Self-adaptive temperature regulation GIS control cabinet intelligent management system

Abstract

The invention discloses an intelligent management system for a self-adaptive temperature regulation GIS control cubicle, which relates to the technical field of GIS control cubicles, and comprises a temperature prediction control unit and a self-adaptive control platform, the temperature prediction control unit operates in the self-adaptive control platform, and the self-adaptive control platform is connected with the GIS control cubicle. The temperature prediction control unit comprises a data acquisition sub-module, a feature extraction sub-module and a prediction sub-module based on an LSTM-RL fusion model, the data acquisition sub-module periodically receives data of the phase change heat management unit, the wireless sensing node and the external environment interface, and the feature extraction sub-module carries out multi-scale time sequence feature analysis; and the prediction submodule outputs the temperature field distribution of each subarea in the next five minutes. By installing the temperature prediction control unit, accurate prediction of future temperature field changes is achieved, the problems that local hot spot response lags behind, the temperature fluctuation amplitude is large, and operation strategies of different cooling elements are difficult to coordinate under multi-source disturbance are solved, temperature fluctuation is reduced, energy consumption is reduced, and the service life of a GIS control cubicle assembly is prolonged.

Description

Technical Field

[0001] The present invention relates to the technical field of GIS control cabinets, and in particular to an adaptive temperature-adjusting intelligent management system for GIS control cabinets. Background Art

[0002] As power density within GIS control cabinets continues to increase, traditional passive cooling or simple fan and oil cooling methods are unable to promptly address local hotspots, resulting in large temperature fluctuations, high energy consumption, and shortened component lifespans. Existing systems generally lack dynamic management of phase change materials, self-learning predictions, and refined airflow control, making it difficult to achieve adaptive optimization for sudden changes in the environment and load.

[0003] Patent CN111177839B discloses an automated converged equipment room layout system and generation method. The above patent can generate cabinet layout and automatic layout of equipment in the cabinet in the computer room based on basic information of the equipment, thermal power monitoring and other factors, and supports human-computer interactive adjustment of cabinet layout and equipment layout.

[0004] The above patent solves the technical problem of the lack of professional, automated and intelligent planning capabilities in the existing computer room equipment layout, but lacks accurate prediction of future temperature field changes of computer room equipment.

[0005] To this end, this application proposes an adaptive temperature regulation GIS control cabinet intelligent management system that can accurately predict future temperature field changes. Summary of the Invention

[0006] The purpose of the present invention is to provide an adaptive temperature regulating GIS control cabinet intelligent management system to solve the technical problems of large temperature fluctuation, high energy consumption and shortened life of local components raised in the above background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solutions: an adaptive temperature-regulating GIS control cabinet intelligent management system, comprising a temperature prediction control unit and an adaptive control platform, wherein the temperature prediction control unit operates in the adaptive control platform and comprises a data acquisition submodule, a feature extraction submodule, and a prediction submodule based on an LSTM-RL fusion model, wherein the data acquisition submodule periodically receives data from a phase change thermal management unit, a wireless sensor node, and an external environment interface, the feature extraction submodule performs multi-scale time series feature analysis, and the prediction submodule outputs the temperature field distribution of each partition in the next 5 minutes;

[0008] The adaptive control platform includes a communication management module, a model management module, a control decision module and an execution and distribution module. The communication management module is responsible for protocol conversion and data synchronization. The model management module maintains the LSTM-RL model and digital twin model for online training and simulation optimization. The control decision module generates adjustment instructions based on model output and multi-objective optimization algorithm. The execution and distribution module distributes the adjustment instructions to the adaptive feedback interfaces of the phase change thermal management unit, the air duct adjustment unit and the wireless distributed sensor network unit respectively.

[0009] Preferably, the phase change thermal management unit includes several phase change material modules, each module encapsulates a polymer-based phase change material with an adjustable melting point, and has built-in temperature, solid phase and liquid phase sensors. The temperature, solid phase and liquid phase sensors collect local phase change state and temperature data in real time and upload them to the adaptive control platform via an internal RS-485 bus;

[0010] The phase change material module in the phase change thermal management unit further includes a microstructured pore structure. By injecting nano-dopants of different polarities, the specific heat capacity and latent heat of melting of the material are dynamically adjusted. The phase change material module feeds back the phase change progress data to the communication management module through built-in solid and liquid phase sensors for fine-tuning of the melting point strategy.

[0011] Preferably, the data acquisition submodule of the temperature prediction control unit adopts time division multiplexing to pull data in parallel from the phase change thermal management unit, the wireless distributed sensor network unit and the external environment interface, and the feature extraction submodule uses the sliding window method to perform a joint frequency domain and time domain analysis on the time series data to enhance the prediction ability of short-term mutations and long-term trends.

[0012] Preferably, the air duct adjustment unit includes a plurality of PWM speed-adjustable brushless DC micro fans, controllable angle electric butterfly valves, and wind speed and pressure sensors. Each fan and air valve are grouped and arranged in a divided air duct partition, and communicate with the adaptive control platform via CAN-bus.

[0013] The execution and dispatching module of the air duct adjustment unit dynamically allocates the working priority of the fan and the air valve according to the "hotspot-air duct mapping matrix" output by the temperature prediction control unit, and monitors the actual airflow data collected by the wind speed and wind pressure sensors in real time for closed-loop error correction with the predicted value.

[0014] Preferably, the wireless distributed sensor network unit includes temperature and humidity sensors, vibration sensors, and current sensors arranged inside and around the cabinet. The sensors send packets to the Mesh wireless network nodes through the self-powered energy collection module, and the Mesh gateway then aggregates the data to the data interface of the adaptive control platform.

[0015] Each node in the wireless distributed sensor network unit integrates a low-power microcontroller and an energy harvesting controller. The energy harvesting controller dynamically switches hotspots and vibration energy sources based on the node power consumption curve, and avoids network conflicts through the TDMA protocol. The node periodically reports its status and responds to the parameter reconfiguration instructions of the adaptive control platform.

[0016] Preferably, the multi-objective optimization algorithm of the adaptive control platform adopts a hybrid architecture of genetic algorithm and particle swarm algorithm to comprehensively evaluate cooling efficiency, energy consumption, phase change material fatigue and system response delay, generate the optimal control strategy, and maintain parameter synchronization between the control decision module and the model management module to ensure consistency between online training and simulation optimization results.

[0017] Preferably, the intelligent management system further includes a digital twin simulation optimization unit, which is deployed in the cloud and receives real-time operation data from the adaptive control platform. It performs multi-condition batch simulation based on cold flow simulation, electrothermal coupling simulation and phase change material performance model, outputs optimal material parameters, air duct layout and control strategy, and updates the results to the local model management module through a secure channel.

[0018] The digital twin simulation optimization unit uses finite element-based electrothermal coupling simulation and fluid-solid coupling cold flow simulation. The simulation results include temperature field distribution, heat transfer efficiency, phase change material phase change curve and air duct pressure distribution. They are regularly sent to the local model management module through a secure API to update the local prediction model and control decision parameters.

[0019] Preferably, a fault diagnosis submodule is also provided in the adaptive control platform. The fault diagnosis submodule uses a Bayesian network and anomaly detection algorithm to identify phase change material failure, fan failure and communication link anomaly based on the time series prediction residual and sensor health index, and pushes early warnings and maintenance suggestions to the operation and maintenance terminal.

[0020] Preferably, the phase change thermal management unit, the air duct adjustment unit and the wireless distributed sensor network unit realize physical layer redundancy and data backup through a unified central information bus. When the communication of any subsystem fails, the remaining subsystems can still maintain optimal control with the local prediction model and automatically transmit offline operation data after the communication is restored.

[0021] Preferably, the intelligent management system adopts a modular design as a whole, and each unit supports hot plugging and configuration expansion. Through self-test at system startup and online digital twin simulation calibration, the mapping and parameter matching between units are automatically completed, realizing plug-and-play and horizontal expansion.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] 1. This invention, equipped with a temperature prediction control unit, accurately predicts future temperature field changes, resolving issues such as delayed response to local hotspots, large temperature fluctuations, and difficulty coordinating the operating strategies of different cooling elements under multi-source disturbances. This reduces temperature fluctuations, lowers energy consumption, and extends the life of GIS control cabinet components.

[0024] 2. The present invention implements automatic routing communication between nodes by installing wireless distributed sensor network units, overcoming the problems of complex wiring, high energy consumption, multiple monitoring blind spots, and network conflicts in traditional sensors, thereby improving system operation and maintenance efficiency and deployment flexibility.

[0025] 3. This invention, through the installation of a digital twin simulation optimization unit, achieves continuous learning and adaptive updates throughout the entire life cycle, solving problems such as insufficient matching between static models and dynamic operating conditions, model aging, and strategy failure in extreme environments. It improves the accuracy of the system model, adapts to complex operating conditions such as extreme high temperatures and rapid load mutations, and reduces system maintenance costs.

[0026] 4. The present invention achieves rapid dynamic response to heat load by installing a phase change thermal management unit, solving the problems of traditional PCM materials with fixed melting point, limited thermal regulation efficiency, and poor adaptability, improving thermal capacity utilization and dynamically adapting to system operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Schematic diagram of the intelligent management process of the present invention. DETAILED DESCRIPTION

[0028] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0029] In the description of the present invention, it should be noted that the terms "upper," "lower," "inner," "outer," "front end," "rear end," "both ends," "one end," "the other end," and the like, indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limiting the present invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0030] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "installed," "provided with," "connected," etc., should be understood in a broad sense. For example, "connected" may refer to a fixed connection, a detachable connection, or an integral connection; it may refer to a mechanical connection or an electrical connection; it may refer to a direct connection or an indirect connection through an intermediate medium; it may refer to internal communication between two components. Those skilled in the art will be able to understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0031] See also Figure 1 The present invention provides an embodiment of an adaptive temperature-adjusting GIS control cabinet intelligent management system, comprising a temperature prediction control unit and an adaptive control platform. The temperature prediction control unit runs in the adaptive control platform. The temperature prediction control unit comprises a data acquisition submodule, a feature extraction submodule, and a prediction submodule based on an LSTM-RL fusion model. The data acquisition submodule periodically receives data from a phase change thermal management unit, a wireless sensor node, and an external environment interface. The feature extraction submodule performs multi-scale time series feature analysis. The prediction submodule outputs the temperature field distribution of each partition in the next 5 minutes.

[0032] The adaptive control platform includes a communication management module, a model management module, a control decision module and an execution and delivery module. The communication management module is responsible for protocol conversion and data synchronization. The model management module maintains the LSTM-RL model and digital twin model for online training and simulation optimization. The control decision module generates adjustment instructions based on model output and multi-objective optimization algorithm. The execution and delivery module sends the adjustment instructions to the adaptive feedback interfaces of the phase change thermal management unit, the air duct adjustment unit and the wireless distributed sensor network unit respectively.

[0033] Furthermore, the data acquisition submodule polls the temperature T of the PCM module from the RS-485 bus of the phase change thermal management unit every T1=5s. PCM and phase transition progress S PCM , synchronously pull the temperature and humidity T of 12 nodes from the wireless sensor network Mesh gateway i 、H i and vibration V i Data, obtain the ambient temperature T from the external environment interface env and humidity H env , encapsulate all data into a timestamp frame Frame k , and write to the local ring buffer Buffer, Buffer depth micro N = 12 frames;

[0034] The feature extraction submodule calculates the moving average, maximum value, minimum value, and mean square error of each temperature signal in the buffer, and the vibration signal V iCalculate the peak-to-peak value and effective value for the PCM module temperature T PCM Perform FFT on the sequence, extract the main frequency amplitude A1 and harmonic amplitude A2, and convert the eigenvector F k The dimension is expanded to D = (number of PCMs × 6) + (number of nodes × 6) + environmental features 4;

[0035] The prediction submodule will take the latest F k Input to LSTM layer, output hidden state h k , h k With the previous moment strategy action a {k-1} After splicing, input reinforcement learning strategy network, output action Q value evaluation and temperature prediction ΔT={ΔT1...ΔT m}, m is the number of air duct partitions, combined baseline temperature T base With ΔT, we can get the temperature prediction sequence T for each partition in the next 5 minutes. pred , T pred Encapsulated as "hotspot-air duct mapping matrix" M hot , identify the priority of high temperature areas in each partition;

[0036] The control decision module will hot and current energy consumption E cur , Remaining life of phase change material L PCM Together as the objective function input genetic + particle swarm hybrid algorithm, the optimization variables include:

[0037] Fan speed V fan 、Butterfly valve angle θ valve , PCM module melting point adjustment ΔP pcm ;

[0038] After 200 generations of iteration, the optimal solution {V fan ,θ valve , ΔP pcm};

[0039] The execution module sends V fan and θ valve The command sends the value to the corresponding node of the air duct adjustment unit, and transmits ΔP pcm The command is sent to the phase change thermal management unit, which automatically fine-tunes the dopant injection amount in the material doping controller and simultaneously sends the next sampling cycle and energy collection mode switching command to the sensor node;

[0040] After each execution unit is completed, the actual operation status is immediately sent back to the adaptive control platform. The platform compares the predicted and actual deviations and converts the residual r k The reward is fed back to the reinforcement learning module and stored in the experience replay pool for continuous online training and updating of model weights.

[0041] See also Figure 1 The present invention provides an embodiment of an adaptive temperature-adjustable GIS control cabinet intelligent management system, wherein the phase change thermal management unit includes a plurality of phase change material modules, each module encapsulating a polymer-based phase change material with an adjustable melting point and having built-in temperature, solid-phase, and liquid-phase sensors. The temperature, solid-phase, and liquid-phase sensors collect local phase change state and temperature data in real time and upload the data to the adaptive control platform via an internal RS-485 bus.

[0042] The phase change material module in the phase change thermal management unit further includes a microstructured pore structure. By injecting nano-dopants of different polarities, the specific heat capacity and latent heat of melting of the material are dynamically adjusted. The phase change material module uses built-in solid and liquid phase sensors to feed back phase change progress data to the communication management module for fine-tuning of the melting point strategy.

[0043] Furthermore, the shell material of the phase change material module is an aluminum alloy shell with dimensions of 80×80×20mm, and the inner cavity is filled with phase change material. The microstructured pore structure uses a micro-injection molding process to form a three-dimensional network channel with a pore size of 1-10μm and a porosity of 30% in the phase change material matrix to improve the heat exchange rate and the diffusion efficiency of the nano-dopant. The phase change material matrix is ​​a composite of polyvinyl alcohol and stearic acid with a basic melting point of 50°C. The nano-dopants are: positive polarity: graphene oxide, which lowers the melting point by 0-3°C; negative polarity: silicon fluoride nanoparticles, which raise the melting point by 0-5°C; the doping range is 0-5% by mass, and is injected via a micro-metering pump.

[0044] Temperature sensor: digital thermometer DS18B20, accuracy ±0.5°C, measurement range -10-80°C; phase change state sensor: solid phase detection: solid phase capacitance sensor based on capacitance change; liquid phase detection: photoelectric liquid phase sensor based on optical refractive index change; all three sensors are connected to the ADC / I / O of the STM32F103 microcontroller inside the module through a 4-wire interface 2 C bus;

[0045] The default sampling period is 2s, which can be adjusted to 1s-10s by issuing instructions from the adaptive control platform. The MCU reads the following data every 2s through the integrated driver: DS18B20 temperature, capacitance sensor C sol , photoelectric liquid phase sensor L liq ; Perform 3-point sliding average filtering on the temperature data to eliminate sudden noise; C sol With L liq Map to the phase transition progress S and generate the local state vector S local =[T mod ,S pcm ] T (T mod : Temperature after filtering; S pcm: phase change progress);

[0046] The adaptive control platform master station sends a register read request to each PCM module every 2 seconds by broadcasting or polling. Each module returns a data frame containing the local temperature and phase change progress. The platform communication management module parses the frame, checks the CRC, and caches it in the "Phase Change Module Status Table" after it is correct.

[0047] The adaptive control platform issues the "PCM melting point offset instruction" ΔP∈[-5°C, +5°C]; the module MCU calculates the required dopant type and mass fraction based on ΔP: if ΔP>0: select negative polarity dopant A, mass fraction = ΔP / 5×5%

[0048] else: Select positive polarity dopant B, mass fraction = |ΔP| / 3×5%

[0049] The nano-dopant solution is injected into the microporous structure through a micro peristaltic pump at an injection rate of 50uL / s and a total volume of ≤200uL. After the injection is completed, the MCU starts a 30s waiting period and re-reads S every 5s. pcm , if the phase transition progress S pcm If the change relative to before injection is less than 5%, the injection is considered to be failed and the error code 0×E1 is returned. Otherwise, a success response is returned and the updated phase change progress is carried in the next reported data frame.

[0050] See also Figure 1 The present invention provides an embodiment of an adaptive temperature-adjusting GIS control cabinet intelligent management system. The data acquisition submodule of the temperature prediction control unit uses time-division multiplexing to concurrently pull data from the phase-change thermal management unit, the wireless distributed sensor network unit, and the external environment interface. The feature extraction submodule uses a sliding window method to perform a joint frequency-domain and time-domain analysis of time series data to enhance the ability to predict short-term mutations and long-term trends.

[0051] The air duct adjustment unit includes multiple PWM speed-adjustable brushless DC micro fans, controllable angle electric butterfly valves, and wind speed and pressure sensors. Each fan and air valve are grouped together and arranged in a divided air duct zone, and communicate with the adaptive control platform via CAN-bus.

[0052] The execution and dispatching module of the air duct adjustment unit dynamically assigns the working priority of fans and air valves based on the "hotspot-air duct mapping matrix" output by the temperature prediction control unit. It also monitors the actual airflow data collected by the wind speed and pressure sensors in real time to perform closed-loop error correction against the predicted values.

[0053] Furthermore, the sampling period, basic frame period T0 = 1s, data collection time periods t1 (phase change unit), t2 (wireless sensor network), and t3 (environmental interface) are 200ms each, and the remaining 400ms is used for local processing and communication idleness; the sliding window length N = 10 frames, the step size ΔN = 2 frames; the frequency domain analysis FFT point number M = 32, the air duct partitioning: a total of m = 4 air duct partitions: upper, middle, lower, and relay; the number of fan-butterfly valve pairs in each zone = 1;

[0054] Duct adjustment unit and execution: Communication architecture: CAN-bus 1Mb / s, bus terminal resistance 120Ω; Node ID allocation: fan ID 0×11-0×14, butterfly valve ID 0×21-0×24, sensor ID 0×31-0×34; control decision module reads M hot , press P i The sequence {i1...i m}; Allocate resources: fan speed V for high priority partition i1 max , butterfly valve opening θ max ; The second highest i2 is set to 0.8V max , 0.8θ max , and so on.

[0055] See also Figure 1 The present invention provides an embodiment of an adaptive temperature-adjustable GIS control cabinet intelligent management system, wherein the wireless distributed sensor network unit includes temperature and humidity sensors, vibration sensors, and current sensors arranged inside and around the cabinet. The sensors send packets to Mesh wireless network nodes through a self-powered energy collection module, and the Mesh gateway then aggregates the data to the data interface of the adaptive control platform.

[0056] Each node in the wireless distributed sensor network integrates a low-power microcontroller and an energy harvesting controller. The energy harvesting controller dynamically switches between hotspots and vibration energy sources based on the node's power consumption curve and uses the TDMA protocol to avoid network conflicts. The node periodically reports its status and responds to the parameter reconfiguration instructions of the adaptive control platform.

[0057] The adaptive control platform's multi-objective optimization algorithm uses a hybrid architecture of genetic algorithm and particle swarm optimization to comprehensively evaluate cooling efficiency, energy consumption, phase change material fatigue, and system response delay to generate an optimal control strategy. It also synchronizes parameters between the control decision module and the model management module to ensure consistency between online training and simulation optimization results.

[0058] Furthermore, the objective function and normalization: Let the decision variable vector be x=[V1,...,V m

[0059] ,θ1,...,θ m,ΔP1,...,ΔP p ], corresponding to m fan speed ratios V i ∈[0,1], butterfly valve opening ratio θ i ∈[0, 1], and p PCM melting point offsets ΔP j ∈[-5, 5]℃. Define four performance indicators:

[0060] Cooling efficiency f1(x), Bigger is better;

[0061] System energy consumption f2(x), The smaller the better;

[0062] Phase change material fatigue f3(x), The smaller the better, α = 0.1 / ℃;

[0063] The system response delay f4(x) measures the time from the issuance of the control command to the stabilization of the wind speed and temperature, recorded as τ(x). The smaller the better.

[0064] Normalize the four targets to f ～ k ∈[0,1], and then use weight w k Weighted summation to obtain single-objective fitness: F x =w1f ～ 1+w2(1-f ～ 2)+w3(1-f ～ 3)+w4(1-f ～ 4), where the typical weights are w1=0.3, w2=0.25, w3=0.2, w4=0.25, ∑w k =1;

[0065] Hybrid Architecture Flow:

[0066] Initialization: Randomly generate GA population Randomly generate PSO particle swarm Initialization speed v j ~U(-1,1);

[0067] GA iteration (first G1 = 30 generations): calculate the fitness of each individual F(x i ); Selection: Roulette wheel to select the parent; Crossover (Probability P c =0.8), single point or uniform crossover; mutation (probability P m =0.05), make a small disturbance to the decision variable; update the population;

[0068] GA optimal top N p Individuals are transferred to PSO: Initialize the PSO particle position y j =x(j) , speed maintenance;

[0069] PSO iteration (I = 50 times): For each particle y j ←y j +v j , where inertia weight w = 0.7, learning factor c1 = c2 = 1.5, random numbers r1, r2 ~ U (0, 1), update individual optimal and the global optimal g best ;

[0070] Fusion and output: After PSO is completed, the optimal control strategy will be output; and the strategy will be written back to the GA population elite library as the seed for the next round of optimization.

[0071] See also Figure 1 The present invention provides an embodiment of an adaptive temperature-adjusting GIS control cabinet intelligent management system, which further includes a digital twin simulation optimization unit. The digital twin simulation optimization unit is deployed in the cloud, receives real-time operating data of the adaptive control platform, performs multi-condition batch simulation based on cold flow simulation, electrothermal coupling simulation, and phase change material performance model, outputs optimal material parameters, air duct layout, and control strategy, and updates the results to the local model management module through a secure channel.

[0072] The digital twin simulation optimization unit uses finite element-based electrothermal coupling simulation and fluid-solid coupling cold flow simulation. The simulation results include temperature field distribution, heat transfer efficiency, phase change material phase change curve, and air duct pressure distribution. These results are regularly sent to the local model management module via a secure API to update the local prediction model and control decision parameters.

[0073] The adaptive control platform also includes a fault diagnosis submodule, which uses a Bayesian network and anomaly detection algorithm to identify phase change material failures, fan failures, and communication link anomalies based on time series prediction residuals and sensor health indicators, and sends warnings and maintenance recommendations to the operation and maintenance terminal.

[0074] Furthermore, multiple working condition batch simulation: working condition parameter range: ambient temperature [-20, 50] ℃, step size 10 ℃; load: 10%-100% rated current, step size 10%; PCM melting point offset: [-5, +5] ℃, step size 2 ℃; fan speed ratio: [0.2, 1], step size 0.2; butterfly valve opening ratio: [0, 1], step size 0.25; simulation process: point thermal coupling simulation: establish the GIS control cabinet geometric model, define the copper busbar, insulation air gap, PCM module thermal properties; set the thermal load boundary conditions, solve the temperature field analysis T(x, t); Fluid-solid coupled cold flow simulation: Import temperature field distribution, couple the air flow field with the structural thermal expansion to simulate the velocity field v(x) and pressure field p(x) in the duct; Phase change material performance simulation: Based on the PCM thermophysical property model, simulate the phase change curve S(t) and latent heat release rate under different melting point offsets; Batch execution and parallelization: Use SLURM job scheduling to generate a total of 6×10×6×5×5=9000 simulation tasks according to the Cartesian product of the working conditions, and assign them to 100 cores for parallel computing; Each task outputs: the maximum temperature field T max , temperature field mean square error σ T Heat exchange efficiency PCM phase change completion time t pcm ;Average pressure drop in the air duct Δp;.

[0075] See also Figure 1 The present invention provides an embodiment of an adaptive temperature-adjustable GIS control cabinet intelligent management system. The phase-change thermal management unit, air duct adjustment unit, and wireless distributed sensor network unit achieve physical layer redundancy and data backup through a unified central information bus. When communication with any subsystem fails, the remaining subsystems can still maintain optimal control using local prediction models and automatically transmit offline operation data after communication is restored.

[0076] The intelligent management system adopts a modular design as a whole. Each unit supports hot swapping and configuration expansion. Through self-testing at system startup and online digital twin simulation calibration, it automatically completes mapping and parameter matching between units, achieving plug-and-play and horizontal expansion.

[0077] Furthermore, the online digital twin calibration process is as follows: calibration trigger conditions: system startup completes self-test; module mapping changes (addition / removal of units); periodically (default 24 hours) or when basic working conditions switch (ambient temperature difference > 10°C); calibration data collection: the platform collects the current position information of each unit, static wind pressure-flow curve, PCM phase change rate curve, and averages 10 samples each; calibration request is issued: the collected data and the latest environmental conditions are encapsulated as JSON and sent to the cloud digital twin interface via HTTPS POST; the cloud calculates the new mapping matrix M map*, the geometric parameters of the air duct partition δX, δY and the model weight update ΔW, the local model management module automatically replaces the mapping table and model weight after receiving them, and takes effect in the next control cycle.

[0078] Working Principle: The temperature and solid-liquid phase sensors built into the phase-change thermal management unit, as well as self-powered temperature, humidity, vibration, and current wireless sensor nodes located inside and outside the cabinet, provide real-time information on hotspot temperature, phase change progress, and environmental conditions.

[0079] In the adaptive control platform, data from each unit is aggregated using time-division multiplexing. A sliding window is used to extract time and frequency domain features. This data is then fed into an LSTM-RL fusion model to generate a five-minute temperature field forecast and a "hotspot-to-duct mapping matrix." Based on this matrix and the current system state, a genetic and particle swarm hybrid multi-objective optimization algorithm is employed to output the optimal fan speed, butterfly valve angle, and phase change material melting point offset strategy.

[0080] The execution and distribution module sends the adjustment instructions to each unit through CAN-bus, RS-485 and Mesh network respectively. Each unit feedbacks the actual wind speed, wind pressure, phase change progress and other operating data. The control platform sends the residual to the RL model for online training, and synchronizes the offline and online operating data to the cloud digital twin to continuously optimize the local model.

[0081] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. An adaptive temperature-adjusting GIS control cabinet intelligent management system, comprising a temperature prediction control unit and an adaptive control platform, characterized by: The temperature prediction control unit runs in the adaptive control platform and includes a data acquisition submodule, a feature extraction submodule, and a prediction submodule based on the LSTM-RL fusion model. The data acquisition submodule periodically receives data from the phase change thermal management unit, the wireless sensor node, and the external environment interface. The feature extraction submodule performs multi-scale time series feature analysis, and the prediction submodule outputs the temperature field distribution of each partition in the next 5 minutes. The adaptive control platform includes a communication management module, a model management module, a control decision module and an execution and distribution module. The communication management module is responsible for protocol conversion and data synchronization. The model management module maintains the LSTM-RL model and digital twin model for online training and simulation optimization. The control decision module generates adjustment instructions based on model output and multi-objective optimization algorithm. The execution and distribution module distributes the adjustment instructions to the adaptive feedback interfaces of the phase change thermal management unit, the air duct adjustment unit and the wireless distributed sensor network unit respectively.

2. The adaptive temperature adjustment GIS control cabinet intelligent management system according to claim 1 is characterized by: The phase change thermal management unit includes several phase change material modules, each of which is encapsulated with a polymer-based phase change material with an adjustable melting point and has built-in temperature, solid phase and liquid phase sensors. The temperature, solid phase and liquid phase sensors collect local phase change state and temperature data in real time and upload them to the adaptive control platform via an internal RS-485 bus. The phase change material module in the phase change thermal management unit further includes a microstructured pore structure. By injecting nano-dopants of different polarities, the specific heat capacity and latent heat of melting of the material are dynamically adjusted. The phase change material module feeds back the phase change progress data to the communication management module through built-in solid and liquid phase sensors for fine-tuning of the melting point strategy.

3. The adaptive temperature adjustment GIS control cabinet intelligent management system according to claim 1 is characterized by: The data acquisition submodule of the temperature prediction control unit adopts time-division multiplexing to pull data in parallel from the phase change thermal management unit, the wireless distributed sensor network unit and the external environment interface. The feature extraction submodule uses the sliding window method to perform a joint frequency domain and time domain analysis on the time series data to enhance the prediction ability of short-term mutations and long-term trends.

4. The adaptive temperature adjustment GIS control cabinet intelligent management system according to claim 1 is characterized by: The air duct adjustment unit includes multiple PWM speed-adjustable brushless DC micro fans, controllable angle electric butterfly valves, and wind speed and pressure sensors. Each fan and air valve are grouped together and arranged in a divided air duct zone, and communicate with the adaptive control platform via CAN-bus. The execution and dispatching module of the air duct control unit dynamically assigns the working priorities of fans and air valves based on the "hotspot-air duct mapping matrix" output by the temperature prediction control unit, and monitors the actual airflow data collected by wind speed and pressure sensors in real time for closed-loop error correction against the predicted values.

5. The adaptive temperature adjustment GIS control cabinet intelligent management system according to claim 1 is characterized by: The wireless distributed sensor network unit includes temperature and humidity sensors, vibration sensors, and current sensors arranged inside and around the cabinet. The sensors send data packets to the Mesh wireless network nodes through the self-powered energy collection module, and the Mesh gateway then aggregates the data to the data interface of the adaptive control platform. Each node in the wireless distributed sensor network unit integrates a low-power microcontroller and an energy harvesting controller. The energy harvesting controller dynamically switches hotspots and vibration energy sources based on the node power consumption curve, and avoids network conflicts through the TDMA protocol. The node periodically reports its status and responds to the parameter reconfiguration instructions of the adaptive control platform.

6. The adaptive temperature adjustment GIS control cabinet intelligent management system according to claim 1 is characterized by: The multi-objective optimization algorithm of the adaptive control platform adopts a hybrid architecture of genetic algorithm and particle swarm algorithm to comprehensively evaluate cooling efficiency, energy consumption, phase change material fatigue and system response delay, generate the optimal control strategy, and maintain parameter synchronization between the control decision module and the model management module to ensure consistency between online training and simulation optimization results.

7. The adaptive temperature adjustment GIS control cabinet intelligent management system according to claim 1 is characterized by: The intelligent management system also includes a digital twin simulation optimization unit, which is deployed in the cloud and receives real-time operating data from the adaptive control platform. It performs multi-condition batch simulation based on cold flow simulation, electrothermal coupling simulation, and phase change material performance model, outputs optimal material parameters, air duct layout, and control strategy, and updates the results to the local model management module through a secure channel. The digital twin simulation optimization unit uses finite element-based electrothermal coupling simulation and fluid-solid coupling cold flow simulation. The simulation results include temperature field distribution, heat transfer efficiency, phase change material phase change curve and air duct pressure distribution. They are regularly sent to the local model management module through a secure API to update the local prediction model and control decision parameters.

8. The adaptive temperature adjustment GIS control cabinet intelligent management system according to claim 1 is characterized by: The adaptive control platform also includes a fault diagnosis submodule, which uses a Bayesian network and anomaly detection algorithm to identify phase change material failures, fan failures, and communication link anomalies based on time series prediction residuals and sensor health indicators, and pushes warnings and maintenance recommendations to the operation and maintenance terminal.

9. The adaptive temperature adjustment GIS control cabinet intelligent management system according to claim 1 is characterized by: The phase change thermal management unit, air duct adjustment unit and wireless distributed sensor network unit achieve physical layer redundancy and data backup through a unified central information bus. When communication of any subsystem fails, the remaining subsystems can still maintain optimal control with local prediction models and automatically transmit offline operation data after communication is restored.

10. The adaptive temperature adjustment GIS control cabinet intelligent management system according to claim 1 is characterized by: The intelligent management system adopts a modular design as a whole. Each unit supports hot plugging and configuration expansion. Through self-testing at system startup and online digital twin simulation calibration, it automatically completes the mapping and parameter matching between units, realizing plug-and-play and horizontal expansion.

Images