Self-adaptive intelligent protection distribution box and control method

Abstract

The invention discloses a self-adaptive intelligent protection distribution box and a control method. The control method comprises the following steps: S1, acquiring environment and operation parameters; s2, performing noise reduction and normalization processing on the acquired data, and extracting feature parameters; s3, calculating an environment-load coupling coefficient K quantitative risk level, predicting future temperature, load rate and risk level through a fusion model, and calling cloud data for calibration when abnormity occurs; s4, generating a dynamic threshold, and generating a hierarchical control instruction through a multi-target weighted decision model; s5, the adaptive protection execution module executes the control instruction and feeds back an execution state in real time; and S6, uploading the data to a cloud management platform, and optimizing parameters and threshold calculation rules. According to the adaptive intelligent protection distribution box and the control method, precise predictive protection of environment-load coupling can be realized, and the protection precision, the power supply stability, the energy consumption optimization capability and the scene adaptability of the distribution box are improved.

Description

Technical Field

[0001] This invention relates to the field of electrical equipment protection technology, and in particular to an adaptive intelligent protection distribution box and control method. Background Technology

[0002] Currently, distribution boxes, as core equipment for power distribution and control, are widely used in industrial production, construction engineering, and new energy fields. Existing distribution boxes mostly employ passive protection methods, which trigger a single protective action based on preset fixed thresholds (such as activating a cooling fan when the temperature reaches a threshold, or directly tripping the circuit breaker when overloaded). This approach has the following technical drawbacks: (1) Single protection dimension: It only controls a few parameters such as temperature and overload, ignoring the coupled influence of multiple factors such as dust, humidity and external impact, which leads to protection failure in complex environments.

[0003] (2) Response lag: The “threshold triggering” mode cannot predict the trend of risk changes and does not quantify the synergistic effect between parameters, which can easily lead to untimely protection.

[0004] (3) Rigid protection methods: heat dissipation, sealing and other protection methods are fixed, threshold parameters cannot be dynamically adjusted, and cannot be adapted to equipment aging and scene differences (such as industrial dust environment and outdoor rain environment), resulting in excessive energy consumption or insufficient protection.

[0005] (4) Crude fault handling: When faults such as short circuits and arcs occur, the overall power outage or single tripping method is often adopted, without graded response, which affects the normal operation of non-faulty branches and reduces the reliability of power supply.

[0006] (5) Insufficient prediction accuracy: Traditional single prediction models do not highlight the impact of key parameters and have large prediction errors for coupled scenarios such as temperature-load and dust-sealing, which cannot support accurate protection.

[0007] Therefore, there is an urgent need to design an intelligent distribution box to address the shortcomings of existing technologies. Summary of the Invention

[0008] The purpose of this invention is to provide an adaptive intelligent protection distribution box and control method. Through dual-model fusion prediction, coupling coefficient quantization, dynamic threshold self-optimization and fault graded response, it achieves accurate predictive protection of environment-load coupling, and improves the protection accuracy, power supply stability, energy consumption optimization capability and scenario adaptability of the distribution box.

[0009] To achieve the above objectives, the present invention provides an adaptive intelligent protection distribution box, including a box body, wherein the box body is provided with: The multi-dimensional sensing module includes a temperature sensor, a humidity sensor, a dust concentration sensor, an arc detection sensor, a three-phase current / voltage sensor, a vibration sensor, and a raindrop sensor, which collect environmental parameters and equipment operating parameters respectively. The adaptive protection execution module includes an adjustable speed ventilation and filtration unit, a multi-mode heat dissipation unit, an electromagnetically driven sealing valve, a load balancing unit, a miniature aerosol fire extinguishing unit, and a branch disconnect switch. The intelligent control module includes a built-in data preprocessing unit, an Attention-LSTM fusion prediction model unit, a coupling coefficient calculation unit, a dynamic threshold generation unit, and a control strategy generation unit. The communication module integrates a 5G module and an Ethernet interface, supports MQTT / HTTP communication protocols, and enables bidirectional data interaction with the cloud management platform and the upper-level power grid.

[0010] Preferably, the enclosure is equipped with a filter ventilation port, and the interior of the enclosure is divided into a main chamber and independent branch chambers. The independent branch chambers are respectively equipped with a miniature aerosol fire extinguishing unit and an electromagnetically driven isolating switch to achieve precise fault isolation.

[0011] Preferably, the adjustable speed ventilation and filtration unit uses a brushless DC motor, the multi-mode heat dissipation unit includes air cooling, heat pipes and micro water cooling components, and the micro aerosol fire extinguishing unit uses non-toxic and environmentally friendly aerosol.

[0012] This invention also provides a control method for an adaptive intelligent protection distribution box, comprising the following steps: S1. The multi-dimensional sensing module collects environmental and operating parameters at a preset frequency and transmits them to the intelligent control module; S2. The data preprocessing unit performs noise reduction and normalization on the collected data and extracts feature parameters. S3. First, calculate the environment-load coupling coefficient K to quantify the risk level, and then use the Attention-LSTM fusion model to predict the future temperature, load rate and risk level. In case of anomalies, call cloud data for calibration. S4. Based on equipment runtime, failure frequency and coupling risk level, generate dynamic thresholds, and generate hierarchical control commands through a multi-objective weighted decision model; S5. The intelligent control module issues control commands, the adaptive protection execution module executes the control commands, and the execution status is fed back to the intelligent control module in real time. S6. The communication module uploads the operating status, prediction results, coupling coefficient, dynamic threshold and control instructions to the cloud management platform. At the same time, it receives the upper-level power grid dispatch instructions and the cloud-optimized weight coefficients, and dynamically updates the model parameters and threshold calculation rules.

[0013] Preferably, in S2, the noise reduction process uses a wavelet thresholding algorithm with the following parameters: selecting a db4 wavelet basis, decomposing into 3 layers, and setting a threshold. ,in, The standard deviation of noise. Given the data length, the signal-to-noise ratio after noise reduction should be ≥30dB; The normalization process uses the Min-Max normalization algorithm, and the formula is: X norm =(XX min ) / (X max -X min ); Map all parameters to the interval [0, 1], where X of each parameter... min / X max The preset sensor measurement range; Feature parameter extraction specifically involves: Temperature change rate ΔT / Δt: Calculate the ratio of the temperature difference to the time difference over 10 consecutive sampling periods, and take the average value; Load factor L: According to the formula L=(I A +I B +I C ) / (3×I rated ) Calculate, rounded to one decimal place, where I rated This is the rated current of the branch circuit; Dust accumulation D acc The formula for integrating the normalized dust concentration over 1 hour is: ; Arc signal strength A s Multiply the normalized peak value of the arc signal by 100 and round to the nearest integer. Vibration frequency f: The vibration acceleration time-domain signal is converted into a frequency-domain signal by fast Fourier transform, and the peak frequency is extracted.

[0014] Preferably, in S3, the environment-load coupling coefficient K is first calculated to quantify the risk level, specifically as follows: Weight coefficient initialization: default α=0.35, β=0.4, γ=0.15, δ=0.1, and the total weight is 1; Coefficient calculation logic: First, normalize each feature parameter to the [0, 1] interval, then substitute them into the formula: K = α × (ΔT / Δt) norm +β×L norm +γ×(D acc / 100) norm +δ×A s norm ; The calculation result is rounded to 3 decimal places, K [0, 1]; Risk level determination: K < 0.4 is low risk, 0.4 ≤ K < 0.7 is medium risk, and K ≥ 0.7 is high risk.

[0015] Preferably, in S3, the Attention-LSTM fusion model is used to predict the temperature, load rate, and risk level for the next 10-60 minutes as follows: Running the Attention-LSTM fusion model: Model Structure: LSTM Layer: 9-dimensional input, 64 hidden neurons, tanh activation function, dropout rate 0.2, 50 iterations; Attention Layer: Uses additive attention mechanism to calculate the similarity between the input vector and the query vector, considering factors such as load rate L, temperature change rate ΔT / Δt, and arc signal intensity A. s Assign dynamic weights; Model training mechanism: Local incremental training is performed every 24 hours using the latest 1000 sets of data to update model weights; global optimal parameters are received from the cloud every 7 days to overwrite local parameters. Prediction results output: Predicted duration selection: Supports five adjustable levels of 10 / 20 / 30 / 45 / 60 minutes, which can be configured via the cloud according to the needs of the scenario; Output: Predicted temperature value T pred Load rate prediction value L pred Coupling risk level K pred ; Accuracy Guarantee: Predicted temperature error ≤ ±2℃, load rate prediction error ≤ ±3%.

[0016] Preferably, in S3, the specific procedure for retrieving cloud data for calibration in case of an anomaly is as follows: Anomaly detection criteria: Compare the prediction results with the average value of historical data for the same period. If the deviation is ≥30%, it is considered an anomaly. Correction process: The communication module sends an anomaly flag and current scene parameters to the cloud. The cloud then retrieves historical prediction data and correction models for similar scenarios, and issues correction coefficients. The local system corrects the error according to the formula below: T pred corrected =T pred ×(1+Δk); L pred corrected =L pred ×(1+Δk); Among them, T pred corrected For the corrected temperature prediction, L pred corrected The corrected load rate prediction value is given by Δk, which is the correction coefficient issued by the cloud, ranging from -0.1 to 0.1.

[0017] Preferably, in S4, the dynamic threshold generated based on device runtime, failure frequency, and coupling risk level is specifically as follows: Basic parameter acquisition: runtime t of the intelligent control module's local storage device, historical failure frequency N; Threshold calculation logic: T1 = 45 - 0.005 × t. If T1 < 40, then take 40℃. T1 is the lower limit of the dynamic temperature threshold. T2 = 60 - 0.008 × t. If T2 < 55, then take 55℃. T2 is the upper limit of the dynamic temperature threshold. L th =70-0.01×N, if L th If <60, then take 60%, L th This is the dynamic load rate threshold; D1 = 2 + 0.1 × K pred Keep one decimal place, and D1 is the lower limit of the dynamic dust threshold; D2 = 5 + 0.3 × K pred Keep one decimal place; D2 is the upper limit of the dynamic dust threshold; V th =0.5+0.1×K pred Round to two decimal places; V th The dynamic vibration safety threshold; Threshold storage update: The dynamic threshold is recalculated once per hour, overwriting historical data and synchronously uploaded to the cloud for backup.

[0018] Preferably, in S4, the generation of hierarchical control instructions through a multi-objective weighted decision model specifically involves: Target weight configuration: Default protection effectiveness weight W1=0.4, energy consumption optimization weight W2=0.3, power supply continuity weight W3=0.3, supports scenario-based adjustment; Decision-making logic and instruction details: Low-coupling-risk scenario: Judgment condition: K pred <0.4 and T pred ≤T1 and L pred <L th Control commands: Air-cooled mode activated, vents opened, load balancing unit in standby mode, fire suppression unit ready; Energy optimization measures: If there are no parameter fluctuations for 30 consecutive minutes, the fan speed is reduced to 0.3 m / s; Intermediate coupling risk scenario: Judgment condition: 0.4≤K pred <0.7 or T1<T pred ≤T2 or L th ≤L pred<85%; Control commands: Air cooling + heat pipe coordinated start-up; if the real-time dust concentration D < D2, the vents will open; the load balancing unit monitors non-critical loads in real time; Power supply continuity guarantee: the migration priority of non-critical loads is lower than that of critical loads. High coupling risk scenario: Judgment condition: K pred ≥0.7 or T pred >T2 or L pred ≥85%; Control commands: Three-stage cooling system (water cooling + air cooling + heat pipes) activated; if real-time dust concentration D ≥ D2, the sealing valve shuts off and internal circulation switches; the load balancing unit forcibly migrates non-critical loads; Enhanced protection measures: Vibration V ≥ V th At times, an additional cavity reinforcement electromagnetic lock is activated to prevent external impacts from causing structural loosening; Arc fault classification response: minor arc A s <0.3: Send a load reduction command, target load rate ≤50%, increase the sampling frequency of the arc detection sensor to 5Hz, monitor continuously for 10 minutes, and restore the original load if no abnormality is found; for moderate arcs 0.3≤A s <0.7: Immediately send a branch isolating switch disconnect command, the fire extinguishing unit enters standby mode, the aerosol spray nozzle is unlocked, and a fault alarm is uploaded simultaneously; Severe arcing A s ≥0.7: When the isolating switch is disconnected, a fire extinguishing unit activation command is sent, which only applies to the faulty branch chamber, while other branches are powered normally; External force impact response: Judgment condition: Vibration magnitude V ≥ V th And the vibration frequency f [10-100Hz]; Control commands: Close the electromagnetic drive sealing valve, reduce the load rate of non-critical branches to below 40%, continuously monitor the vibration sensor for 5 minutes, and if the vibration level drops to V... th If the following steps are taken, the sealing valve will be restored to the open state; otherwise, the protection state will be maintained and an alarm will be uploaded.

[0019] Therefore, the beneficial effects of the above-mentioned adaptive intelligent protection distribution box and control method of the present invention are as follows: (1) This invention uses coupling coefficient calculation and dual-model fusion prediction to predict multi-factor collaborative risks 10-60 minutes in advance, with a prediction error of ≤5%, making the protection response more targeted.

[0020] (2) The present invention adopts a dynamic threshold self-optimization mechanism to adapt to equipment aging and scene differences, avoiding over-protection or insufficient protection; adopts arc fault graded processing, which can reduce unnecessary branch disconnection and improve power supply continuity; adopts local edge computing + cloud collaborative optimization, continuously improving model accuracy, supporting remote monitoring and strategy adjustment, and significantly improving operation and maintenance efficiency.

[0021] (3) The multi-objective weighted decision model of the present invention takes into account both protection and energy consumption. In low-risk scenarios, it automatically reduces the heat dissipation wind speed and shuts down redundant heat dissipation units, resulting in a significant reduction in energy consumption.

[0022] (4) This invention supports cloud-based adjustment of weight coefficients and target weights to meet the personalized needs of different scenarios such as industrial plants, outdoor base stations, and new energy power plants.

[0023] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0024] Figure 1 This is a schematic diagram illustrating the composition of an embodiment of an adaptive intelligent protection distribution box according to the present invention; Figure 2 This is a schematic diagram illustrating the steps of an embodiment of the control method for an adaptive intelligent protection distribution box according to the present invention. Detailed Implementation

[0025] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0026] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0027] Example 1: like Figure 1 As shown, the present invention provides an adaptive intelligent protection distribution box, including a box body made of high-strength flame-retardant material. The box body is divided into a main chamber and independent branch chambers. The independent branch chambers are respectively equipped with a micro aerosol fire extinguishing unit and an electromagnetic drive isolating switch to achieve precise fault isolation. The detachable filter ventilation port is equipped with a metal filter screen and an activated carbon filter layer, which takes into account both ventilation and primary dust prevention. The filter layer can be removed and replaced individually.

[0028] The enclosure houses a multi-dimensional sensing module, an adaptive protection execution module, an intelligent control module, and a communication module. The multi-dimensional sensing module integrates eight types of high-precision sensors, covering all environmental and operational parameters: temperature sensor (measurement range -40℃-125℃, accuracy ±0.5℃), humidity sensor (measurement range 0-100%RH, accuracy ±3%RH), and dust concentration sensor (measurement range 0-10mg / m³). 3 Accuracy ±0.1 mg / m 3 The system includes a raindrop sensor (detection sensitivity ≥ 0.1 mm / h), a vibration sensor (measurement range 0-5 g, accuracy ± 0.01 g), a three-phase current / voltage sensor (current measurement range 0-1000 A, voltage 0-10 kV, accuracy ± 0.2%), and an arc detection sensor (response time ≤ 10 μs, detection range 0-10 kV), with a sampling frequency of 1-5 Hz (adjustable via the intelligent control module) to meet real-time monitoring requirements.

[0029] The adaptive protection execution module includes: Adjustable speed ventilation and filtration unit: adopts a brushless DC motor with stepless adjustment of wind speed from 0.5-2m / s to adapt to different dust concentrations and heat dissipation requirements.

[0030] Multi-mode cooling unit: Air cooling (axial fan, airflow ≥100m³ / h) 3 / h) + heat pipe (copper heat pipe array, thermal conductivity ≥400W / (m·K)) + micro water cooling (mini circulating water pump + copper heat dissipation coil, heat dissipation power ≥500W) to achieve three-level gradient heat dissipation.

[0031] Electromagnetically driven sealing valve: Employs an electromagnetic actuator with a response time ≤0.3s, a sealing rating up to IP65, and an air leakage rate ≤0.1m when closed. 3 / h.

[0032] Load balancing unit: The intelligent power distributor supports dynamic switching of 3-8 branches, with a load adjustment response time of ≤1s and a rated carrying current of ≥200A.

[0033] Miniature aerosol fire extinguishing unit: Uses non-toxic and environmentally friendly aerosol (free of halon), with the spray volume precisely matched to the branch chamber volume (≥50g spray volume per 10L chamber) and the spray time ≤5s.

[0034] Branch disconnector: electromagnetically driven circuit breaker, breaking time ≤0.05s, rated short-circuit breaking capacity ≥50kA.

[0035] The intelligent control module uses STM32H743 as the main control chip (480MHz main frequency, 1MB memory), and has built-in data preprocessing unit, Attention-LSTM fusion prediction model unit (9-dimensional input, 64 neurons in the LSTM hidden layer, 2-dimensional output of the Attention layer), coupling coefficient calculation unit, dynamic threshold generation unit, and control strategy generation unit. It supports edge computing and has a local data processing latency of ≤100ms.

[0036] The communication module integrates a 5G module (supporting SA / NSA dual-mode, data transmission rate ≥100Mbps) and an Ethernet interface (Gigabit Ethernet), supports MQTT / HTTP communication protocols, and enables bidirectional data interaction with the cloud management platform and the upper-level power grid.

[0037] The core of the control method for the aforementioned distribution box lies in "precise prediction + dynamic adaptation + hierarchical response," such as... Figure 2 As shown, through the closed-loop collaboration of six steps S1-S6, multi-dimensional parameter acquisition, processing, prediction, decision-making, execution, and optimization are achieved. The specific technical solutions for each step are as follows: S1. The multi-dimensional sensing module collects environmental and operating parameters at a preset frequency and transmits them to the intelligent control module.

[0038] Acquisition triggering mechanism: The intelligent control module sends acquisition commands to the multi-dimensional sensing module at 1-5Hz, with a default of 3Hz. It supports dynamic adjustment of the sampling frequency according to the coupling risk level (automatically increasing to 5Hz when the risk is high and decreasing to 1Hz when the risk is low).

[0039] Data acquisition logic: Environmental parameters: The temperature sensor collects one set of temperature values ​​T every sampling period, the humidity sensor collects humidity values ​​H synchronously, the dust concentration sensor continuously collects 5 data points and takes the average value as D, the raindrop sensor outputs a digital signal R (0=no rain, 1=rain), and the vibration sensor collects vibration acceleration V (unit g) and calculates the vibration frequency f. Operating parameters: Three-phase current / voltage sensors synchronously acquire the A / B / C three-phase current I. A / I B / I C Voltage U A / U B / U C Calculate the three-phase imbalance (≤5% is normal), and the arc detection sensor collects the peak value A of the arc signal and records the signal duration.

[0040] Data transmission protocol: The collected data is transmitted to the intelligent control module using the SPI communication protocol (transmission rate ≥10Mbps). Each data frame contains sensor ID, collection timestamp, data value, and checksum (CRC32) to ensure data integrity (if the check fails, the data is collected again).

[0041] S2. The data preprocessing unit performs noise reduction and normalization on the collected data and extracts feature parameters, specifically: Noise reduction: A wavelet thresholding noise reduction algorithm is used, with the following parameters: db4 wavelet basis selected, 3 decomposition layers, and threshold value set. ,in, The standard deviation of noise. The data length is specified, and the signal-to-noise ratio after noise reduction is ≥30dB.

[0042] Normalization: The Min-Max normalization algorithm is used, and the formula is as follows: X norm =(XX min ) / (X max -X min ) Map all parameters to the interval [0, 1], where X of each parameter... min / X max Preset to the sensor measurement range (e.g., temperature X) min =-40℃, X max =125℃).

[0043] Feature parameter extraction specifically involves: Temperature change rate ΔT / Δt: Calculate the ratio of the temperature difference to the time difference over 10 consecutive sampling periods, and take the average value (unit: ℃ / min). Load factor L: According to the formula L=(I A +I B +I C ) / (3×I rated ) Calculate, rounded to one decimal place, where I rated This is the rated current of the branch circuit; Dust accumulation D acc The formula for integrating the normalized dust concentration over 1 hour is: Unit: mg·h / m 3 ; Arc signal strength A s Multiply the normalized peak value of the arc signal by 100 and round to the nearest integer (range 0-100, corresponding to A). s =0-1.0); Vibration frequency f: The vibration acceleration time-domain signal is converted into a frequency-domain signal by Fast Fourier Transform (FFT), and the peak frequency (unit: Hz) is extracted.

[0044] S3. First, calculate the environment-load coupling coefficient K to quantify the risk level, and then use the Attention-LSTM fusion model to predict the future temperature, load rate and risk level. In case of anomalies, call cloud data for calibration.

[0045] Calculation of the environment-load coupling coefficient K: Weight coefficient initialization: default α=0.35 (temperature change rate weight), β=0.4 (load rate weight), γ=0.15 (dust accumulation weight), δ=0.1 (arc signal intensity weight), and the total weight is 1.

[0046] Coefficient calculation logic: First, normalize each feature parameter to the [0, 1] interval, then substitute them into the formula: K = α × (ΔT / Δt) norm +β×L norm +γ×(D acc / 100) norm +δ×A s norm The calculation result is rounded to 3 decimal places, K [0, 1].

[0047] Risk level determination: K < 0.4 is low risk, 0.4 ≤ K < 0.7 is medium risk, and K ≥ 0.7 is high risk.

[0048] Running the Attention-LSTM fusion model: Model structure: LSTM layer: 9-dimensional input (5 feature parameters + coupling coefficient K + humidity H) norm +Vibration magnitude V norm +Raindrop signal R), number of hidden layer neurons 64, activation function is tanh, dropout rate is 0.2, number of iterations 50; Attention layer: Employs an additive attention mechanism to calculate the similarity between the input vector and the query vector, considering factors such as load rate L, temperature change rate ΔT / Δt, and arc signal intensity A. s Assign dynamic weights (weight coefficients are normalized using the softmax function, ranging from 0.2 to 0.5).

[0049] Model training mechanism: Local incremental training is performed every 24 hours using the latest 1000 sets of data to update model weights; global optimal parameters are received from the cloud every 7 days to overwrite local parameters.

[0050] Prediction results output: Predicted duration selection: Supports five adjustable levels of 10 / 20 / 30 / 45 / 60 minutes (default 30 minutes), which can be configured via the cloud according to the needs of the scenario.

[0051] Output: Predicted temperature value T pred (Outputs a temperature value every 5 minutes), load rate prediction value L pred Coupling risk level K pred (Consistent with the determination rule for the environment-load coupling coefficient K).

[0052] Accuracy Guarantee: Predicted temperature error ≤ ±2℃, load rate prediction error ≤ ±3%.

[0053] Anomaly correction mechanism: Anomaly detection criteria: Compare the predicted results with the average value of the data for the same period in the past 30 days. If the deviation is ≥30%, it is considered an anomaly.

[0054] Correction process: The communication module sends an anomaly flag and current scene parameters (temperature, humidity, dust, load, etc.) to the cloud. The cloud then retrieves historical prediction data and correction models from similar scenes (scene similarity ≥ 85%), and issues correction coefficients. The local system corrects according to the formula below: T pred corrected =T pred ×(1+Δk) L pred corrected =L pred ×(1+Δk) Among them, T pred corrected For the corrected temperature prediction, L pred corrected The corrected load rate prediction value is given by Δk, which is the correction coefficient issued by the cloud, ranging from -0.1 to 0.1.

[0055] S4. Based on equipment runtime, failure frequency and coupling risk level, generate dynamic thresholds, and generate hierarchical control commands through a multi-objective weighted decision model; Dynamic threshold calculation: Basic parameter acquisition: runtime t of the local storage device of the intelligent control module (cumulative from the first power-on, unit: days), historical fault frequency N (cumulative number of faults in the past year, N≤10).

[0056] Threshold calculation logic: T1 = 45 - 0.005 × t. If T1 < 40, then take 40℃. T1 is the lower limit of the dynamic temperature threshold. T2 = 60 - 0.008 × t. If T2 < 55, then take 55℃. T2 is the upper limit of the dynamic temperature threshold. L th =70-0.01×N, if L thIf <60, then take 60%, L th This is the dynamic load rate threshold; D1 = 2 + 0.1 × K pred Keep one decimal place, and D1 is the lower limit of the dynamic dust threshold; D2 = 5 + 0.3 × K pred Keep one decimal place; D2 is the upper limit of the dynamic dust threshold; V th =0.5+0.1×K pred Round to two decimal places; V th This is the dynamic vibration safety threshold.

[0057] Threshold storage update: The dynamic threshold is recalculated once per hour, overwriting historical data and synchronously uploaded to the cloud for backup.

[0058] Multi-objective weighted decision-making and control instruction generation: Target weight configuration: Default protection effectiveness weight W1=0.4, energy consumption optimization weight W2=0.3, power supply continuity weight W3=0.3, supports scenario-based adjustment (industrial plant W1=0.5, outdoor base station W2=0.4, new energy power station W3=0.4).

[0059] Decision-making logic and instruction details: Low-coupling risk scenarios: Judgment condition: K pred <0.4 and T pred ≤T1 and L pred <L th ; Control commands: Air-cooled mode activated (fan speed 0.5m / s), vents opened, load balancing unit in standby, fire suppression unit ready; Energy consumption optimization measures: If there are no parameter fluctuations for 30 consecutive minutes, the fan speed will be reduced to 0.3m / s.

[0060] In-coupling risk scenarios: Judgment condition: 0.4≤K pred <0.7 or T1<T pred ≤T2 or L th ≤L pred <85%; Control commands: Air cooling + heat pipe coordinated start (fan speed 1.2m / s, heat pipe heat sink fully deployed), if real-time dust concentration D < D2, the vents will open, and the load balancing unit will monitor non-critical loads in real time; Power continuity assurance: Non-critical load migration is given lower priority than critical load migration.

[0061] High-coupling-risk scenarios: Judgment condition: K pred≥0.7 or T pred >T2 or L pred ≥85%; Control commands: Start water cooling + air cooling + heat pipe three-stage heat dissipation (water pump speed 2000rpm, fan speed 2m / s), if the real-time dust concentration D≥D2, close the sealing valve and switch to internal circulation, and the load balancing unit will forcibly migrate non-critical loads; Protective reinforcement measures: Vibration amount V ≥ V th At times, an additional cavity reinforcement electromagnetic lock is activated to prevent external impacts from causing structural loosening.

[0062] Arc fault graded response: Slight electric arc A s <0.3: Send a load reduction command, with a target load rate ≤50%. Increase the sampling frequency of the arc detection sensor to 5Hz and monitor continuously for 10 minutes. If no abnormality is found, restore the original load. Moderate arc 0.3≤A s <0.7: Immediately send a branch isolating switch disconnect command, the fire extinguishing unit enters standby status, the aerosol spray nozzle is unlocked, and a fault alarm is uploaded simultaneously; Severe arc A s ≥0.7: At the same time as disconnecting the isolating switch, a fire extinguishing unit start command is sent (spraying time ≤5s), which only acts on the faulty branch chamber, while other branches are powered normally.

[0063] External impact response: Judgment condition: Vibration quantity V ≥ V th And the vibration frequency f [10-100Hz] (excluding environmental vibration interference); Control commands: Close the electromagnetically driven sealing valve (sealing rating IP65), reduce the load rate of non-critical branches to below 40%, continuously monitor with the vibration sensor for 5 minutes, and if the vibration level drops to V... th If the following steps are taken, the sealing valve will be restored to the open state; otherwise, the protection state will be maintained and an alarm will be uploaded.

[0064] S5. The intelligent control module issues control commands, the adaptive protection execution module executes the control commands, and the execution status is fed back to the intelligent control module in real time.

[0065] Command transmission: The intelligent control module sends control commands to the adaptive protection execution module via the CAN bus (transmission rate ≥ 500kbps). The command format is "Device ID + Command Code + Parameter + Check Bit".

[0066] Execution process monitoring: The heat dissipation unit provides real-time feedback on fan speed, water pump operating status, and heat pipe temperature. If the operating status does not match the command (e.g., the fan does not reach the set speed), it will retry within 3 seconds. If it fails to retry 3 times, it will report a fault. Sealing valve: Feedback on open / closed status (0 = closed, 1 = open), real-time monitoring of air leakage rate when closed (≤0.1m). 3 / h is normal); Load balancing unit: Feedback on branch load migration results (migration success / failure) and the current load rate of each branch; Fire extinguishing unit: provides feedback on ready status and spraying status, and provides feedback on the remaining amount of medium after spraying (reminds replacement when ≤20%).

[0067] Feedback data processing: The intelligent control module verifies the feedback data and stores key execution results (such as fault isolation time and heat dissipation mode switching records) for a duration of ≥90 days.

[0068] S6, cloud-based collaboration and parameter optimization.

[0069] Data Upload: The communication module uploads operational data at a frequency of once per minute, including: Real-time parameters: current environment parameters, running parameters, coupling coefficient K, dynamic threshold; Forecast data: temperature forecast curve, load rate forecast, risk level; Control data: Control command content, execution status, and feedback results; Upload protocol: MQTT protocol (QoS level 2) is used, and data is transmitted after compression (compression ratio ≥ 3:1).

[0070] Cloud-based interaction: Receive dispatch instructions: load adjustment instructions issued by the upper-level power grid (such as energy consumption control during peak and valley electricity price periods) and cloud-optimized weighting coefficients (α, β, γ, δ). Model parameter update: Receive the globally optimal parameters of the Attention-LSTM model from the cloud every 7 days, overwrite the old parameters locally and reinitialize the model; Threshold rule optimization: The cloud adjusts the coefficients in the dynamic threshold calculation formula based on device operation data (such as failure frequency and environmental adaptability) (e.g., increasing the weight of humidity on dust threshold in high humidity environments).

[0071] Local update: After receiving data from the cloud, the intelligent control module immediately updates the relevant parameters and synchronously stores the update log (including update time, update content, and cloud signature) to ensure traceability.

[0072] Therefore, the present invention adopts the above-mentioned adaptive intelligent protection distribution box and control method, which achieves precise, intelligent and scenario-based protection of the distribution box through the synergy of dual-model fusion prediction, coupled risk quantification, dynamic threshold self-optimization and fault graded response, solves many defects of traditional technology and has broad application prospects.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An adaptive intelligent protection distribution box, characterized in that, Includes a housing, and inside the housing are: The multi-dimensional sensing module includes a temperature sensor, a humidity sensor, a dust concentration sensor, an arc detection sensor, a three-phase current / voltage sensor, a vibration sensor, and a raindrop sensor, which collect environmental parameters and equipment operating parameters respectively. The adaptive protection execution module includes an adjustable speed ventilation and filtration unit, a multi-mode heat dissipation unit, an electromagnetically driven sealing valve, a load balancing unit, a miniature aerosol fire extinguishing unit, and a branch disconnect switch. The intelligent control module includes a built-in data preprocessing unit, an Attention-LSTM fusion prediction model unit, a coupling coefficient calculation unit, a dynamic threshold generation unit, and a control strategy generation unit. The communication module integrates a 5G module and an Ethernet interface, supports MQTT / HTTP communication protocols, and enables bidirectional data interaction with the cloud management platform and the upper-level power grid.

2. The adaptive intelligent protection distribution box according to claim 1, characterized in that: The enclosure is equipped with a filter ventilation port, and the interior is divided into a main chamber and independent branch chambers. Each independent branch chamber is equipped with a miniature aerosol fire extinguishing unit and an electromagnetically driven isolating switch to achieve precise fault isolation.

3. The adaptive intelligent protection distribution box according to claim 1, characterized in that: The adjustable-speed ventilation and filtration unit uses a brushless DC motor, the multi-mode heat dissipation unit includes air cooling, heat pipes and micro water cooling components, and the micro aerosol fire extinguishing unit uses non-toxic and environmentally friendly aerosol.

4. A control method for an adaptive intelligent protection distribution box, characterized in that, Includes the following steps: S1. The multi-dimensional sensing module collects environmental and operating parameters at a preset frequency and transmits them to the intelligent control module; S2. The data preprocessing unit performs noise reduction and normalization on the collected data and extracts feature parameters. S3. First, calculate the environment-load coupling coefficient K to quantify the risk level, and then use the Attention-LSTM fusion model to predict the future temperature, load rate and risk level. In case of anomalies, call cloud data for calibration. S4. Based on equipment runtime, failure frequency and coupling risk level, generate dynamic thresholds, and generate hierarchical control commands through a multi-objective weighted decision model; S5. The intelligent control module issues control commands, the adaptive protection execution module executes the control commands, and the execution status is fed back to the intelligent control module in real time. S6. The communication module uploads the operating status, prediction results, coupling coefficient, dynamic threshold and control instructions to the cloud management platform. At the same time, it receives the upper-level power grid dispatch instructions and the cloud-optimized weight coefficients, and dynamically updates the model parameters and threshold calculation rules.

5. The control method for an adaptive intelligent protection distribution box according to claim 4, characterized in that: In S2, the noise reduction process uses a wavelet thresholding algorithm. The specific parameters are: selecting a db4 wavelet basis, decomposing into 3 layers, and setting a threshold. ,in, The standard deviation of noise. Given the data length, the signal-to-noise ratio after noise reduction should be ≥30dB; The normalization process uses the Min-Max normalization algorithm, and the formula is: X norm =(X-X min ) / (X max -X min ); Map all parameters to the interval [0, 1], where X of each parameter... min / X max The preset sensor measurement range; Feature parameter extraction specifically involves: Temperature change rate ΔT / Δt: Calculate the ratio of the temperature difference to the time difference over 10 consecutive sampling periods, and take the average value; Load factor L: According to the formula L=(I A +I B +I C ) / (3×I rated ) Calculate, rounded to one decimal place, where I rated This is the rated current of the branch circuit; Dust accumulation D acc The formula for integrating the normalized dust concentration over 1 hour is: ; Arc signal strength A s Multiply the normalized peak value of the arc signal by 100 and round to the nearest integer. Vibration frequency f: The vibration acceleration time-domain signal is converted into a frequency-domain signal by fast Fourier transform, and the peak frequency is extracted.

6. The control method for an adaptive intelligent protection distribution box according to claim 5, characterized in that: In S3, the environment-load coupling coefficient K is first calculated to quantify the risk level, specifically as follows: Weight coefficient initialization: default α=0.35, β=0.4, γ=0.15, δ=0.1, and the total weight is 1; Coefficient calculation logic: First, normalize each feature parameter to the [0, 1] interval, then substitute them into the formula: K=α×(ΔT / Δt) norm +β×L norm +γ×(D acc / 100) norm +δ×A s norm ; The calculation result is rounded to 3 decimal places, K [0, 1]; Risk level determination: K < 0.4 is low risk, 0.4 ≤ K < 0.7 is medium risk, and K ≥ 0.7 is high risk.

7. The control method for an adaptive intelligent protection distribution box according to claim 6, characterized in that: In S3, the Attention-LSTM fusion model is used to predict the temperature, load rate, and risk level for the next 10-60 minutes as follows: Running the Attention-LSTM fusion model: Model Structure: LSTM Layer: 9-dimensional input, 64 hidden neurons, tanh activation function, dropout rate 0.2, 50 iterations; Attention Layer: Uses additive attention mechanism to calculate the similarity between the input vector and the query vector, considering factors such as load rate L, temperature change rate ΔT / Δt, and arc signal intensity A. s Assign dynamic weights; Model training mechanism: Local incremental training is performed every 24 hours using the latest 1000 sets of data to update model weights; global optimal parameters are received from the cloud every 7 days to overwrite local parameters. Prediction results output: Predicted duration selection: Supports five adjustable levels of 10 / 20 / 30 / 45 / 60 minutes, which can be configured via the cloud according to the needs of the scenario; Output: Predicted temperature value T pred Load rate prediction value L pred Coupling risk level K pred ; Accuracy Guarantee: Predicted temperature error ≤ ±2℃, load rate prediction error ≤ ±3%.

8. The control method for an adaptive intelligent protection distribution box according to claim 7, characterized in that: In S3, the specific steps for retrieving cloud data for calibration in case of an anomaly are as follows: Anomaly detection criteria: Compare the prediction results with the average value of historical data for the same period. If the deviation is ≥30%, it is considered an anomaly. Correction process: The communication module sends an anomaly flag and current scene parameters to the cloud. The cloud then retrieves historical prediction data and correction models for similar scenarios, and issues correction coefficients. The local system corrects the error according to the formula below: T pred corrected =T pred ×(1+Δk); L pred corrected =L pred ×(1+Δk); Among them, T pred corrected To correct the temperature prediction value, L pred corrected The corrected load rate prediction value is given by Δk, which is the correction coefficient issued by the cloud, ranging from -0.1 to 0.

1.

9. The control method for an adaptive intelligent protection distribution box according to claim 8, characterized in that: In S4, the dynamic threshold generated based on device runtime, failure frequency, and coupling risk level is as follows: Basic parameter acquisition: runtime t of the intelligent control module's local storage device, historical failure frequency N; Threshold calculation logic: T1 = 45 - 0.005 × t. If T1 < 40, then take 40℃. T1 is the lower limit of the dynamic temperature threshold. T2 = 60 - 0.008 × t. If T2 < 55, then take 55℃. T2 is the upper limit of the dynamic temperature threshold. L th =70-0.01×N, if L th If <60, then take 60%, L th This is the dynamic load rate threshold; D1 = 2 + 0.1 × K pred Keep one decimal place, and D1 is the lower limit of the dynamic dust threshold; D2 = 5 + 0.3 × K pred Keep one decimal place; D2 is the upper limit of the dynamic dust threshold; V th =0.5+0.1×K pred Round to two decimal places; V th The dynamic vibration safety threshold; Threshold storage update: The dynamic threshold is recalculated once per hour, overwriting historical data and synchronously uploaded to the cloud for backup.

10. The control method for an adaptive intelligent protection distribution box according to claim 9, characterized in that: In S4, the generation of hierarchical control instructions through a multi-objective weighted decision model is specifically as follows: Target weight configuration: Default protection effectiveness weight W1=0.4, energy consumption optimization weight W2=0.3, power supply continuity weight W3=0.3, supports scenario-based adjustment; Decision-making logic and instruction details: Low-coupling-risk scenario: Judgment condition: K pred <0.4 and T pred ≤T1 and L pred <L th Control commands: Air-cooled mode activated, vents opened, load balancing unit in standby mode, fire suppression unit ready; Energy optimization measures: If there are no parameter fluctuations for 30 consecutive minutes, the fan speed is reduced to 0.3 m / s; Intermediate coupling risk scenario: Judgment condition: 0.4≤K pred <0.7 or T1 < T pred ≤T2 or L th ≤L pred <85%; Control commands: Air cooling + heat pipe coordinated start-up; if the real-time dust concentration D < D2, the vents will open; the load balancing unit monitors non-critical loads in real time; Power supply continuity guarantee: the migration priority of non-critical loads is lower than that of critical loads. High coupling risk scenario: Judgment condition: K pred ≥0.7 or T pred >T2 or L pred ≥85%; Control commands: Three-stage cooling system (water cooling + air cooling + heat pipes) activated; if real-time dust concentration D ≥ D2, the sealing valve shuts off and internal circulation switches; the load balancing unit forcibly migrates non-critical loads; Enhanced protection measures: Vibration V ≥ V th At times, an additional cavity reinforcement electromagnetic lock is activated to prevent external impacts from causing structural loosening; Arc fault classification response: minor arc A s <0.3: Send a load reduction command, target load rate ≤50%, increase the sampling frequency of the arc detection sensor to 5Hz, monitor continuously for 10 minutes, and restore the original load if no abnormality is found; for moderate arcs 0.3≤A s <0.7: Immediately send a branch isolating switch disconnect command, the fire extinguishing unit enters standby mode, the aerosol spray nozzle is unlocked, and a fault alarm is uploaded simultaneously; Severe arcing A s ≥0.7: When the isolating switch is disconnected, a fire extinguishing unit activation command is sent, which only applies to the faulty branch chamber, while other branches are powered normally; External force impact response: Judgment condition: Vibration magnitude V ≥ V th And the vibration frequency f [10-100Hz]; Control commands: Close the electromagnetic drive sealing valve, reduce the load rate of non-critical branches to below 40%, continuously monitor the vibration sensor for 5 minutes, and if the vibration level drops to V... th If the following steps are taken, the sealing valve will be restored to the open state; otherwise, the protection state will be maintained and an alarm will be uploaded.

Images