An industrial compact isolating switch precision on-off control system

Abstract

The application discloses an industrial compact isolating switch precise on-off control system, and relates to the technical field of industrial isolating switch on-off control.The system comprises an integrated multi-parameter acquisition module, an edge computing control layer, a main control execution layer and a closed-loop verification module.The integrated multi-parameter acquisition module is adapted to the size of the reserved cavity of the isolating switch operating mechanism, is directly embedded after IP67 protection level packaging, and can collect operation torque, moving contact displacement, arc spectrum and environmental temperature and humidity in the on-off process in real time.The edge computing control layer is arranged on the ARM architecture embedded processing unit of the field industrial control cabinet, receives the collected data, completes preprocessing, and constructs an on-off state feature sequence.The application does not need to greatly transform the original isolating switch structure, adapts to the integrated needs of the stock transformation and new production equipment, the collection accuracy of the sensing unit is not affected by installation errors and running wear, the on-off reliability of the isolating switch can be effectively improved, and the operation and maintenance cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of industrial disconnector switch opening and closing control technology, and in particular to a precise opening and closing control system for a compact industrial disconnector switch. Background Technology

[0002] In 10kV to 35kV industrial high-voltage power distribution scenarios, compact disconnect switches have become core power distribution components for industrial power distribution circuit switching and equipment maintenance isolation due to their small size, high integration, and compatibility with the installation requirements of compact switchgear and prefabricated substations. With the development trend of intelligent and less-manned industrial power distribution, the industry has put forward higher requirements for the reliability and controllability of the opening and closing actions of compact disconnect switches. The core requirement is to achieve precise control of opening and closing actions, prediction of abnormal risks, and closed-loop verification of the entire process under extremely small installation space constraints, so as to avoid power distribution faults and safety accidents caused by incomplete opening and closing, arcing, and mechanical jamming.

[0003] Currently, the mainstream compact disconnector switch opening and closing control scheme is mainly based on mechanical travel trigger control. Its working principle involves installing mechanical travel switches at preset travel points on the operating mechanism. When the moving contact reaches the preset position, it triggers the travel switch's on / off signal, directly cutting off the driving power to the operating mechanism to complete the opening and closing action. This scheme is easy to implement and has low hardware costs, and is currently widely used in the control logic of existing industrial compact disconnectors. However, this scheme has significant drawbacks. On the one hand, the triggering accuracy of the travel switch is greatly affected by installation errors and long-term wear, easily leading to trigger offset and incomplete opening or closing. On the other hand, it can only identify whether the preset travel has been reached, and cannot detect potential risks such as abnormal torque, contact arcing, and mechanism jamming during the opening and closing process. Furthermore, compact disconnectors have very little reserved installation space, making it difficult to add additional expansion detection components, thus failing to meet the requirements of intelligent control.

[0004] Another existing solution is the external sensor-based electrical control scheme. This involves installing independent displacement and current sensors outside the disconnector to collect simple opening and closing status parameters. These parameters are then uploaded to the control cabinet to trigger the electrical control of the opening and closing actions. Compared to mechanical solutions, this offers improved control accuracy and can identify some obvious opening and closing anomalies. However, this solution requires installation space outside the disconnector, which does not meet the integration requirements of compact power distribution equipment. The external sensor wiring is susceptible to strong electromagnetic interference in industrial environments, leading to data distortion. Furthermore, it can only collect a few basic parameters, making it unable to predict risks during the opening and closing process. There is also no subsequent closed-loop verification mechanism, allowing only post-incident troubleshooting after an anomaly occurs. This makes it unsuitable for the operational needs of industrial power distribution scenarios with minimal human intervention.

[0005] The shortcomings of the existing solutions have become the core bottleneck restricting the intelligent upgrading of compact disconnect switches and improving the reliability of industrial power distribution operation. There is an urgent need to develop a switching control solution that is suitable for compact installation space and can achieve precise control throughout the entire process. Summary of the Invention

[0006] This invention proposes a precise opening and closing control system for industrial compact disconnect switches to solve the problems mentioned in the prior art, such as low opening and closing accuracy, poor adaptability, lack of risk prediction and closed-loop verification mechanism in existing industrial compact disconnect switch opening and closing control systems.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: an industrial compact disconnector precise opening and closing control system, applied to industrial high-voltage power distribution scenarios with rated voltages of 10kV to 35kV, including a data acquisition unit, a control unit, and an execution unit. The data acquisition unit is an integrated multi-parameter data acquisition module, adapted to the reserved cavity size of the industrial compact disconnector operating mechanism not exceeding 120mm×80mm×40mm, and encapsulated with IP67 protection level and embedded in the reserved cavity. It is used to acquire multiple parameters in real time during the opening and closing process of the disconnector, wherein the operating torque acquisition range is 0 to 200N·m with a resolution of not less than 0.1N·m, the moving contact displacement acquisition range is 0 to 150mm with a resolution of not less than 0.01mm, the inter-contact arc spectrum acquisition covers the visible to near-infrared band from 300nm to 800nm, and the temperature and humidity acquisition range is -40℃ to 85℃ and 0 to 100%RH. The control unit is an edge computing control layer deployed in the embedded processing unit of the industrial control cabinet on site. The embedded processing unit adopts ARM Cortex-M7 or above architecture and has a computing power of not less than 1000DMIPS. It is used to receive the raw acquisition data uploaded by the integrated multi-parameter acquisition module, and after completing the data preprocessing, it constructs a dimensionally unified separation and combination state feature sequence, and calls the pre-deployed lightweight separation and combination state prediction model to output separation and combination action control parameters. The execution unit is the main control execution layer, with a servo drive response delay of no more than 10ms. It is electrically connected to the servo operating mechanism of the industrial compact disconnect switch and is used to receive the opening and closing action control parameters output by the edge computing control layer and output the corresponding drive signal to control the servo operating mechanism to perform the opening and closing action. The control system also includes a closed-loop verification module with a data acquisition delay of no more than 50ms. It is connected to the integrated multi-parameter acquisition module and the main control execution layer signal respectively. It is used to perform a pass / fail verification of the status parameters of the disconnecting switch after the opening and closing action is completed. If the verification fails, it outputs a correction control command to the main control execution layer.

[0008] Preferably, the integrated multi-parameter acquisition module is an integrated packaged MEMS sensor array, including a micro-torque sensing unit, a laser micro-displacement sensing unit, a micro-spectral sensing unit, and a temperature and humidity sensing unit. The micro-torque sensing unit is attached to the output shaft surface of the operating mechanism to collect torque data. The laser micro-displacement sensing unit's emitting end is aligned with the transmission link end face of the moving contact to collect displacement data. The micro-spectral sensing unit's light-collecting window faces the gap between the moving and stationary contacts of the disconnector to collect arc spectrum data. The temperature and humidity sensing unit is arranged at the edge of the acquisition module to directly contact the ambient air to collect temperature and humidity data. All the wiring of the sensing units is built into the PCB board of the acquisition module, with no external exposed sensing wiring. The overall weight does not exceed 200g, and it will not increase the operating load of the operating mechanism.

[0009] Preferably, the integrated multi-parameter acquisition module adopts a trigger-based acquisition mode. Under normal conditions, it is in a low-power sleep state with a sleep power consumption of no more than 10μA. It automatically wakes up after receiving the disconnection / opening trigger signal, with a wake-up time of no more than 2ms. The sampling frequency is dynamically adjusted according to the progress of the disconnection / opening action. The specific adjustment rules are as follows: in the initial stage of the disconnection / opening action, when the moving contact moves less than 10% of the total stroke, the sampling frequency is 1kHz; in the middle stage of the disconnection / opening action, when the moving contact moves from 10% to 90% of the total stroke, the sampling frequency is increased to 10kHz; at the end of the disconnection / opening action, when the moving contact moves more than 90% of the total stroke, the sampling frequency drops back to 2kHz; and it automatically returns to the sleep state within 1s after the disconnection / opening action is completed.

[0010] Preferably, the data preprocessing logic of the edge computing control layer includes sliding window denoising, multi-parameter temporal alignment, and outlier removal. Sliding window denoising uses a Hamming window with a width of 5 to remove high-frequency electromagnetic interference noise from the collected data. Multi-parameter temporal alignment uses an alignment method based on the timestamp of the displacement parameter, matching the sampling points of torque, spectrum, temperature, and humidity to the temporal nodes of the displacement parameter one by one according to the timestamp. Outlier removal uses the 3σ criterion to remove outlier sampling points that deviate from the normal range by 3 times the standard deviation. The preprocessed multi-parameters are fused according to preset weight coefficients. Specifically, the weight coefficients are: 0.3 for operating torque, 0.4 for moving contact displacement, 0.2 for arc spectrum parameters, and 0.1 for temperature and humidity parameters. After weighted fusion, a separation and combination state feature sequence with a dimension of 1×128 is generated.

[0011] Preferably, the separation / combination state prediction model is a lightweight CNN and GRU fusion neural network model, wherein the lightweight CNN is responsible for extracting the spatial features of the separation / combination state feature sequence, and the GRU is responsible for extracting the temporal features of the separation / combination state feature sequence. The outputs of the two branches are concatenated and input into a fully connected layer to obtain a prediction result in three dimensions, namely the probability of separation / combination action in place, the level of jamming risk, and the expected arc extinguishing. The separation / combination state prediction model is pre-trained with more than 100,000 historical separation / combination samples of disconnecting switches covering different temperatures, humidity levels, aging levels, and load conditions. The prediction accuracy is not less than 99%. After compression using INT8 quantization, it is deployed on an embedded processing unit. The compressed model size does not exceed 2MB, and the single-frame inference latency does not exceed 1ms, which meets the requirements of real-time control on site.

[0012] Preferably, the drive signal adjustment logic of the main control execution layer is as follows: the jamming risk level is pre-divided into 1 to 5 levels, with the higher the level, the higher the jamming risk. The preset jamming risk level threshold is 3. When the jamming risk level is predicted to be higher than 3, the servo drive power is immediately cut off and a jamming alarm signal is output simultaneously through two methods: on-site indicator lights and host computer alarms. The preset arc extinguishing expectation threshold is 200ms. When the predicted arc extinguishing expectation exceeds 200ms, the moving contact stops for 150ms when it travels to 2mm away from the final position to wait for the arc to extinguish before completing the remaining stroke. The output torque adjustment range of the servo operating mechanism is 50% to 150% of the rated torque, and the travel speed adjustment range is 30% to 120% of the rated speed. The dynamic matching is based on the predicted probability of the separation and engagement actions being completed. The higher the probability of completion, the closer the output torque and travel speed are to the rated values.

[0013] Preferably, the pass / fail verification logic of the closed-loop verification module is as follows: After the opening and closing action is completed, three parameters of the mining isolation switch are measured: the final position of the moving contact, the contact resistance between the contacts, and the termination value of the operating torque. The pass / fail threshold for the final position of the moving contact is that the deviation from the standard position does not exceed ±0.2mm, the pass / fail threshold for the contact resistance between the contacts is not greater than 100μΩ, and the pass / fail threshold for the termination value of the operating torque is 80% to 120% of the rated termination torque. The three mining parameters are compared with the pre-stored pass / fail threshold range one by one. If any one of them does not meet the threshold requirement, a fine-tuning command in the corresponding correction direction is output to control the servo operating mechanism to perform the correction action. The minimum fine-tuning step size of the correction action is 0.05mm, and a maximum of 3 correction actions are allowed. If the threshold requirement is still not met after 3 corrections, the action is stopped immediately and a maintenance alarm signal is output. At the same time, the full data of this opening and closing is uploaded to the remote terminal.

[0014] Preferably, it also includes a remote operation and maintenance layer. The remote operation and maintenance layer communicates with the edge computing control layer and the main control execution layer through an industrial internet gateway that supports multiple industrial communication protocols such as Modbus and OPC UA. Data upload is encrypted using the national cryptographic SM4 algorithm and is used to upload full process data and alarm information of all separation and combination actions. It receives remotely issued manual control commands and model update packages. The issuance of manual control commands requires two levels of authorization verification. The model update package adopts an incremental update method. The update process will not affect the normal control logic on site. After the update is completed, it is automatically verified. If the verification fails, it will automatically roll back to the old version of the model, so there will be no control interruption. This completes the remote incremental update of the edge-side separation and combination state prediction model.

[0015] Preferably, the remote operation and maintenance layer has a built-in fault tracing module for storing all historical data on opening and closing operations for a period of not less than 3 years. When an opening or closing failure occurs, the cosine similarity matching algorithm is automatically used to match historical failure samples under the same temperature, humidity, load, and usage duration conditions. Samples with a similarity higher than 95% directly output the corresponding root cause of the failure, including insufficient lubrication of the operating mechanism, contact oxidation, and shaft wear. At the same time, corresponding operation and maintenance suggestions are output, including replenishing grease, polishing contacts, and replacing shafts, which greatly reduces the difficulty of fault diagnosis for operation and maintenance personnel.

[0016] Preferably, the system also includes an interlocking protection module, which is connected to the control system signals of other associated high-voltage electrical equipment on site. The associated high-voltage electrical equipment includes circuit breakers, grounding switches, and disconnect switches in adjacent bays for the corresponding circuits. The interlocking protection module outputs a hard contact signal with a response delay of no more than 5ms and a priority higher than all remote control commands. When the disconnecting switch's opening and closing action is not completed or the verification fails, the interlocking blocking signal is output to prohibit other associated high-voltage electrical equipment from performing start-stop operations. The interlocking blocking signal is only released after the disconnecting switch's opening and closing action is completed and the closed-loop verification is passed, effectively avoiding high-voltage safety accidents caused by misoperation.

[0017] Compared with existing technologies, the beneficial effects of this invention are: This invention integrates a sensing unit that combines micro-stroke acquisition and torque sensing into the original cavity of the disconnector switch operating mechanism, replacing the triggering logic of the traditional mechanical limit switch. It does not require additional installation space inside or outside the disconnector switch, nor does it require significant modification to the original disconnector switch structure. It effectively solves the problems of insufficient triggering accuracy of mechanical limit switches due to installation errors and long-term wear during operation, as well as the problems of external sensing solutions occupying installation space and having poor adaptability.

[0018] This invention pre-collects the operating reference parameters of the operating mechanism before the separation and opening actions are executed, dynamically matches the trigger threshold of the corresponding separation and opening actions, and compares the deviation between the collected parameters and the reference threshold in real time during the separation and opening process. It can identify potential hidden dangers such as mechanism jamming, torque exceeding limits, and abnormal contact discharge at the incipient stage of abnormal risks, and terminate abnormal actions in a timely manner and issue alarms. This effectively solves the problem that the existing technology cannot predict potential risks in the separation and opening process and can only troubleshoot faults after the fact.

[0019] This invention performs a closed-loop confirmation of the opening and closing effect by executing dual verification logic of infrared verification of contact position and power distribution circuit continuity verification after the opening and closing action is completed. The opening and closing action is only determined to be completed when both verifications pass. This effectively solves the problem of misjudgment of incomplete opening and closing that is prone to occur in the prior art which only relies on travel signal to determine the opening and closing status.

[0020] This invention is compatible with industrial compact disconnect switches of various voltage levels. It can be directly integrated into the control logic of newly manufactured compact disconnect switches, or used for the intelligent transformation of existing disconnect switches. Its resistance to strong electromagnetic interference and dust pollution is suitable for the operating environment of various industrial power distribution sites. It can meet the development needs of industrial power distribution with less manpower and intelligent operation and maintenance. It does not require additional special operation and maintenance equipment, has low deployment cost, and has high industry promotion value. Attached Figure Description

[0021] Figure 1 This is the overall flowchart of the precise opening and closing control of the disconnector switch proposed in this invention; Figure 2 This is a logic diagram of the multi-parameter fusion and preprocessing of the acquisition module proposed in this invention; Figure 3 This is a reasoning flowchart for the lightweight separation and combination state prediction model proposed in this invention; Figure 4 This is the closed-loop correction and interlocking protection linkage control diagram proposed in this invention; Figure 5 This is a diagram of the closed-loop system for remote operation and maintenance and fault tracing proposed in this invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Reference Figures 1 to 5This invention discloses a precise opening and closing control system for a compact industrial disconnector switch, applicable to industrial high-voltage power distribution scenarios with rated voltages from 10kV to 35kV. The system includes a data acquisition unit, a control unit, and an execution unit. The data acquisition unit is an integrated multi-parameter data acquisition module. Its housing is injection molded from modified PA66 with 30% glass fiber, with silicone rubber sealing rings embedded at the edges. Internally, it is encapsulated with thermally conductive epoxy resin to achieve IP67 protection. The overall dimensions are 116mm × 77mm × 36mm, perfectly fitting the pre-reserved cavity size of no more than 120mm × 80mm × 40mm for the operating mechanism of a compact industrial disconnector switch. It can be directly embedded in the pre-reserved cavity without additional drilling modifications. It is used to acquire multiple parameters during the opening and closing process of the disconnector switch in real time. The operating torque acquisition uses a foil strain gauge sensor chip with a data acquisition range of 0 to 200 N·m and a resolution of no less than 0.1 N·m. The displacement of the moving contact is acquired using a laser triangulation chip, with an acquisition range of 0 to 150 mm and a resolution of no less than 0.01 mm. The arc spectrum between the contacts is acquired using a miniature CMOS spectral sensor chip, covering the visible to near-infrared band from 300 nm to 800 nm. Temperature and humidity are acquired using an integrated MEMS temperature and humidity chip, with an acquisition range of -40℃ to 85℃ and 0 to 100%RH. The control unit is an edge computing control layer deployed in the embedded processing unit of the industrial control cabinet. The embedded processing unit uses an STM32H743 ARM Cortex-M7 architecture chip with a computing power of 1024 DMIPS, meeting the requirement of no less than 1000 DMIPS. It is used to receive the raw acquired data uploaded by the integrated multi-parameter acquisition module, complete the data preprocessing, construct a dimensionally unified separation and combination state feature sequence, and call the pre-deployed lightweight separation and combination state prediction model to output the separation and combination action control parameters.

[0024] The execution unit is the main control execution layer, which adopts the MINAS A6 series servo driver. The servo drive response delay is 8ms, which meets the requirement of not more than 10ms. It is electrically connected to the servo operating mechanism of the industrial compact disconnect switch through a shielded power line. It is used to receive the opening and closing action control parameters output by the edge computing control layer and output the corresponding PWM drive signal to control the servo operating mechanism to perform the opening and closing action.

[0025] The control system also includes a closed-loop verification module, which adopts hardware-triggered data acquisition logic with a data acquisition delay of 32ms, meeting the requirement of not exceeding 50ms. It is connected to the integrated multi-parameter acquisition module and the main control execution layer via SPI bus signals. It is used to acquire the status parameters of the isolating switch after the opening and closing action is completed to complete the qualification verification. If the verification fails, it outputs a correction control command to the main control execution layer.

[0026] In this invention, the integrated multi-parameter acquisition module is an integrated packaged MEMS sensing array, including a micro-torque sensing unit, a laser micro-displacement sensing unit, a micro-spectral sensing unit, and a temperature and humidity sensing unit. The micro-torque sensing unit uses a BF350-3AA type foil strain gauge, which is attached to the non-force-bearing side surface of the output shaft of the operating mechanism with high-temperature anaerobic adhesive to collect torque data. After the attachment is completed, zero-point calibration is performed by a torque calibrator. The laser micro-displacement sensing unit uses a VL53L5CX laser ranging chip. The transmitting end is aligned with the end face of the transmission rod of the moving contact. High reflective aluminum foil is pre-attached to the end face of the transmission rod to improve the ranging signal-to-noise ratio and collect displacement data. The micro-spectral sensing unit uses a C12880MA micro-spectral chip. A quartz glass lens is embedded in the light-collecting window, which is directly facing the gap between the moving and stationary contacts of the disconnecting switch to collect arc spectrum data. The temperature and humidity sensing unit uses an SHT30 integrated temperature and humidity chip, which is located on the edge of the PCB board of the acquisition module. The corresponding ventilation hole on the module housing allows direct contact with the ambient air to collect temperature and humidity data. All the wiring of the sensing unit is built into the internal wiring layer of the 4-layer PCB board, with no external exposed sensing wiring. The overall weight is 187g, which meets the requirement of not exceeding 200g and will not increase the operating load of the mechanism.

[0027] In this invention, the integrated multi-parameter acquisition module adopts a trigger-based acquisition mode. Under normal conditions, it is in a low-power sleep state with a sleep power consumption of 7.2μA, meeting the requirement of not exceeding 10μA. Upon receiving a trigger signal for the opening or closing of the isolating switch, it automatically wakes up with a wake-up time of 1.7ms, meeting the requirement of not exceeding 2ms. The sampling frequency is dynamically adjusted according to the progress of the opening or closing action. The specific adjustment rule is implemented using the following formula: ; In the above formula, For real-time sampling frequency, The base sampling frequency is set to 1kHz. This is the stroke coefficient, representing the distance the moving contact travels. When less than 10% of the total distance S Takes the value 1, when When the total distance S is 10% to 90% The value is 10, when When it is greater than 90% of the total distance S The value is 2. The system automatically returns to sleep mode within 1 second after the splitting and merging actions are completed. The code snippet for the core sampling control logic is as follows: void adjust_sample_freq(uint16_t current_s, uint16_t total_s) { float rate = (float)current_s / total_s; if(rate<0.1) set_sample_freq(1000); else if(rate>= 0.1&&rate<= 0.9) set_sample_freq(10000); else set_sample_freq(2000); } In this invention, the data preprocessing logic of the edge computing control layer includes sliding window denoising, multi-parameter temporal alignment, and outlier removal. The sliding window denoising uses a Hamming window with a width of 5 and a window coefficient of [0.08, 0.54, 1, 0.54, 0.08]. It performs convolution operations with 5 consecutive sampling points to remove high-frequency electromagnetic interference noise from the collected data. The multi-parameter temporal alignment uses an alignment method based on the timestamp of the displacement parameter. The sampling points of torque, spectrum, temperature and humidity are matched one by one to the temporal nodes of the displacement parameter according to the timestamp using a linear interpolation algorithm. Outlier removal employs the 3σ criterion, calculating the mean μ and standard deviation σ of 20 consecutive sampling points for each parameter, and removing outlier sampling points deviating from the interval [μ-3σ, μ+3σ]. The preprocessed multi-class parameters are then fused using a preset weighting coefficient, employing the following formula: ; In the above formula, To fuse feature values, This is the normalized operating torque value. This is the normalized displacement of the moving contact. These are the normalized arc spectrum parameters. These are the normalized temperature and humidity coupling parameters. After weighted fusion, a 1×128 sequence of split and merge state features is generated.

[0028] In this invention, the separation and merging state prediction model is a neural network model that integrates a lightweight CNN and a GRU. The lightweight CNN uses inverted residual blocks from MobileNetV2 with a kernel size of 3×3 and a stride of 1, and is responsible for extracting the spatial features of the separation and merging state feature sequence. The GRU uses a two-layer network structure with a hidden layer dimension of 64, and is responsible for extracting the temporal features of the separation and merging state feature sequence. The outputs of the two branches are concatenated and then input into a fully connected layer with a dimension of 3 to obtain prediction results in three dimensions: the probability of the separation and merging action being in place, the level of jamming risk, and the expected arc extinguishing.

[0029] The disconnection / opening state prediction model was pre-trained using 120,000 historical disconnection / opening samples covering different temperatures, humidity levels, aging degrees, and load conditions of disconnecting switches. The prediction accuracy reached 99.2%, meeting the requirement of at least 99%. After compression using INT8 quantization, the model was deployed on an embedded processing unit, resulting in a compressed model size of 1.7MB, meeting the requirement of not exceeding 2MB. The single-frame inference latency was 0.8ms, meeting the requirement of not exceeding 1ms, fully meeting the requirements of real-time on-site control. The code snippet for the core inference logic is as follows: TfLiteStatus predict(TfLiteTensor input, TfLiteTensor output) { TfLiteStatus invoke_status = interpreter->Invoke(); if(invoke_status == kTfLiteOk) { float output_data = output->data.f; float arrive_prob = output_data[0]; uint8_t jam_level = (uint8_t)output_data[1]; uint16_t arc_time = (uint16_t)output_data[2]; } return invoke_status; } In this invention, the drive signal adjustment logic of the main control execution layer is as follows: the jamming risk level is pre-divided into 1 to 5 levels, with the higher the level, the higher the jamming risk. The preset jamming risk level threshold is 3. When the jamming risk level is predicted to be higher than 3, the servo drive power is immediately cut off and the jamming alarm signal is output simultaneously through two methods: flashing red indicator lights on site and pop-up window on the host computer + SMS alarm. The preset arc extinguishing expectation threshold is 200ms. When the arc extinguishing expectation is predicted to exceed 200ms, the moving contact stops for 150ms when it travels to 2mm away from the final position to wait for the arc to extinguish before completing the remaining stroke. The output torque of the servo operating mechanism is adjustable from 50% to 150% of the rated torque, and the travel speed is adjustable from 30% to 120% of the rated speed. It is dynamically matched according to the predicted probability of the separation and engagement actions. The higher the probability of engagement, the closer the output torque and travel speed are to the rated values: when the probability of engagement is ≥99%, the output torque is set to 100% of the rated value and the travel speed is set to 100% of the rated value; when the probability of engagement is 95%~99%, the output torque is set to 120% of the rated value and the travel speed is set to 80% of the rated value; when the probability of engagement is <95%, the output torque is set to 150% of the rated value and the travel speed is set to 50% of the rated value.

[0030] In this invention, the qualification verification logic of the closed-loop verification module is as follows: After the opening and closing action is completed, three types of parameters are measured: the final position of the moving contact of the mining disconnector, the contact resistance between the contacts, and the termination value of the operating torque. The qualification threshold for the final position of the moving contact is that the deviation from the standard position does not exceed ±0.2mm. The contact resistance between the contacts is measured using the four-wire method, and the qualification threshold is not greater than 100μΩ. The qualification threshold for the termination value of the operating torque is 80% to 120% of the rated termination torque. The three mining parameters are compared with the pre-stored qualification threshold range one by one. If any one of them does not meet the threshold requirement, a fine-tuning command in the corresponding correction direction is output to control the servo operating mechanism to perform the correction action. The minimum fine-tuning step size of the correction action is 0.05mm, and a maximum of 3 correction actions are allowed. If the threshold requirement is still not met after 3 corrections, the action is stopped immediately and a maintenance alarm signal is output. At the same time, the full data of this opening and closing is uploaded to the remote terminal.

[0031] This invention also includes a remote operation and maintenance layer. The remote operation and maintenance layer communicates with the edge computing control layer and the main control execution layer through an industrial internet gateway that supports multiple industrial communication protocols such as Modbus RTU and OPC UA. Data upload is encrypted using the national cryptographic SM4 algorithm, and the encryption key is automatically rotated every 24 hours. It is used to upload full process data and alarm information of all separation and combination actions, and to receive remotely issued manual control instructions and model update packages. The issuance of manual control instructions requires two levels of authorization verification by the administrator and operation and maintenance personnel. Independent keys must be entered for the instructions to take effect. The model update package adopts an incremental update method, which only transmits the changed weight parameters. The update process runs in the background and will not affect the normal control logic on site. After the update is completed, it will automatically use 100 preset test cases to complete the verification. If the verification fails, it will automatically roll back to the old version of the model. There will be no control interruption. It can complete the remote incremental update of the edge-side split and merge state prediction model.

[0032] In this invention, the remote operation and maintenance layer has a built-in fault tracing module for storing all historical disconnection and connection actions' full data for 3.5 years, meeting the requirement of at least 3 years. When a disconnection or connection fault occurs, a cosine similarity matching algorithm is automatically used to match historical fault samples under the same temperature, humidity, load, and usage duration conditions. The cosine similarity calculation formula is as follows: ; In the above formula, The feature vector of the current fault. The feature vector of historical fault samples, The cosine similarity of two vectors is used to directly output the root cause of the failure when the similarity is higher than 95%, including insufficient lubrication of the operating mechanism, contact oxidation, shaft wear, etc. At the same time, the corresponding maintenance suggestions are output, including adding high-temperature grease, polishing the contacts, replacing the shaft, etc., which greatly reduces the difficulty of troubleshooting for maintenance personnel.

[0033] This invention also includes an interlocking protection module, which is connected to the control system signals of other associated high-voltage electrical equipment on site. The associated high-voltage electrical equipment includes vacuum circuit breakers, grounding switches, and disconnect switches in adjacent bays for the corresponding circuits. The interlocking protection module outputs a passive hard-contact signal with a response delay of 3ms, meeting the requirement of not exceeding 5ms. It has a higher priority than all remote control commands. When the disconnecting switch's opening and closing action is not completed or the verification fails, the interlocking blocking signal is output to prohibit other associated high-voltage electrical equipment from performing start-stop operations. The interlocking blocking signal is only released after the disconnecting switch's opening and closing action is completed and the closed-loop verification is passed, effectively avoiding high-voltage safety accidents caused by misoperation.

[0034] Scenario Example 1: Intelligent Upgrading of Existing 35kV Power Distribution Room in a Metallurgical Plant The 35kV power distribution room in a steel metallurgical plant has been in operation for 7 years. The compact disconnect switch inside uses a traditional electromagnetic operating mechanism and has no intelligent control capability. It frequently experiences problems such as incomplete opening and closing, and contact jamming and burning. The internal space of the power distribution room is narrow, and the reserved cavity for the disconnect switch operating mechanism is only 118mm×78mm×38mm, which makes it impossible to install external sensing equipment. The renovation requirements are that the original disconnect switch structure cannot be changed and the downtime should not exceed 2 hours.

[0035] The modification only requires removing the end cover of the isolating switch operating mechanism, directly embedding the integrated multi-parameter acquisition module into the reserved cavity, and connecting the wiring along the original reserved wiring hole to the embedded processing unit of the industrial control cabinet. The servo driver of the main control execution layer directly replaces the control unit of the original electromagnetic operating mechanism and directly interfaces with the original servo operating mechanism without modifying other structures. The entire modification process takes 1 hour and 40 minutes.

[0036] The acquisition module is fully compatible with the reserved cavity size, without occupying extra space, thus solving the problem of the inability to install external sensing solutions; the trigger-based acquisition and dynamic sampling frequency logic reduce power consumption while ensuring acquisition accuracy; the opening and closing state prediction model can identify jamming and arc timeout risks in advance, avoiding contact erosion; the closed-loop verification logic avoids misjudgment of incomplete opening and closing; the interlocking protection module avoids high-voltage accidents caused by misoperation, adapting to the strong electromagnetic interference and high reliability operation requirements of metallurgical plant areas.

[0037] The integrated protection system detected an overcurrent fault in the downstream circuit and issued a disconnect switch tripping command. After the acquisition module was awakened, parameters were collected at a dynamic frequency. After preprocessing by the edge computing layer, a feature sequence was generated. The model predicted a 98.7% probability of arrival, a level 2 risk of jamming, and an expected arc extinguishing time of 170ms. The main control execution layer output a drive signal with 120% rated torque and 80% rated speed. After the tripping was completed, the closed-loop verification returned a displacement deviation of +0.12mm, a contact resistance of 47μΩ, and a torque termination value of 107% of the rated value. All of these were qualified. The interlock was released, and feedback was sent to the integrated protection system that the tripping was completed. The entire process took 2.1 seconds and required no manual intervention.

[0038] Scenario Example 2: Factory Integration of a Newly Built 10kV Distributed Photovoltaic Substation A new 10MW distributed photovoltaic booster station project adopts an indoor compact high-voltage switchgear. The disconnecting switch is required to have remote operation and maintenance and fault tracing capabilities to meet the operation and maintenance needs of the photovoltaic station in a remote location with few personnel. The cavity reserved for the operating mechanism of the disconnecting switch inside the high-voltage switchgear is 120mm×80mm×40mm. All control components are required to be integrated inside the cabinet without the need for additional equipment.

[0039] During the production stage of the disconnect switch, the integrated multi-parameter acquisition module is directly embedded in the reserved cavity of the operating mechanism. The edge computing control layer is integrated into the comprehensive protection device of the high-voltage cabinet. The main control execution layer is integrated with the servo operating mechanism. The remote operation and maintenance layer is connected to the cloud operation and maintenance platform through the industrial Internet gateway of the photovoltaic power station. The interlocking protection module is directly hard-wired to the control circuit of the circuit breaker and grounding switch, without the need for additional wiring.

[0040] All components are fully integrated into the existing structure without occupying additional installation space; the lightweight INT8 quantized model is adapted to embedded low-computing-power environments, and its real-time performance meets requirements; the remote operation and maintenance layer enables remote control and incremental model updates without requiring on-site maintenance personnel; the fault tracing module can quickly locate the cause of the fault, reducing the operation and maintenance costs of remote sites; the interlocking protection module avoids photovoltaic grid-connection accidents caused by misoperation, and is adapted to the operation requirements of new energy sites.

[0041] The cloud-based operation and maintenance platform issues a command to open the disconnect switch that is shut down at night. After two levels of authorization verification, the command is sent to the edge computing layer. The data acquisition module is activated to collect parameters. The model predicts that the probability of the switch being in place is 99.3%, the risk of jamming is level 1, and the expected arc extinguishing time is 120ms. The main control execution layer outputs drive signals with 100% rated torque and 100% rated speed. After the switch is opened, all closed-loop verifications are passed, the interlocking is released, and all data is uploaded to the remote operation and maintenance layer for storage. The entire process does not require manual on-site operation.

[0042] refer to Figure 1 This diagram illustrates the entire lifecycle of the system, from perception to execution and closed-loop verification. The process begins with the integrated acquisition module sensing torque, displacement, spectrum, and temperature and humidity in real time. After the data is uploaded to the edge computing control layer, it undergoes preprocessing such as noise reduction and alignment, and the neural network model outputs control parameters. The execution layer drives the servo mechanism to complete the action based on the parameters, and then the verification module performs final status verification. If the verification is successful, the operation is completed and the data is uploaded; if it fails, fine-tuning correction or fault alarm is triggered. The entire process demonstrates the tight coupling between edge intelligence and hardware execution, solving the problems of low opening and closing accuracy and unmonitored status of traditional disconnect switches.

[0043] refer to Figure 2 This diagram illustrates the core data processing logic involved in this invention. The four types of parameters (torque, displacement, spectrum, and temperature / humidity) acquired by the acquisition module are first denoised using a Hamming window. Subsequently, the system uses the displacement parameter as the time reference axis to perform high-precision time-series alignment of the other three discrete parameters, and uses the 3σ criterion to eliminate abnormal jump values ​​caused by electromagnetic interference. Finally, feature fusion is performed according to preset weight coefficients to generate a feature sequence with a unified dimension. This logic ensures that the data input to the prediction model has extremely high spatiotemporal correlation and accuracy, providing the data foundation for achieving precise control.

[0044] refer to Figure 3 This figure details the architecture of a neural network model based on the fusion of CNN and GRU. The model employs a dual-branch structure: the convolutional network branch is responsible for capturing spatial topological features in the feature sequence, while the recurrent network branch is responsible for extracting dynamic temporal evolution features during the action process. The two feature paths are concatenated and fused before the fully connected layer, ultimately outputting prediction results in three dimensions: the probability of separation and reunification, the level of jamming risk, and the expected arc extinguishing. This model undergoes INT8 quantization compression, which significantly reduces computational resource consumption while maintaining high accuracy, making it possible to achieve millisecond-level real-time inference on power-constrained embedded processing units.

[0045] refer to Figure 4This figure illustrates the safety assurance mechanism of the present invention. After the action is executed, the verification module compares whether the position of the moving contact, the contact resistance, and the termination torque are within the qualified threshold. If a deviation occurs, the system will calculate the fine-tuning step size for step correction. At the same time, the interlocking protection module monitors the opening and closing status throughout the process. During the period when the action is not completed or the verification is not qualified, the associated circuit breaker and grounding switch are forcibly locked through hard contact signals to prevent illegal operations such as opening and closing under load. This dual redundancy design of physical interlocking and software verification greatly improves the safety and reliability of switch operation in industrial high-voltage scenarios.

[0046] refer to Figure 5 This diagram illustrates the system's logical architecture at the cloud management and intelligent operation and maintenance (O&M) levels. The remote O&M layer collects full-scale process data from the field via an industrial gateway and uses a cosine similarity algorithm to compare real-time fault characteristics with a historical sample database. When the matching degree reaches a high threshold, the system can automatically locate the root cause of the fault and generate O&M recommendations. Simultaneously, the platform supports remote incremental updates of the predictive model, ensuring that the model optimization process does not interfere with real-time field control through a dual-version rollback mechanism. This system achieves a closed-loop information flow from field control to remote intelligent diagnostics, reducing the maintenance costs of the power distribution system.

[0047] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A precise opening and closing control system for an industrial compact disconnecting switch, characterized in that, include: The acquisition unit is an integrated multi-parameter acquisition module, which is embedded in the reserved cavity of the industrial compact disconnect switch operating mechanism. It is used to acquire in real time the operating torque, moving contact displacement, arc spectrum parameters between contacts, and temperature and humidity data of the environment where the disconnect switch is located during the opening and closing process of the disconnect switch. The control unit is an edge computing control layer deployed in the embedded processing unit of the industrial control cabinet on site. It is used to receive the raw acquisition data uploaded by the integrated multi-parameter acquisition module, construct the separation and combination state feature sequence after data preprocessing, and call the pre-deployed separation and combination state prediction model to output the separation and combination action control parameters. The execution unit is the main control execution layer, which is electrically connected to the servo operating mechanism of the industrial compact disconnect switch. It is used to receive the opening and closing action control parameters output by the edge computing control layer and output the corresponding drive signal to control the servo operating mechanism to perform the opening and closing action. The control system also includes a closed-loop verification module, which is connected to the integrated multi-parameter acquisition module and the main control execution layer signal respectively. It is used to collect the status parameters of the disconnecting switch after the opening and closing action is completed to complete the qualification verification. If the verification fails, it outputs a correction control command to the main control execution layer.

2. The industrial compact disconnector precise opening and closing control system according to claim 1, characterized in that, The integrated multi-parameter acquisition module is an integrated packaged MEMS sensor array, including a micro-torque sensing unit, a laser micro-displacement sensing unit, a micro-spectral sensing unit, and a temperature and humidity sensing unit. Its overall size is adapted to the reserved cavity of the industrial compact disconnect switch operating mechanism, and there are no external sensing wires.

3. The industrial compact disconnector precise opening and closing control system according to claim 2, characterized in that, The integrated multi-parameter acquisition module adopts a trigger-based acquisition mode. Under normal conditions, it is in a low-power sleep state. It automatically wakes up after receiving the disconnection and opening trigger signal of the isolating switch. The sampling frequency is dynamically adjusted according to the progress of the disconnection and opening action. After the disconnection and opening action is completed, it automatically returns to the sleep state.

4. The industrial compact disconnector precise opening and closing control system according to claim 1, characterized in that, The data preprocessing logic of the edge computing control layer includes sliding window denoising, multi-parameter temporal alignment, and outlier removal. The preprocessed multi-parameters are fused according to preset weight coefficients to generate a dimensionally unified separation and combination state feature sequence.

5. The industrial compact disconnector precise opening and closing control system according to claim 4, characterized in that, The separation and merging state prediction model is a lightweight neural network model that integrates CNN and GRU. The input is the separation and merging state feature sequence, and the output is the prediction results in three dimensions: the probability of the separation and merging action being in place, the level of jamming risk, and the expected arc extinguishing. After quantization and compression, the separation and merging state prediction model is deployed in an embedded processing unit, and the single-frame inference latency meets the requirements of real-time control on site.

6. The industrial compact disconnector precise opening and closing control system according to claim 5, characterized in that, The drive signal adjustment logic of the main control execution layer is as follows: when the predicted jamming risk level is higher than the preset threshold, the splitting and merging action is paused and a jamming alarm signal is output; when the predicted arc extinguishing is expected to be longer than the preset threshold, the arc extinguishing waiting time of the splitting and merging action is extended; and the output torque and travel speed of the servo operating mechanism are dynamically adjusted according to the predicted probability of the splitting and merging action being completed.

7. The industrial compact disconnector precise opening and closing control system according to claim 1, characterized in that, The specific logic of the closed-loop verification module is as follows: After the opening and closing action is completed, the final position of the moving contact of the isolation switch, the contact resistance between the contacts, and the termination value of the operating torque are compared with the pre-stored qualified threshold range one by one. If any item does not meet the threshold requirement, a fine-tuning command in the corresponding correction direction is output to control the servo operating mechanism to perform the correction action. If the error still occurs after performing the correction action a maximum of a preset number of times, a maintenance alarm signal is output.

8. The industrial compact disconnector precise opening and closing control system according to claim 1, characterized in that, It also includes a remote operation and maintenance layer, which communicates with the edge computing control layer and the main control execution layer through an industrial internet gateway. It is used to upload full process data and alarm information of all separation and combination actions, receive remote manual control instructions and model update packages, and complete the remote incremental update of the edge-side separation and combination state prediction model.

9. The industrial compact disconnector precision opening and closing control system according to claim 8, characterized in that, The remote operation and maintenance layer has a built-in fault tracing module, which stores all historical data on the separation and connection actions. When a separation and connection failure occurs, it automatically matches historical fault samples under the same operating conditions, locates the root cause of the failure, and outputs corresponding operation and maintenance suggestions.

10. The industrial compact disconnector precise opening and closing control system according to claim 1, characterized in that, It also includes an interlocking protection module, which is connected to the control system signals of other related high-voltage electrical equipment on site. When the opening and closing action of the disconnecting switch is not completed or the verification is not qualified, the interlocking blocking signal is output to prohibit other related high-voltage electrical equipment from performing start and stop operations.

Images