Cooperative control system of primary and secondary deep fusion pole-mounted circuit breaker

Abstract

The invention relates to the technical field of circuit breaker control, in particular to a cooperative control system of a primary and secondary deep fusion pole-mounted circuit breaker, which comprises a perceptual coding module, an intelligent optimization module, a cooperative decision module, a digital twinning module, a causal reasoning module, a learning evolution module, a semantic communication module and an execution feedback module, multi-modal data acquisition is realized through a neuromorphic sensor, a protection strategy is generated in combination with quantum optimization and chaos detection, collaborative decision is realized by using multi-agent reinforcement learning, strategy verification is performed by means of digital twinning and meta-learning, a decision process is optimized based on causal reasoning, and continuous evolution of a model is realized through neural architecture search. And finally, a standardized control instruction is generated through semantic communication, and closed-loop control is formed, so that the fault early warning accuracy, the control response speed and the equipment service life are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of circuit breaker control technology, and in particular to a collaborative control system for a pole-mounted circuit breaker with deep integration of primary and secondary circuits. Background Technology

[0002] In the current field of circuit breaker protection and control, traditional systems mainly rely on single sensor data and preset thresholds for protection decisions, resulting in significant technical bottlenecks. Existing systems lack multimodal data fusion capabilities, making it difficult to comprehensively perceive the operating status of equipment; control strategies are mostly statically set, unable to adapt to dynamic changes in the power grid; fault early warning relies on simple threshold judgments, lacking the ability to identify early latent defects; the decision-making process lacks causal verification mechanisms, making it difficult to guarantee the scientific validity of control strategies; and the systems generally lack self-learning capabilities, hindering continuous optimization and upgrades. Furthermore, poor coordination between functional modules creates information silos, leading to slow system response and low control accuracy. With the advancement of new power system construction, power grid operating conditions are becoming increasingly complex. Traditional protection and control systems can no longer meet actual needs in terms of fault early warning timeliness, control accuracy, and system adaptability. There is an urgent need to construct a new generation of protection and control systems that integrate intelligent perception, collaborative decision-making, causal verification, and continuous evolution.

[0003] Chinese Patent Publication No. CN119864759A discloses an automated protection and control system, method, and device for pole-mounted circuit breakers. The system includes a grid connection request receiving module, a circuit breaker protection parameter setting module, a parameter correction module, and a protection control module. By determining whether the distributed power source is connecting to the grid for the first time, it dynamically sets and corrects protection parameters based on its scale information and environmental data, and optimizes action delay based on communication distance and status data. This invention covers steps such as grid connection request receiving, parameter initialization and correction, and anomaly detection. It uses IoT technology to collect multi-dimensional data in real time, quantitatively assesses the impact of inrush current and environmental fluctuations, and dynamically matches protection parameter thresholds and delays, effectively solving the problem of false alarms caused by inrush current when distributed power sources are connected, thus improving the stability of power grid operation and the accuracy of protection actions. However, this solution still suffers from problems such as delayed fault warnings, insufficient control accuracy, and poor system adaptability due to a lack of multimodal perception capabilities, isolated decision-making processes, and a lack of causal verification and adaptive evolution mechanisms. Summary of the Invention

[0004] To address this, the present invention provides a collaborative control system for pole-mounted circuit breakers with deep fusion of primary and secondary circuits, which overcomes the problems of delayed fault warning, insufficient control accuracy, and poor system adaptability caused by the lack of multimodal perception capabilities, isolated decision-making processes, lack of causal verification and adaptive evolution mechanisms in the prior art.

[0005] To achieve the above objectives, the present invention provides a collaborative control system for a primary and secondary deeply integrated pole-mounted circuit breaker, comprising: The perception coding module is used to collect multimodal operation data through neuromorphic sensors and perform pulse coding processing to obtain a pulse coding dataset. The intelligent optimization module is used to generate optimized protection strategies and early fault warnings based on the pulse coding dataset using quantum optimization algorithms and chaotic detection methods. The collaborative decision-making module is used to construct a multi-agent reinforcement learning framework, and generate a collaborative operation strategy through game theory collaboration based on the optimized protection strategy and early fault warning. A digital twin module is used to construct a digital twin of the circuit breaker and to verify and adapt the cooperative operation strategy through a meta-learning algorithm to obtain a verified control strategy. The causal reasoning module is used to evaluate the effectiveness of the verified control strategy based on the causal model, and to optimize the decision-making process through counterfactual reasoning to obtain the final control strategy. The learning evolution module is used to control the continuous optimization and evolution of the model through neural architecture search and weight consolidation algorithms; The semantic communication module is used to convert the final control strategy into semantic instructions, perform verification and reasoning through a knowledge graph, and generate standardized control instructions. The execution feedback module is used to execute the standardized control commands and collect status data to feed back to the sensing and encoding module, forming a closed-loop control system.

[0006] Furthermore, the perception coding module acquires multimodal operational data through a neuromorphic sensor and performs pulse coding processing to obtain a pulse coding dataset including: The neuromorphic sensor array is activated to collect multimodal operation data and obtain raw pulse data; the raw pulse data is processed by pulse coding to obtain encoded pulse sequences; the encoded pulse sequences are encapsulated to obtain pulse-coded datasets.

[0007] Furthermore, based on the pulse coding dataset, the intelligent optimization module generates optimized protection strategies and early fault warnings using quantum optimization algorithms and chaotic detection methods, including: The pulse code dataset is decoded and reconstructed to obtain multi-dimensional operating feature parameters. Quantum optimization calculations are performed based on the feature parameter set to obtain optimized protection settings. Chaotic fault detection is performed based on the feature parameter set to obtain early fault warning information. The optimized protection settings and early fault warning information are integrated to generate a protection strategy package.

[0008] Furthermore, the collaborative decision-making module is used to construct a multi-agent reinforcement learning framework, and based on the optimized protection strategy and early fault warning, generates a collaborative operation strategy through game theory cooperation, including: The multi-agent system is initialized to obtain a reinforcement learning environment framework. The optimized protection strategy and early fault warning are integrated to obtain the agent's prior knowledge. The multi-agent system is trained in a game-theoretic cooperative manner to obtain a preliminary cooperative strategy. The stability of the preliminary cooperative strategy is optimized to obtain the final cooperative operation strategy.

[0009] Furthermore, the digital twin module is used to construct a digital twin of the circuit breaker and to verify and adapt the cooperative operation strategy through a meta-learning algorithm, resulting in a verified control strategy including: Multi-dimensional modeling of the circuit breaker physical entity is performed to obtain an initial digital twin. The initial digital twin is then integrated with a meta-learning framework to obtain an adaptive digital twin. The collaborative operation strategy is verified using a digital twin to obtain a strategy performance evaluation. Based on the strategy performance evaluation, meta-learning adaptation is performed to obtain a verified control strategy.

[0010] Furthermore, the causal reasoning module evaluates the effectiveness of the verified control strategy based on a causal model, and optimizes the decision-making process through counterfactual reasoning to obtain the final control strategy, which includes: A causal graph model is constructed based on the verified control strategy and related operational data to obtain a structured causal network. The effect of the strategy is evaluated based on the structured causal network to obtain the causal effect quantification result. Counterfactual scenarios are generated and reasoned for the verified control strategy to obtain optimization decision suggestions. The strategy is iteratively optimized based on the counterfactual reasoning result to obtain the final control strategy.

[0011] Furthermore, the learning evolution module achieves continuous optimization and evolution of the control model through neural architecture search and weight consolidation algorithms, including: The current control model is evaluated for performance and analyzed for architecture to obtain model optimization requirements. Based on the model optimization requirements, a neural architecture search is performed to obtain an optimized model structure. The optimized model structure is then subjected to weight consolidation training to obtain a stable evolutionary model. The stable evolutionary model is then deployed and verified online to obtain the final optimized model.

[0012] Furthermore, the semantic communication module converts the final control strategy into semantic instructions, verifies and reasons using a knowledge graph, and generates standardized control instructions, including: The final control strategy is semantically parsed and transformed to obtain initial semantic instructions. The initial semantic instructions are then subjected to knowledge graph consistency verification to obtain semantic verification results. Based on the semantic verification results, knowledge graph reasoning optimization is performed to obtain optimized semantic instructions. The optimized semantic instructions are then standardized, encoded, and encapsulated to obtain standardized control instructions.

[0013] Furthermore, the execution feedback module executes the standardized control commands and collects status data to feed back to the sensing encoding module, forming a closed-loop control system including: The standardized control commands are parsed and executed to obtain control execution results. After execution, the device status is collected by multiple sensors to obtain raw status data. The raw status data is preprocessed and standardized to obtain formatted feedback data. The formatted feedback data is fed back to the sensing and encoding module in real time to form a closed-loop control cycle.

[0014] Furthermore, the standardized control instructions are parsed and executed to obtain control execution results. Specifically, the operation commands and parameters in the standardized control instructions are parsed, the circuit breaker operating mechanism is controlled to complete the specified action through the actuator drive circuit, and the execution status is monitored in real time, and the execution success or failure result is recorded. After execution, the device status is collected by multiple sensors to obtain raw status data. Specifically, the current, voltage, temperature rise, and mechanical displacement parameters of the circuit breaker during operation are collected by current sensors, voltage sensors, temperature sensors, and position sensors. The raw status data is obtained with a high-frequency sampling rate to ensure the comprehensiveness and real-time nature of the data. The original state data is preprocessed and standardized to obtain formatted feedback data. Specifically, the Kalman filter algorithm is used to eliminate sensor noise, the data is time-stamp aligned and dimension-normalized, and converted into a standardized format compatible with the input interface of the sensing coding module to form formatted feedback data. The formatted feedback data is fed back to the sensing and encoding module in real time to form a closed-loop control cycle. Specifically, the formatted feedback data is sent to the input buffer of the sensing and encoding module through a low-latency communication protocol to trigger a new round of pulse coding processing, thereby realizing iterative optimization of the closed-loop control system of "execution-feedback-sensing-decision". Compared with existing technologies, the beneficial effects of this invention are as follows: The system establishes a multimodal data acquisition system based on neuromorphic sensors through a perception coding module, providing high-quality pulse-coded data for intelligent decision-making; the system also achieves accurate early warning of faults and generation of optimized protection strategies through an intelligent optimization module combining quantum optimization and chaotic detection technologies; the system also constructs a multi-agent reinforcement learning framework through a collaborative decision-making module, forming optimal collaborative operation strategies through game theory cooperation; the system also establishes a high-fidelity circuit breaker model through a digital twin module, and uses meta-learning algorithms to achieve rapid verification and adaptation of strategies; the system also ensures the scientificity and reliability of control strategies through a causal reasoning module based on causal models and counterfactual reasoning; the system also achieves continuous optimization and self-evolution of the control model through a learning evolution module using neural architecture search and weight consolidation; the system also transforms control strategies into standardized instructions through a semantic communication module, and ensures the accuracy and consistency of instructions through a knowledge graph; the system also achieves precise execution of control instructions through an execution feedback module, and feeds back state data to the sensing end in real time, forming a complete closed-loop control system. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the collaborative control system for the secondary deep fusion pole-mounted circuit breaker in this embodiment. Detailed Implementation

[0016] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0017] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0018] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0019] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0020] Please see Figure 1 The diagram shown is a structural schematic of the collaborative control system for the secondary deep fusion pole-mounted circuit breaker in this embodiment. The system includes: The perception coding module is used to collect multimodal operation data through neuromorphic sensors and perform pulse coding processing to obtain a pulse coding dataset. The intelligent optimization module is used to generate optimized protection strategies and early fault warnings based on the pulse coding dataset using quantum optimization algorithms and chaotic detection methods. The intelligent optimization module is connected to the sensing coding module. The collaborative decision-making module is used to construct a multi-agent reinforcement learning framework. Based on the optimized protection strategy and early fault warning, it generates a collaborative operation strategy through game-theoretic cooperation. The collaborative decision-making module is connected to the intelligent optimization module. A digital twin module is used to construct a digital twin of the circuit breaker and to verify and adapt the cooperative operation strategy through a meta-learning algorithm to obtain a verified control strategy. The digital twin module is connected to the cooperative decision-making module. The causal reasoning module is used to evaluate the effectiveness of the verified control strategy based on the causal model, and to optimize the decision-making process through counterfactual reasoning to obtain the final control strategy. The causal reasoning module is connected to the digital twin module. The learning evolution module is used to achieve continuous optimization and evolution of the control model through neural architecture search and weight consolidation algorithms. The learning evolution module is connected to the causal inference module. A semantic communication module is used to convert the final control strategy into semantic instructions, perform verification and reasoning through a knowledge graph, and generate standardized control instructions. The semantic communication module is connected to the learning and evolution module. The execution feedback module is used to execute the standardized control commands and collect status data to feed back to the perception encoding module to form a closed-loop control system. The execution feedback module is connected to the semantic communication module.

[0021] Specifically, the system is applied to the collaborative control terminal of a pole-mounted circuit breaker with deep fusion of primary and secondary components. The system establishes a multimodal data acquisition system based on neuromorphic sensors through a perception coding module, providing high-quality pulse-coded data for intelligent decision-making. The system also achieves accurate early warning of faults and generates optimized protection strategies through an intelligent optimization module combining quantum optimization and chaotic detection technologies. Furthermore, the system constructs a multi-agent reinforcement learning framework through a collaborative decision-making module, forming optimal collaborative operation strategies through game theory cooperation. The system also establishes a high-fidelity circuit breaker model through a digital twin module and utilizes meta-learning algorithms to achieve rapid strategy verification and adaptation. The system ensures the scientific validity and reliability of the control strategy through a causal reasoning module based on causal models and counterfactual reasoning. The system achieves continuous optimization and self-evolution of the control model through a learning evolution module using neural architecture search and weight consolidation. The system transforms the control strategy into standardized instructions through a semantic communication module and ensures the accuracy and consistency of the instructions through a knowledge graph. Finally, the system achieves precise execution of control instructions through an execution feedback module and feeds back state data to the sensing end in real time, forming a complete closed-loop control system.

[0022] Specifically, the perception coding module collects multimodal operational data through a neuromorphic sensor and performs pulse coding processing to obtain a pulse coding dataset including: The neuromorphic sensor array is activated to collect multimodal operation data and obtain raw pulse data; the raw pulse data is processed by pulse coding to obtain encoded pulse sequences; the encoded pulse sequences are encapsulated to obtain pulse-coded datasets.

[0023] Specifically, the neuromorphic sensor array is activated to collect multimodal operating data and obtain raw pulse data. Specifically, the memristor-based pulse neural network sensor chip is activated to collect electrical, mechanical and temperature data in an event-driven mode. Data acquisition is triggered only when a state change exceeds a threshold, and raw pulse data containing multimodal operating characteristics is output. The original pulse data is processed by pulse coding to obtain the encoded pulse sequence. Specifically, the continuous analog signal is converted into a sparse pulse sequence, the pulse time-dependent plasticity learning rule is applied, the feature extraction and signal compression ratio are adaptively optimized, and the pulse-coded time-series pulse sequence is output. The encoded pulse sequence is encapsulated to obtain a pulse-coded dataset. Specifically, timestamps and device identifiers are added to the pulse sequence, the data is encapsulated according to a preset format, the data integrity and consistency are verified, and a standardized pulse-coded dataset is output.

[0024] Specifically, the intelligent optimization module, based on the pulse coding dataset, generates optimized protection strategies and early fault warnings through quantum optimization algorithms and chaotic detection methods, including: The pulse code dataset is decoded and reconstructed to obtain multi-dimensional operating feature parameters. Quantum optimization calculations are performed based on the feature parameter set to obtain optimized protection settings. Chaotic fault detection is performed based on the feature parameter set to obtain early fault warning information. The optimized protection settings and early fault warning information are integrated to generate a protection strategy package.

[0025] Specifically, the pulse coding dataset is decoded and reconstructed to obtain multi-dimensional operating feature parameters. Specifically, the sparse pulse sequence is reconstructed into continuous operating parameters, electrical feature parameters, mechanical feature parameters and temperature feature parameters are extracted, the statistical characteristics and changing trends of each feature parameter are calculated, and a feature parameter set containing multi-dimensional features is output. Based on the set of feature parameters, quantum optimization calculations are performed to obtain optimized protection settings. Specifically, the quantum annealing algorithm is used to solve a multi-objective optimization problem, balancing the fault detection sensitivity and selectivity requirements, and outputting the optimized protection settings parameters. Based on the set of feature parameters, chaotic fault detection is performed to obtain early fault warning information. Specifically, a Duffing oscillator chaotic system is constructed to detect weak fault features. The sensitivity of the chaotic system to initial conditions is used to identify early faults, detect high-resistance grounding faults and insulation degradation precursors, and output early fault warning information containing fault type and risk level. The optimized protection settings and early fault warning information are integrated to generate a protection strategy package. Specifically, the optimized protection settings and fault warning information are correlated and analyzed to generate a protection strategy that includes setting, warning threshold and handling suggestions. The integrity and consistency of the protection strategy are verified, and a standardized optimized protection strategy and early fault warning package are output.

[0026] Specifically, the collaborative decision-making module is used to construct a multi-agent reinforcement learning framework, and based on the optimized protection strategy and early fault warning, generates a collaborative operation strategy through game theory cooperation, including: The multi-agent system is initialized to obtain a reinforcement learning environment framework. The optimized protection strategy and early fault warning are integrated to obtain the agent's prior knowledge. The multi-agent system is trained in a game-theoretic cooperative manner to obtain a preliminary cooperative strategy. The stability of the preliminary cooperative strategy is optimized to obtain the final cooperative operation strategy.

[0027] Specifically, the multi-agent system is initialized to obtain a reinforcement learning environment framework, which involves defining multiple agents (such as circuit breaker control units, protection device units, etc.), setting up the state space (including equipment operating parameters and environmental variables), action space (such as switching operations and parameter adjustments), and reward function (based on safety and efficiency indicators), and constructing a distributed reinforcement learning environment. The optimized protection strategy and early fault warning are integrated to obtain the agent's prior knowledge. Specifically, the optimized protection strategy is used as the agent's initial policy constraint, and the early fault warning is used as an additional input to the state space. This information is encoded into the agent's prior probability distribution through knowledge distillation to enhance the guidance of the learning process. To obtain a preliminary cooperative strategy, the multi-agent cooperative strategy is trained through game theory collaboration. Specifically, the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm is used, combined with the concept of Nash equilibrium in game theory, to design a cooperative reward mechanism (such as sharing global goal rewards and individual competitive rewards). Through iterative training, the agents learn cooperative and competitive behaviors and generate a preliminary cooperative strategy. The initial cooperative strategy is optimized for stability to obtain the final cooperative operation strategy. Specifically, the strategy avoids policy oscillation or getting stuck in local optima by using opponent modeling and policy regularization methods. The robustness of the strategy is improved by using an experience replay buffer and target network technology, and the cooperative operation strategy that satisfies multi-objective optimization is output.

[0028] Specifically, the digital twin module is used to construct a digital twin of the circuit breaker and to verify and adapt the cooperative operation strategy through a meta-learning algorithm, resulting in a verified control strategy including: Multi-dimensional modeling of the circuit breaker physical entity is performed to obtain an initial digital twin. The initial digital twin is then integrated with a meta-learning framework to obtain an adaptive digital twin. The collaborative operation strategy is verified using a digital twin to obtain a strategy performance evaluation. Based on the strategy performance evaluation, meta-learning adaptation is performed to obtain a verified control strategy.

[0029] Specifically, the physical entity of the circuit breaker is modeled in multiple dimensions to obtain an initial digital twin. Specifically, based on the electrical parameters, mechanical structural characteristics and thermodynamic behavior of the circuit breaker, combined with historical operating data, a high-fidelity simulation model is constructed through finite element analysis and system identification methods to form an initial digital twin that can reflect the status of the circuit breaker in real time. The initial digital twin is integrated with a meta-learning framework to obtain an adaptive digital twin. Specifically, the model-independent meta-learning (MAML) algorithm is used to combine the digital twin with a meta-learner. Through multi-task training, the twin can quickly adapt to different operating conditions and policy inputs, forming an adaptive digital twin with self-learning capabilities. The cooperative operation strategy is verified by digital twin to obtain a strategy performance evaluation. Specifically, the cooperative operation strategy is simulated and executed on an adaptive digital twin, key indicators (such as operation delay, arc suppression effect, and mechanical wear degree) are collected, and the strategy performance under multiple scenarios is compared through a meta-learner to generate a strategy performance evaluation that includes robustness score and risk prediction. Meta-learning adaptation is performed based on policy performance evaluation to obtain a validated control policy. Specifically, the rapid adaptability of the meta-learner is used to adjust the parameters of the cooperative operation policy (such as action timing and force control threshold) according to the performance evaluation results. The policy generalization is optimized through gradient update and policy distillation. Finally, the validated control policy is output after digital twin verification and adaptation.

[0030] Specifically, the causal reasoning module evaluates the effectiveness of the verified control strategy based on a causal model, and optimizes the decision-making process through counterfactual reasoning to obtain the final control strategy, including: A causal graph model is constructed based on the verified control strategy and related operational data to obtain a structured causal network. The effect of the strategy is evaluated based on the structured causal network to obtain the causal effect quantification result. Counterfactual scenarios are generated and reasoned for the verified control strategy to obtain optimization decision suggestions. The strategy is iteratively optimized based on the counterfactual reasoning result to obtain the final control strategy.

[0031] Specifically, a causal graph model is constructed on the verified control strategy and related operating data to obtain a structured causal network. Specifically, based on the circuit breaker's historical operating data and the parameters of the verified control strategy, a causal discovery algorithm (such as the PC algorithm or FCI algorithm) is used to identify the causal relationship between key variables (such as operating timing, current intensity, and mechanical stress) and results (such as failure rate and equipment life), and a causal network in the form of a directed acyclic graph (DAG) is constructed. Based on the structured causal network, the strategy effect is evaluated to obtain the causal effect quantification results. Specifically, through intervention reasoning (Do-calculus) and potential outcome model, the average treatment effect (ATE) and conditional average treatment effect (CATE) of the control strategy on key indicators (such as operation success rate and energy consumption efficiency) are quantified and verified. The effective components and potential defects in the strategy are identified, and an evaluation report containing benefit scores and risk probabilities is generated. The verified control strategy is subjected to counterfactual scenario generation and reasoning to obtain optimized decision suggestions. Specifically, the counterfactual reasoning framework (such as counterfactual calculation based on structural equation model) is used to simulate the potential results of taking alternative actions (such as adjusting operation delay or changing force control parameters) under the same initial conditions, compare the expected benefits and risks of different decisions, and output optimized suggestions including the direction of strategy modification and the range of parameter adjustment. Based on the counterfactual reasoning results, the strategy is iteratively optimized to obtain the final control strategy. Specifically, the parameters of the verified control strategy are adjusted by combining the optimization decision suggestions and using gradient descent or Bayesian optimization methods to ensure that the strategy achieves optimal performance in both causal effect and counterfactual performance, while satisfying safety and reliability constraints. Finally, the final control strategy that has been verified by causality and optimized by counterfactual reasoning is generated.

[0032] Specifically, the learning evolution module achieves continuous optimization and evolution of the control model through neural architecture search and weight consolidation algorithms, including: The current control model is evaluated for performance and analyzed for architecture to obtain model optimization requirements. Based on the model optimization requirements, a neural architecture search is performed to obtain an optimized model structure. The optimized model structure is then subjected to weight consolidation training to obtain a stable evolutionary model. The stable evolutionary model is then deployed and verified online to obtain the final optimized model.

[0033] Specifically, the current control model is evaluated for performance and its architecture is analyzed to obtain the model optimization requirements. Specifically, the performance indicators (such as response delay and control accuracy) of the control model under the current operating conditions are evaluated through the validation set, the matching degree between the model structure complexity and performance is analyzed, and the model components and parameters that need to be optimized are identified. Based on the model optimization requirements, a neural architecture search is performed to obtain an optimized model structure. Specifically, a neural architecture search method based on reinforcement learning is adopted, with control accuracy and inference speed as optimization objectives. Within a preset search space, the optimal number of network layers, connection methods, and activation functions are automatically explored to generate an optimized model structure that balances performance and efficiency. The optimized model structure is subjected to weight consolidation training to obtain a stable evolutionary model. Specifically, the elastic weight consolidation algorithm is applied to apply regularization constraints to important parameters when training new tasks to protect learned knowledge from being overwritten, while allowing non-important parameters to be updated freely, so as to achieve continuous accumulation of model capabilities without catastrophic forgetting. The stable evolutionary model is deployed and verified online to obtain the final optimized model. Specifically, the optimized model is tested in parallel in the simulation environment and the actual system. The performance of the old and new models is compared through A / B testing. After confirming the performance improvement, the model is hot-updated to achieve continuous evolution of the control model.

[0034] Specifically, the semantic communication module converts the final control strategy into semantic instructions, verifies and reasons using a knowledge graph, and generates standardized control instructions, including: The final control strategy is semantically parsed and transformed to obtain initial semantic instructions. The initial semantic instructions are then subjected to knowledge graph consistency verification to obtain semantic verification results. Based on the semantic verification results, knowledge graph reasoning optimization is performed to obtain optimized semantic instructions. The optimized semantic instructions are then standardized, encoded, and encapsulated to obtain standardized control instructions.

[0035] Specifically, the final control strategy is semantically parsed and transformed to obtain initial semantic instructions. Specifically, natural language processing technology is used to convert the operation commands, parameter settings and conditional logic in the final control strategy into ontology-based semantic representations to form structured initial semantic instructions, including action type, target object and execution parameters. The initial semantic instructions are subjected to knowledge graph consistency verification to obtain semantic verification results. Specifically, the initial semantic instructions are matched with the pre-built circuit breaker domain knowledge graph. The concepts, attributes and relationships in the instructions are checked for consistency with the domain knowledge through graph query and rule reasoning. Semantic conflicts or logical errors are identified and marked. Based on the semantic verification results, knowledge graph reasoning optimization is performed to obtain optimized semantic instructions. Specifically, the semantic instructions are logically optimized and parameters are adjusted using the causal reasoning and rule engine of the knowledge graph, based on domain constraints and historical operation data, to generate safer and more efficient optimized semantic instructions. The optimized semantic instructions are standardized through encoding and encapsulation to obtain standardized control instructions. Specifically, the optimized semantic instructions are converted into a machine-executable standardized format (such as JSON-LD or OPC UA protocol), and timestamps, priorities, and checksums are added to ensure the integrity, parsing, and interoperability of the instructions, ultimately generating standardized control instructions.

[0036] Specifically, the execution feedback module executes the standardized control commands and collects status data to feed back to the sensing encoding module, forming a closed-loop control system, including: The standardized control commands are parsed and executed to obtain control execution results. After execution, the device status is collected by multiple sensors to obtain raw status data. The raw status data is preprocessed and standardized to obtain formatted feedback data. The formatted feedback data is fed back to the sensing and encoding module in real time to form a closed-loop control cycle.

[0037] Specifically, the standardized control instructions are parsed and executed to obtain control execution results. Specifically, the operation commands and parameters (such as circuit breaker opening and closing commands, force control thresholds) in the standardized control instructions are parsed, the circuit breaker operating mechanism is controlled to complete the specified actions through the actuator drive circuit, and the execution status (such as action time, position feedback) is monitored in real time, and the execution success or failure result is recorded. After execution, the device status is collected by multiple sensors to obtain raw status data. Specifically, the current, voltage, temperature rise, and mechanical displacement parameters of the circuit breaker during operation are collected by current sensors, voltage sensors, temperature sensors, and position sensors. Raw status data is obtained at a high frequency sampling rate (e.g., 10kHz) to ensure data comprehensiveness and real-time performance. The original state data is preprocessed and standardized to obtain formatted feedback data. Specifically, the Kalman filter algorithm is used to eliminate sensor noise, the data is time-stamp aligned and dimension-normalized, and converted into a standardized format (such as a Tensor or JSON sequence) compatible with the input interface of the sensing coding module to form formatted feedback data. The formatted feedback data is fed back to the sensing and coding module in real time to form a closed-loop control cycle. Specifically, the formatted feedback data is sent to the input buffer of the sensing and coding module through a low-latency communication protocol (such as MQTT or OPC UA) to trigger a new round of pulse coding processing, thereby realizing the iterative optimization of the closed-loop control system of "execution-feedback-sensing-decision".

[0038] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A collaborative control system of a primary and secondary deeply integrated pole-mounted circuit breaker, characterized in that, The method comprises the following steps: a perception coding module for collecting multi-modal operation data through a neuromorphic sensor and performing pulse coding processing to obtain a pulse coding data set; an intelligent optimization module for generating an optimized protection strategy and early fault warning based on the pulse coding data set through a quantum optimization algorithm and a chaos detection method; a collaborative decision-making module for constructing a multi-agent reinforcement learning framework and generating a collaborative operation strategy through game cooperation based on the optimized protection strategy and early fault warning; a digital twin module for constructing a circuit breaker digital twin and verifying and adapting the collaborative operation strategy through a meta-learning algorithm to obtain a verified control strategy; a causal reasoning module for evaluating the effect of the verified control strategy based on a causal model and optimizing the decision-making process through counterfactual reasoning to obtain a final control strategy; a learning evolution module for realizing continuous optimization and evolution of the control model through neural architecture search and weight consolidation algorithm; a semantic communication module for converting the final control strategy into semantic instructions, verifying and reasoning through a knowledge graph, and generating standardized control instructions; an execution feedback module for executing the standardized control instructions and collecting state data feedback to the perception coding module to form a closed-loop control system.

2. The coordinated control system of a primary and secondary deeply fused pole cutout according to claim 1, characterized in that, The perception coding module collects multi-modal operation data through a neuromorphic sensor and performs pulse coding processing to obtain a pulse coding data set, which includes: starting a neuromorphic sensor array to collect multi-modal operation data and obtain raw pulse data; performing pulse coding processing on the raw pulse data to obtain encoded pulse sequences; and performing data encapsulation on the encoded pulse sequences to obtain a pulse coding data set.

3. The coordinated control system of a primary and secondary deeply fused pole cutout according to claim 1, characterized in that, The intelligent optimization module generates an optimized protection strategy and early fault warning based on the pulse coding data set through a quantum optimization algorithm and a chaos detection method, which includes: decoding and reconstructing the pulse coding data set to obtain multi-dimensional operation feature parameters, performing quantum optimization calculation based on the feature parameter set to obtain an optimized protection setting, performing chaos fault detection based on the feature parameter set to obtain early fault warning information, and integrating the optimized protection setting and early fault warning information to generate a protection strategy package.

4. The coordinated control system of a primary and secondary deeply fused pole cutout according to claim 1, characterized in that, The collaborative decision-making module constructs a multi-agent reinforcement learning framework and generates a collaborative operation strategy through game cooperation based on the optimized protection strategy and early fault warning, which includes: initializing a multi-agent system to obtain a reinforcement learning environment framework, integrating the optimized protection strategy and early fault warning to obtain agent priori knowledge, training the multi-agent through game cooperation to obtain a preliminary collaborative strategy, and optimizing the stability of the preliminary collaborative strategy to obtain a final collaborative operation strategy.

5. The coordinated control system of a primary and secondary deeply fused pole cutout according to claim 1, characterized in that, The digital twin module constructs a circuit breaker digital twin and verifies and adapts the collaborative operation strategy through a meta-learning algorithm to obtain a verified control strategy, which includes: The multi-dimensional modeling of the circuit breaker physical entity is performed to obtain an initial digital twin, a meta-learning framework is integrated for the initial digital twin to obtain an adaptive digital twin, the collaborative operation strategy is verified by the digital twin to obtain a strategy performance evaluation, and meta-learning adaptation is performed based on the strategy performance evaluation to obtain a verified control strategy.

6. The coordinated control system of a primary and secondary deeply fused pole cutout according to claim 1, characterized in that, The effect evaluation of the verified control strategy is performed by the causal model based on the causal reasoning module, and the decision-making process is optimized through counterfactual reasoning to obtain a final control strategy. The causal graph model is constructed based on the verified control strategy and related operation data to obtain a structured causal network, the strategy effect evaluation is performed based on the structured causal network to obtain a causal effect quantization result, the counterfactual scenario generation and reasoning are performed for the verified control strategy to obtain an optimized decision-making suggestion, and the strategy iterative optimization is performed based on the counterfactual reasoning result to obtain a final control strategy.

7. The coordinated control system of a primary and secondary deeply fused pole cutout according to claim 1, characterized in that, The learning evolution module realizes the continuous optimization and evolution of the control model through neural architecture search and weight consolidation algorithm, including: The performance evaluation and architecture analysis are performed for the current control model to obtain model optimization requirements, the neural architecture search is performed based on the model optimization requirements to obtain an optimized model structure, the weight consolidation training is performed for the optimized model structure to obtain a stable evolution model, and the online deployment verification is performed for the stable evolution model to obtain a final optimized model.

8. The coordinated control system of a primary and secondary deeply fused pole cutout according to claim 1, characterized in that, The semantic communication module converts the final control strategy into semantic instructions, verifies and reasons through a knowledge graph to generate standardized control instructions, including: The semantic analysis and conversion are performed for the final control strategy to obtain initial semantic instructions, the knowledge graph consistency verification is performed for the initial semantic instructions to obtain a semantic verification result, the knowledge graph reasoning optimization is performed based on the semantic verification result to obtain optimized semantic instructions, and the standardized encoding and packaging are performed for the optimized semantic instructions to obtain standardized control instructions.

9. The coordinated control system of a primary and secondary deeply fused pole cutout according to claim 1, characterized in that, The execution feedback module executes the standardized control instructions and collects state data feedback to the perception coding module to form a closed-loop control system, including: The analysis and execution are performed for the standardized control instructions to obtain control execution results, the multi-sensor data collection is performed for the device state after execution to obtain original state data, the preprocessing and standardization are performed for the original state data to obtain formatted feedback data, and the formatted feedback data is fed back to the perception coding module in real time to form a closed-loop control cycle.

10. The coordinated control system of a primary and secondary deeply fused pole cutout according to claim 1, characterized in that, The analysis and execution are performed for the standardized control instructions to obtain control execution results, specifically: the operation commands and parameters in the standardized control instructions are analyzed, the specified actions are completed through the actuator driving circuit control circuit breaker operating mechanism, and the execution status is monitored in real time to record the execution success or failure results; The multi-sensor data collection is performed for the device state after execution to obtain original state data, specifically: the current, voltage, temperature rise, and mechanical displacement parameters of the circuit breaker during operation are collected through current sensors, voltage sensors, temperature sensors, and position sensors, the original state data is obtained at a high-frequency sampling rate to ensure data comprehensiveness and real-time performance; The original state data is pre-processed and standardized to obtain formatted feedback data, specifically: the Kalman filtering algorithm is used to eliminate sensor noise, the data is time-stamped and dimensionless, and is converted into a standardized format compatible with the input interface of the perception coding module to form formatted feedback data; The formatted feedback data is fed back to the perception coding module in real time to form a closed-loop control cycle, specifically: the formatted feedback data is sent to the input buffer of the perception coding module through a low-delay communication protocol, triggering a new round of pulse coding processing, realizing the iterative optimization of the "execution-feedback-perception-decision" closed-loop control system.

Images