crossing point. The preset value of the counter is converted from the current amplitude value ΔI to the corresponding trigger value through a lookup table method. Delay angle; when the counter reaches the set value, it triggers the pulse transformer to generate a drive pulse with a defined phase shift, controlling the bidirectional thyristor to work within the set conduction angle range; the actual output current is monitored in real time by the Hall sensor in the closed-loop feedback loop, and the measured value is sent to the PID compensator (composed of an analog operational amplifier TL084) to compare with the command value. Its output dynamically adjusts the trigger delay angle until the measured current stabilizes within ±1% of the target value; after the adjustment is completed, the system automatically triggers a new round of temperature prediction. When the predicted value still exceeds the limit, this process is iteratively executed until the predicted temperature values ​​of all nodes are lower than the allowable temperature rise value.

[0097] In one embodiment, the step of matching the target SCR regulator according to the feeder loop identifier in the correction instruction includes:

[0098] Based on the determination of any target node location exceeding the temperature rise value, read the preset mapping table between the feeder loop identifier and the connection port of the thyristor regulator;

[0099] The physical control port of the target thyristor regulator is located through the mapping table, and the physical control port is mapped to the node location.

[0100] In this embodiment, a programmable logic device (PLD) is deployed on the back panel of the column head cabinet. Internally, a feeder loop-thyristor regulator mapping table is embedded. This table uses the physical coordinates of the busbar nodes as index keys (e.g., node K1 corresponds to coordinates (x1, y1)) and stores three sets of associated parameters: target feeder loop identifier (ASCII encoding), thyristor regulator slot number (4-bit binary), and control port electrical characteristics (including drive voltage 12V / trigger current 50mA, etc.). When an over-temperature node position signal is received, the CPLD performs a two-stage addressing operation: first, it uses the node coordinate indexer (with a built-in two-dimensional lookup table)... The target feeder loop is located, and the loop identifier is converted into an 8-bit address bus signal by a hardware decoder. Then, the address signal is converted into a strobe signal for the physical control port by a slot distributor. The strobe signal is isolated by an optocoupler and then drives the corresponding thyristor regulator control loop. The control port adopts a design to prevent mis-plugging. Each port includes a mechanical keyway positioning structure and an electrical interlock pin (which triggers a normally closed relay to cut off the output when not connected correctly) to ensure the reliability of the physical connection. Finally, the status of the activated control port is displayed in real time by LED indicator (red / green dual color) and fed back to the central monitoring unit to generate an operation log.

[0101] In one embodiment, the pre-training method for the temperature prediction model includes:

[0102] Under normal operating conditions, the temperature values ​​of key nodes of the busbar are continuously collected by distributed contact temperature sensors, and the current carrying capacity data is synchronously measured by the current transformer of the feeder circuit. A complete sequence data package containing the first temperature sequence data with the current timestamp and its bound current carrying capacity data, as well as the second temperature sequence data corresponding to the previous cycle timestamp, is generated at fixed time intervals.

[0103] The sequence data packets are stored in the local cache area of ​​the column head cabinet in real time. When the amount of cached data reaches the preset storage capacity, the model training enable signal is triggered.

[0104] In response to the enable signal, all sequence data packets are extracted from the buffer, and the second temperature sequence data in each data packet is extracted as the input feature set, and the first temperature sequence data corresponding to the timestamp is used as the output tag set.

[0105] The initial model parameter matrix is ​​constructed and loaded into the matrix operation processor built into the column head cabinet. The input feature set is input into the processor in chronological order to perform matrix multiplication and addition operations, generating a preliminary predicted temperature sequence.

[0106] The absolute error between the preliminary predicted temperature sequence and the output label set is calculated. When the average absolute error exceeds the allowable error threshold, the parameter correction amount is generated through the error backpropagation controller.

[0107] Update the model parameter matrix using the parameter correction amount and repeat the matrix operation and error calculation process until the mean absolute error stabilizes within the allowable error threshold.

[0108] The final optimized model parameter matrix is ​​burned into the read-only memory of the column head cabinet, the training enable signal is turned off, and the buffer is cleared.

[0109] In this embodiment, a dedicated training subsystem is integrated into the control compartment of the busbar control panel. This system consists of a data acquisition module, a cache management unit, a matrix operation accelerator, and a parameter programming interface. During normal operation of the busbar, distributed contact temperature sensors collect key node temperature values ​​at 5-second intervals, while synchronous feeder loop current transformers measure current-carrying data. The data packaging engine encapsulates the current node temperature sequence, corresponding current-carrying capacity, and historical temperature sequence from the same time point in the previous cycle (marked as the second temperature sequence) into a 512-byte data packet. The packet header includes an IRIG-B timestamp, the number of nodes, and a CRC checksum. The data packet is transmitted to a dual-port buffer (8GB capacity, organized in a circular queue structure) via a DMA channel. When the amount of buffered data reaches a preset threshold (e.g., 6GB), a hardware counter overflows, triggering a training enable signal (active high).

[0110] The training enable signal activates the dedicated computing channel of the matrix operation processor (Xilinx Versal ACAP), and its workflow consists of four stages:

[0111] The parser reads data packets sequentially from the buffer, extracts the second temperature sequence from each packet as the input feature matrix X (dimension N×n, where N is the number of samples and n is the number of nodes), and outputs the corresponding first temperature sequence as the label matrix Y (dimension N×n).

[0112] The feature matrix X and the initial parameter matrix W (initial values ​​are random Gaussian distributions) are input into the MAC (multiply-accumulate) operation unit to perform a linear transformation:

[0113]

[0114] The bias matrix B is pre-stored in the register file, and the calculation result Z is nonlinearly mapped by the Sigmoid function hardware circuit (implemented by Taylor series approximation) to generate the prediction matrix. ;

[0115] The comparator calculates the predicted value. The absolute error matrix E of label Y, when the average error When the temperature exceeds a threshold (e.g., 2.0℃), the error backpropagation controller (a differentiating circuit based on an analog operational amplifier) ​​generates a parameter gradient:

[0116]

[0117] in The learning rate (adjustable by a hardware potentiometer). For the output of the Sigmoid derivative circuit;

[0118] The gradient ΔW updates the parameter matrix W←W through the accumulator. ΔW, iteratively execute the above process until the average error stabilizes within the threshold.

[0119] After training is complete, the parameter matrix W is burned into the non-volatile region of the ferroelectric memory via the JTAG interface, and the cache pointer and status register are reset simultaneously. The entire training process is monitored by a hardware watchdog, which forcibly terminates the training upon timeout and issues an alarm.

[0120] Reference Appendix Figure 2 The present invention provides a device for predicting the temperature rise and dynamically adjusting the current of a busbar in a row-head cabinet. The device includes:

[0121] The acquisition unit is used to acquire the surface temperature sequence of the busbar and the corresponding current carrying capacity data at the time, and generate sequence data;

[0122] A calculation unit is used to calculate the temperature rise rate per unit time based on the sequence data;

[0123] The model unit is used to input the temperature rise rate into a pre-trained temperature prediction model and obtain the predicted busbar temperature value under a future set time period output by the temperature prediction model.

[0124] The instruction unit is used to generate a current carrying capacity correction instruction when the predicted temperature value exceeds the allowable temperature rise value of the busbar material, wherein the threshold value of the temperature rise value is preset in advance;

[0125] The action unit is configured to reduce the feeder loop current according to the current carrying capacity correction command until the predicted temperature value is lower than the allowable temperature rise value.

[0126] In summary, the method of this invention generates sequence data by acquiring the surface temperature sequence of the busbar and the corresponding current-carrying data at the corresponding time; calculates the temperature rise rate per unit time based on the sequence data; inputs the temperature rise rate into a pre-trained temperature prediction model to obtain the predicted busbar temperature value for a future set time period output by the temperature prediction model; when the predicted temperature value exceeds the allowable temperature rise value of the busbar material, a current-carrying correction instruction is generated, wherein the threshold of the temperature rise value is preset in advance; and the feeder circuit current is reduced according to the current-carrying correction instruction until the predicted temperature value is lower than the allowable temperature rise value; thereby achieving the effect of initiating adjustment before overheating occurs through temperature rise trend prediction, and avoiding thermal deformation of the busbar material.

[0127] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for predicting temperature rise and dynamically adjusting current carrying capacity of busbars in a power supply unit, characterized in that, Includes the following steps: Obtain the surface temperature sequence of the busbar and the corresponding current carrying capacity data at the corresponding time, and generate sequence data; The rate of temperature rise per unit time is calculated based on the sequence data; The temperature rise rate is input into a pre-trained temperature prediction model to obtain the busbar temperature prediction value for a future set time period output by the temperature prediction model. This includes combining the temperature rise rate of all nodes at the current timestamp with the bound current carrying capacity data into a model input vector; inputting the input vector into the pre-trained temperature prediction model; and obtaining the temperature prediction value of all nodes for a future set time span output by the temperature prediction model. The prediction process of the temperature prediction model is as follows: the temperature rise rate and current carrying capacity data in the input vector are combined linearly with weighted summation to generate a feature matrix; the feature matrix is ​​multiplied and summed with a pre-stored model parameter matrix; and the temperature prediction value of all nodes for a future set time span is output through a linear mapping function. When the predicted temperature exceeds the allowable temperature rise value of the busbar material, a current carrying capacity correction command is generated, and the threshold value of the temperature rise value is preset in advance. The feeder loop current is reduced according to the current carrying capacity correction command until the predicted temperature value is lower than the allowable temperature rise value.

2. The method for predicting the temperature rise and dynamically adjusting the current carrying capacity of the busbar in the head unit according to claim 1, characterized in that, The steps for obtaining the busbar surface temperature sequence include: Temperature values ​​of nodes on the inner wall of the busbar are collected through distributed contacts of pre-set temperature sensors inside the busbar. Record the temperature measurements at each node at fixed time intervals; The temperature values ​​of each node at the same timestamp are integrated into the first temperature sequence data. The second temperature sequence data corresponding to the previous period timestamp of the first temperature sequence data is retained, while the temperature sequence content of other timestamps is forgotten.

3. The method for predicting the temperature rise and dynamically adjusting the current carrying capacity of the busbar in the head unit according to claim 2, characterized in that, The steps for obtaining the current carrying capacity data of the busbar at a corresponding time and generating sequence data include: The current value of each circuit is measured in real time by pre-installing current transformers in the feeder circuits inside the cabinet busbar. The current values ​​of each circuit are converted into digital signals of current carrying capacity, and the mapping relationship between the conversion timestamp and the current carrying capacity data is recorded; According to the timestamp of temperature acquisition, the first temperature sequence data is bound to the current carrying capacity digital signal to obtain the sequence data.

4. The method for predicting the temperature rise and dynamically adjusting the current carrying capacity of the busbar in the head unit according to claim 3, characterized in that, The step of calculating the temperature rise rate per unit time based on the sequence data includes: Extract the first temperature sequence data with the current timestamp and its associated current carrying capacity data from the sequence data; Obtain the second temperature sequence data from the previous period's timestamp; Calculate the temperature difference between all nodes of the busbar in the first temperature sequence data and the second temperature sequence data; Divide the temperature difference by a fixed time interval to obtain the temperature rise rate of all nodes of the busbar.

5. The method for predicting the temperature rise and dynamically adjusting the current carrying capacity of the busbar in the head unit according to claim 1, characterized in that, When the predicted temperature exceeds the allowable temperature rise value of the busbar material, the step of generating a current carrying capacity correction command includes: Compare the predicted temperature values ​​of all nodes with the allowable temperature rise values ​​of the busbar material; When the predicted temperature of any target node exceeds the allowable temperature rise value, a current carrying capacity correction instruction containing the identifier of the feeder loop associated with that node is generated, wherein the correction instruction contains the current amplitude value to be reduced.

6. The method for predicting the temperature rise and dynamically adjusting the current carrying capacity of the busbar in the head unit according to claim 5, characterized in that, The step of reducing the feeder loop current according to the current carrying capacity correction command until the predicted temperature value is lower than the allowable temperature rise value includes: Match the target thyristor regulator according to the feeder loop identifier in the correction instruction; The thyristor regulator is controlled to reduce the output current by the required reduction value.

7. The method for predicting the temperature rise and dynamically adjusting the current carrying capacity of the busbar in the head unit according to claim 6, characterized in that, The step of matching the target thyristor regulator according to the feeder loop identifier in the correction instruction includes: Based on the determination of any target node location exceeding the temperature rise value, read the preset mapping table between the feeder loop identifier and the connection port of the thyristor regulator; The physical control port of the target thyristor regulator is located through the mapping table, and the physical control port is mapped to the node location.

8. The method for predicting the temperature rise and dynamically adjusting the current carrying capacity of the busbar in the head unit according to claim 1, characterized in that, The pre-training method for the temperature prediction model includes: Under normal operating conditions, the temperature values ​​of key nodes of the busbar are continuously collected by distributed contact temperature sensors, and the current carrying capacity data is synchronously measured by the current transformer of the feeder circuit. A complete sequence data package containing the first temperature sequence data with the current timestamp and its bound current carrying capacity data, as well as the second temperature sequence data corresponding to the previous cycle timestamp, is generated at fixed time intervals. The sequence data packets are stored in the local cache area of ​​the column head cabinet in real time. When the amount of cached data reaches the preset storage capacity, the model training enable signal is triggered. In response to the enable signal, all sequence data packets are extracted from the buffer, and the second temperature sequence data in each data packet is extracted as the input feature set, and the first temperature sequence data corresponding to the timestamp is used as the output tag set. The initial model parameter matrix is ​​constructed and loaded into the matrix operation processor built into the column head cabinet. The input feature set is input into the processor in chronological order to perform matrix multiplication and addition operations, generating a preliminary predicted temperature sequence. The absolute error between the preliminary predicted temperature sequence and the output label set is calculated. When the average absolute error exceeds the allowable error threshold, the parameter correction amount is generated through the error backpropagation controller. Update the model parameter matrix using the parameter correction amount and repeat the matrix operation and error calculation process until the mean absolute error stabilizes within the allowable error threshold. The final optimized model parameter matrix is ​​burned into the read-only memory of the column head cabinet, the training enable signal is turned off, and the buffer is cleared.

9. A device for predicting temperature rise and dynamically adjusting current carrying capacity of busbars in a row-head cabinet, characterized in that, include: The acquisition unit is used to acquire the surface temperature sequence of the busbar and the corresponding current carrying capacity data at the time, and generate sequence data; A calculation unit is used to calculate the temperature rise rate per unit time based on the sequence data; The model unit is used to input the temperature rise rate into a pre-trained temperature prediction model and obtain the predicted busbar temperature for a future set time period output by the temperature prediction model. This includes: combining the temperature rise rates of all nodes at the current timestamp with the bound current carrying capacity data into a model input vector; inputting the input vector into the pre-trained temperature prediction model; and obtaining the predicted temperature values ​​of all nodes for a future set time span output by the temperature prediction model. The prediction process of the temperature prediction model is as follows: generating a feature matrix by linearly weighting the temperature rise rate and current carrying capacity data in the input vector; performing multiplication and accumulation operations on the feature matrix and a pre-stored model parameter matrix; and outputting the predicted temperature values ​​of all nodes for a future set time span through a linear mapping function. The instruction unit is used to generate a current carrying capacity correction instruction when the predicted temperature value exceeds the allowable temperature rise value of the busbar material, wherein the threshold value of the temperature rise value is preset in advance; The action unit is configured to reduce the feeder loop current according to the current carrying capacity correction command until the predicted temperature value is lower than the allowable temperature rise value.

Images