A Circuit Breaker Energy Consumption Monitoring and Control Method Based on Edge Computing

Abstract

This invention discloses a circuit breaker energy consumption monitoring and control method based on edge computing, comprising the following steps: deploying an edge computing unit inside the circuit breaker; collecting and preprocessing the circuit breaker's electrical parameters; extracting energy consumption state factors and constructing a state factor sequence; constructing a state phase diagram; determining whether the current energy consumption state has an evolution trend towards the intervention prediction zone; when the determination result indicates a trend towards entering the intervention prediction zone, the edge computing unit generates a control command; the edge computing unit executes corresponding control actions according to the control command. This invention constructs a state phase diagram based on edge computing to achieve local energy consumption monitoring and control of the circuit breaker, possessing advantages such as fast response, accurate identification, and lightweight deployment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system intelligent operation and maintenance, and particularly relates to a circuit breaker energy consumption monitoring and control method based on edge computing. BACKGROUND

[0002] In the traditional power distribution system, the circuit breaker mainly undertakes the overload and short-circuit protection functions of the electrical circuit. With the development of intelligent power distribution, some high-end circuit breakers begin to integrate data acquisition and remote communication capabilities, realizing the uploading and remote monitoring of basic electrical parameters such as current and voltage. However, the existing intelligent circuit breakers mostly rely on a central server to centrally process the collected data, and the control instructions are issued by the central system based on the processing results, which leads to significant delay in energy consumption state recognition and control response, and it is difficult to meet the requirements of real-time and safety in complex load scenarios.

[0003] On the other hand, the existing energy consumption monitoring methods usually use periodic sampling and static threshold alarm mechanisms, lack modeling and prediction capabilities for the energy consumption evolution trend of electrical equipment, and are prone to false positives or false negatives. Moreover, the control strategies are mostly rule-driven, lacking dynamic intervention logic based on state evolution. Therefore, how to implement a lightweight and high-response local intelligent analysis and control mechanism within the circuit breaker body, break through the center dependency, improve real-time decision-making capabilities, and achieve more precise energy consumption management through state modeling and dynamic control strategies, has become a key problem that needs to be solved in the field of intelligent circuit breaker energy consumption monitoring and control. SUMMARY

[0004] One object of the present application is to provide a circuit breaker energy consumption monitoring and control method based on edge computing. The present application constructs a state phase diagram based on edge computing, extracts energy consumption state factors and realizes evolution trend judgment, generates control intentions locally and links control execution mechanisms to complete real-time identification and response control of energy consumption abnormalities. Compared with the traditional cloud processing method, the present application has the advantages of fast response, accurate identification, closed-loop control, and low resource occupation, which improves the intelligent level and operation reliability of the circuit breaker in distributed scenarios, and is suitable for energy consumption safety monitoring and adaptive control in various power distribution systems.

[0005] According to the circuit breaker energy consumption monitoring and control method based on edge computing of the present application, the following steps are included: S1, deploying an edge computing unit inside the circuit breaker; S2, collecting electrical parameters of the circuit breaker, constructing an electrical parameter sequence, and pre-processing the electrical parameter sequence; S3, extracting energy consumption state factors based on the pre-processed electrical parameter sequence, and constructing a state factor sequence; S4. Map the state factor sequence to nodes in the state phase graph. By calculating the evolutionary similarity and trend direction between state factors, establish directed edges between nodes to form the state phase graph. S5. Based on the trend consistency and weight of the adjacent edges of the current node in the state phase diagram, determine whether the current energy consumption state has an evolution trend toward the intervention prediction zone. S6. When the judgment result indicates that there is a trend to enter the intervention prediction zone, the edge computing unit generates control commands based on the current node's status, evolution path, and local control resource status. S7. The edge computing unit executes corresponding control actions according to the control instructions.

[0006] Optionally, the electrical parameters include current, voltage, active power, reactive power, power factor, and frequency, and the preprocessing includes outlier removal, moving average filtering, data completion, time alignment, unit unification, and normalization.

[0007] Optionally, S3 specifically includes: S31. Input the preprocessed electrical parameter sequence into the state factor extraction process of the edge computing unit, set the sliding time window length and sliding step size, and construct the electrical parameter sampling subsequence in each window; S32. Within each sliding time window, extract energy consumption state factors based on the electrical parameter sampling subsequence to form a state factor vector, including: The fluctuation amplitude factor is used to reflect the overall fluctuation level of each electrical parameter within the current window. ; in, Indicates volatility factor. Indicates the number of dimensions of electrical parameters. This indicates taking the maximum value. This indicates taking the minimum value. Indicates the first The first time point Individual electrical parameter values; The rate of change factor is used to characterize how quickly electrical parameters change. ; in, Represents the rate of change factor. Indicates the length of the sliding time window. Indicates the first The first time point Individual electrical parameter values; The oscillation period factor represents the frequency of reversal in the direction of change of electrical parameters per unit time. ; in, Indicates the oscillation cycle factor. This is a logical judgment function; it outputs 1 if the condition is met, and 0 otherwise. Indicates the first The first time point Individual electrical parameter values; The trend slope factor indicates the strength of the directional trend of the overall electrical parameter change: ; in, Indicates the trend slope factor. Indicates the first The linear slope of each electrical parameter obtained by least-squares fitting within the current window; S33. Organize the state factor vectors extracted from all sliding windows in chronological order to construct a state factor sequence; S34. All state factor vectors will be cached in the temporal structure of the edge computing unit in chronological order and timestamped synchronously.

[0008] Optionally, S4 specifically includes: S41. Extract the state factor vector from the state factor sequence; S42. Map each state factor vector to a node in the state phase diagram and construct a node set, wherein each node in the node set uniquely corresponds to a state factor vector. S43. For any two nodes in the node set, calculate the state factor difference vector, which is used to measure the magnitude and direction of state evolution. S44. Introduce an edge-side attention weight mechanism to model the sensitivity of the state factor difference vector using normalization, and obtain the attention weights: ; in, Indicates the first Attention weights for each state factor, Indicates the first The static importance coefficient of each state factor. State factor At the node and The state factor difference vector between them This represents the natural exponential function. Indicates the first The static importance coefficient of each state factor. State factor At the node and The state factor difference vector between them; S45. Calculate the Euclidean distance between state factor vectors and calculate the trend consistency coefficient, where the trend consistency coefficient represents the cosine value of the angle between the state factor difference vector and the predefined evolution direction vector; Select the time periods with stable breaker load and stable energy consumption from the historical operation data. In these periods, extract the corresponding state factor vectors, calculate the average adjacent difference vector of the state factor vectors, and obtain the predefined evolution direction vector; S46. Integrate the state similarity, trend consistency, and attention weights to calculate the directed edge weight of the directed edge: ; where represents the weight of the directed edge from node to node , represents the exponential function, represents the squared Euclidean distance of the state factor difference vector, represents the distance scale adjustment parameter, represents the state factor difference vector between nodes and , represents the state factor difference vector between nodes and ; S47. Each node only retains the top edges with the largest weights among its outgoing edges as valid connections to form a sparse directed graph structure, obtaining the state phase diagram , where represents the state phase diagram, represents the node set, represents the directed edge set, represents the edge weight set.

[0009] Optionally, the specific steps of S5 include: S51. For each edge in the state phase diagram, determine whether the target node belongs to the intervention prediction area. If it does, record it as the potential intervention edge set, and at the same time record the trend consistency coefficient and edge weight of this edge; The determination of whether the target node belongs to the intervention prediction area includes: is the set warning threshold for the energy consumption state factor. Any state node exceeding the warning threshold range is marked as an intervention prediction node; S52. Define the evolution intervention trend score: ; where The evolutionary intervention trend score describes the overall degree to which the current state's evolutionary path points towards a high-risk area. Indicates from node Pointing to node The directed edge, Represents the set of potential intervention edges. Representing an edge The weight, Representing an edge The trend consistency coefficient; S53. Setting dynamic thresholds based on evolutionary intervention trend scores: ; in, Indicates dynamic threshold. This represents the mean score of historical evolutionary intervention trends. The standard deviation of the historical evolution trend score This represents the adjustment coefficient, used to adjust the intervention sensitivity; S54. Compare the evolutionary intervention trend score with the dynamic threshold. If the evolutionary intervention trend score is greater than or equal to the dynamic threshold, it is determined that the current energy consumption state has a significant evolutionary trend toward the intervention prediction zone. Otherwise, maintain state monitoring and update the state phase map window. The intervention prediction zone is a predefined set of high-risk states that reflects potential abnormal energy consumption patterns.

[0010] Optionally, S6 specifically includes: S61. When the edge computing unit determines that the current energy consumption state node has an evolution trend toward the intervention prediction area, it first obtains the state factor vector corresponding to this node. S62. Based on the position of the current node in the state phase diagram, use the Dijkstra method to extract all intermediate nodes on the shortest path from the node to the intervention prediction area, and accumulate the edge weights between each adjacent node in the path to obtain the path risk score of the evolution path, thereby quantifying the path risk level of the current state leading to the high-risk area. S63. The edge computing unit constructs a control intent vector based on the current node's state factor vector and path risk score. The control intent vector includes circuit breaker control intent, current limiting control intent, load switching intent, and alarm prompt diagram, which are used to represent the necessity of triggering various control operations. The generation of the control intent vector is calculated based on a linear combination of the weighted score of the state factor vector and the path risk score. S64. Obtain the local control resources available to the circuit breaker at the current moment. The local control resources include hardware channel status and electrical control allowable conditions. Based on the local control resources, eliminate control items that cannot be executed immediately, and retain only control items that currently have the conditions to be executed and whose control intent strength exceeds a set threshold as candidate control operations. The control intent strength is an element component of the control intent vector. S65. The candidate control operations are prioritized according to the intensity of the control intent to form the final control instruction sequence. The control instructions include circuit breaker control instructions, current limiting control instructions, load switching control instructions, and alarm prompt control instructions.

[0011] The beneficial effects of this invention are: This invention integrates an edge computing unit within the circuit breaker, enabling localized processing of electrical parameter acquisition, energy consumption status identification, and control command execution. This overcomes the limitations of traditional smart circuit breakers that rely on a central server for analysis and decision-making, improving the real-time performance and independence of energy consumption monitoring and response control. Furthermore, this invention introduces an electrical parameter preprocessing and state factor extraction mechanism, constructing a state factor sequence and mapping it to a state phase diagram. Through comprehensive modeling of evolutionary similarity and trend direction, it achieves accurate identification of energy consumption state evolution trends, providing reliable prior support for subsequent control strategies.

[0012] Furthermore, this invention introduces an edge-side attention mechanism and a sparse connection strategy in the construction of the state phase diagram, which improves the accuracy of state modeling while reducing computational complexity, adapting to the operating environment where circuit breaker edge resources are limited. For the intervention and control of energy consumption anomalies, this invention generates control intentions by fusing state factors and path risk information, and generates optimal control commands by combining local control resource constraints, forming a closed-loop response chain of "identification-judgment-decision-execution," enhancing the proactive management capability of high-risk energy consumption modes. Overall, this invention realizes the evolution of circuit breakers from data acquisition terminals to local intelligent decision-making nodes, improving the operational safety, response timeliness, and control accuracy of the power system. Attached Figure Description

[0013] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Fig. 1 This is a flowchart of a circuit breaker energy consumption monitoring and control method based on edge computing proposed in this invention; Fig. 2 This is a schematic diagram illustrating the state phase diagram construction and edge-side attention-weighted directed edge establishment of a circuit breaker energy consumption monitoring and control method based on edge computing proposed in this invention. Detailed Implementation

[0014] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0015] refer to Figs. 1-2 A method for monitoring and controlling the energy consumption of circuit breakers based on edge computing, comprising the following steps: S1. Deploy edge computing units inside the circuit breaker; S2. Collect the electrical parameters of the circuit breaker, construct the electrical parameter sequence, and preprocess the electrical parameter sequence; S3. Extract energy consumption state factors based on the preprocessed electrical parameter sequence and construct a state factor sequence; S4. Map the state factor sequence to nodes in the state phase graph. By calculating the evolutionary similarity and trend direction between state factors, establish directed edges between nodes to form the state phase graph. S5. Based on the trend consistency and weight of the adjacent edges of the current node in the state phase diagram, determine whether the current energy consumption state has an evolution trend toward the intervention prediction zone. S6. When the judgment result indicates that there is a trend to enter the intervention prediction zone, the edge computing unit generates control commands based on the current node's status, evolution path, and local control resource status. S7. The edge computing unit executes corresponding control actions according to the control instructions.

[0016] This invention establishes a complete link between data acquisition, status identification, and control response by deploying an edge computing unit inside the circuit breaker. This enables the circuit breaker to autonomously complete the closed loop of energy consumption monitoring and control without the participation of a central server. The entire process covers electrical parameter acquisition, state factor construction, evolution trend analysis, and control execution, achieving high-frequency monitoring of electricity consumption behavior and rapid risk response. This improves the real-time performance and reliability of the power distribution system. The method is adaptable to deployment requirements in multiple scenarios and has the advantages of low latency, low dependency, and intelligent response, making it suitable for the distributed management trend of modern power systems.

[0017] In this embodiment, the electrical parameters include current, voltage, active power, reactive power, power factor, and frequency. The preprocessing includes outlier removal, moving average filtering, data completion, time alignment, unit unification, and normalization.

[0018] This invention introduces a refined data preprocessing workflow in the energy consumption monitoring process. By performing anomaly removal, moving average filtering, data completion, time alignment, and dimensional unification and normalization on multi-dimensional electrical parameters such as current, voltage, active power, reactive power, power factor, and frequency, the input data has high continuity and consistency. This processing workflow effectively eliminates interference caused by sensor fluctuations, instantaneous anomalies, or communication jitter in the data, thereby ensuring the stability and accuracy of subsequent state recognition and trend analysis. It is a fundamental guarantee step for realizing edge intelligent judgment.

[0019] In this embodiment, S3 specifically includes: S31. Input the preprocessed electrical parameter sequence into the state factor extraction process of the edge computing unit, set the sliding time window length and sliding step size, and construct the electrical parameter sampling subsequence in each window; S32. Within each sliding time window, extract energy consumption state factors based on the electrical parameter sampling subsequence to form a state factor vector, including: The fluctuation amplitude factor is used to reflect the overall fluctuation level of each electrical parameter within the current window. ; in, Indicates volatility factor. Indicates the number of dimensions of electrical parameters. This indicates taking the maximum value. This indicates taking the minimum value. Indicates the first The first time point Individual electrical parameter values; The rate of change factor is used to characterize how quickly electrical parameters change. ; in, Represents the rate of change factor. Indicates the length of the sliding time window. Indicates the first The first time point Individual electrical parameter values; The oscillation period factor represents the frequency of reversal in the direction of change of electrical parameters per unit time. ; in, Indicates the oscillation cycle factor. This is a logical judgment function; it outputs 1 if the condition is met, and 0 otherwise. Indicates the first The first time point Individual electrical parameter values; The trend slope factor indicates the strength of the directional trend of the overall electrical parameter change: ; in, Indicates the trend slope factor. Indicates the first The linear slope of each electrical parameter obtained by least-squares fitting within the current window; S33. Organize the state factor vectors extracted from all sliding windows in chronological order to construct a state factor sequence; S34. All state factor vectors will be cached in the temporal structure of the edge computing unit in chronological order and timestamped synchronously.

[0020] This invention constructs a state factor extraction process through a sliding time window, dynamically extracting key indicators such as fluctuation amplitude, rate of change, oscillation period, and trend slope from the processed electrical parameters, and forming a structured state factor sequence to comprehensively characterize the current energy consumption state of the circuit breaker. This method integrates short-term dynamic and long-term trend information, effectively enhancing the perception of complex load changes, and provides clear temporal logic for subsequent graph structure modeling through sequence organization, ensuring efficient and accurate energy consumption behavior identification even when edge device resources are limited.

[0021] In this embodiment, S4 specifically includes: S41. Extract the state factor vector from the state factor sequence; S42. Map each state factor vector to a node in the state phase diagram and construct a node set, wherein each node in the node set uniquely corresponds to a state factor vector. S43. For any two nodes in the node set, calculate the state factor difference vector, which is used to measure the magnitude and direction of state evolution. S44. Introduce an edge-side attention weight mechanism to model the sensitivity of the state factor difference vector using normalization, and obtain the attention weights: ; in, Indicates the first Attention weights for each state factor, Indicates the first The static importance coefficient of each state factor. State factor At the node and The state factor difference vector between them This represents the natural exponential function. Indicates the first The static importance coefficient of each state factor. State factor At the node and The state factor difference vector between them; S45. Calculate the Euclidean distance between state factor vectors and calculate the trend consistency coefficient, which represents the cosine of the angle between the state factor difference vector and the predefined evolution direction vector. The circuit breaker load is stable and energy consumption is stable during the historical operation data. During these periods, the corresponding state factor vectors are extracted, and the average adjacent difference vector of the state factor vectors is calculated to obtain the predefined evolution direction vector. S46. Combine state similarity, trend consistency, and attention weights to calculate directed edges. Directed edge weights: ; in, Indicates from node To the node The weight of the directed edge. Represents an exponential function. The squared Euclidean distance represents the vector of state factor differences. This indicates the distance scale adjustment parameter. Represents a node and The state factor difference vector between them Represents a node and The trend consistency coefficient between them; S47. Each node retains only the edge with the largest weight among its outgoing edges. Edges serve as effective connections, forming a sparse directed graph structure, resulting in a state-phase graph. ,in Represents the state phase diagram. Represents a set of nodes. Represents a set of directed edges. Let represent the set of edge weights.

[0022] This invention models the energy consumption state evolution process of a circuit breaker using a graph structure by mapping state factor sequences to nodes and constructing a state phase graph, forming a state topology that is easy to track and analyze. In particular, an edge-side attention mechanism is introduced to model the state factor difference vector, dynamically allocating the attention level of each state factor, so that the edge weights in the graph more realistically reflect the importance and consistency of state transitions. This mechanism, combined with historical evolution direction information, enhances the ability to identify potential risk paths, and improves the discriminative ability and generalization performance of trend graph construction while ensuring controllable computational overhead.

[0023] In this embodiment, S5 specifically includes: S51. For each edge in the state phase diagram, determine whether the target node belongs to the intervention prediction zone. If it does, record it as a potential intervention edge set, and record the trend consistency coefficient and edge weight of this edge. The determination of whether a target node belongs to the intervention prediction zone includes: Set an early warning threshold for the energy consumption state factor, and mark any state node that exceeds the early warning threshold as an intervention prediction node; S52. Define the evolutionary intervention trend score: ; in, The evolutionary intervention trend score describes the overall degree to which the current state's evolutionary path points towards a high-risk area. Indicates from node Pointing to node The directed edge, Represents the set of potential intervention edges. Representing an edge The weight, Representing an edge The trend consistency coefficient; S53. Setting dynamic thresholds based on evolutionary intervention trend scores: ; in, Indicates dynamic threshold. This represents the mean score of historical evolutionary intervention trends. The standard deviation of the historical evolution trend score This represents the adjustment coefficient, used to adjust the intervention sensitivity; S54. Compare the evolutionary intervention trend score with the dynamic threshold. If the evolutionary intervention trend score is greater than or equal to the dynamic threshold, it is determined that the current energy consumption state has a significant evolutionary trend toward the intervention prediction zone. Otherwise, maintain state monitoring and update the state phase map window. The intervention prediction zone is a predefined set of high-risk states that reflects potential abnormal energy consumption patterns.

[0024] This invention uses an evolutionary trend scoring mechanism to model the edge relationships in the state phase diagram for risk perception, constructs an intervention prediction zone, and dynamically judges whether the current state is evolving towards a high-risk area. By introducing a trend consistency coefficient, historical evolution distribution statistics, and a dynamic threshold mechanism, the system can automatically adjust the intervention sensitivity to avoid false alarms and missed alarms. This method improves the accuracy of energy consumption anomaly judgment, enabling circuit breakers to intervene and control in advance in the early stages of state evolution, and enhancing the initiative, foresight, and adaptability of electricity consumption behavior monitoring.

[0025] In this embodiment, S6 specifically includes: S61. When the edge computing unit determines that the current energy consumption state node has an evolution trend toward the intervention prediction area, it first obtains the state factor vector corresponding to this node. S62. Based on the position of the current node in the state phase diagram, use the Dijkstra method to extract all intermediate nodes on the shortest path from the node to the intervention prediction area, and accumulate the edge weights between each adjacent node in the path to obtain the path risk score of the evolution path, thereby quantifying the path risk level of the current state leading to the high-risk area. S63. The edge computing unit constructs a control intent vector based on the current node's state factor vector and path risk score. The control intent vector includes circuit breaker control intent, current limiting control intent, load switching intent, and alarm prompt diagram, which are used to represent the necessity of triggering various control operations. The generation of the control intent vector is calculated based on a linear combination of the weighted score of the state factor vector and the path risk score. S64. Obtain the local control resources available to the circuit breaker at the current moment. The local control resources include hardware channel status and electrical control allowable conditions. Based on the local control resources, eliminate control items that cannot be executed immediately, and retain only control items that currently have the conditions to be executed and whose control intent strength exceeds a set threshold as candidate control operations. The control intent strength is an element component of the control intent vector. S65. The candidate control operations are prioritized according to the intensity of the control intent to form the final control instruction sequence. The control instructions include circuit breaker control instructions, current limiting control instructions, load switching control instructions, and alarm prompt control instructions.

[0026] This invention constructs a control intent vector and integrates current state factors and path risk information to achieve local generation and hierarchical judgment of control commands, effectively improving the control decision-making capability of edge computing units in complex states. By combining local hardware status and control allowable conditions to screen control resource constraints, the generated control commands are ensured to be executable and optimal. This method dynamically adjusts the control strategy under the premise of meeting safety requirements, and has the capabilities of precise triggering, rapid response and multi-level linkage, effectively avoiding the problems of over-control and control lag, and ensuring the stable operation of the power supply system.

[0027] Example 1: To verify the feasibility of this invention in practice, it was applied to a power distribution station system with high power load density in an industrial park. This power distribution station is mainly responsible for supplying power to multiple production workshops and power equipment. The equipment operates under complex conditions, with frequent start-ups and shutdowns, sudden load changes, and power factor fluctuations. Traditional centralized energy consumption monitoring and control schemes are significantly inadequate in terms of real-time processing, sensitivity to changes in electrical parameters, and efficiency of on-site intervention response. They cannot identify potential high energy consumption trends and hidden load anomalies in a timely manner, and are unable to effectively prevent equipment overload, inefficient operation, and potential power distribution safety hazards.

[0028] In practical applications, the edge computing unit of this invention is directly integrated into the selected circuit breaker. By connecting the electrical parameter sampling circuit and the control actuator, the immediacy and closed-loop nature of the acquisition and control are ensured. The edge computing unit collects multi-dimensional electrical parameter data of the circuit breaker in real time, including current, voltage, active power, reactive power, power factor and frequency. The data is then cleaned, aligned, dimensionally unified and normalized through a preset data preprocessing process to obtain a continuous and complete sequence of electrical parameters. Based on this, the system calculates state factor vectors, such as fluctuation amplitude, rate of change, oscillation period and trend slope, within each sliding time window to characterize the dynamic changes in energy consumption during equipment operation.

[0029] Furthermore, the system constructs a state phase diagram using a sequence of state factors and assigns importance weights to different factors in the evolution path through an edge-side attention mechanism, thus constructing a sparse directed graph structure that reflects the direction and similarity of energy consumption state evolution. During operation, when the system detects that the current state node has a tendency to evolve towards a predefined intervention prediction zone, that is, when it determines that the current operating state of the equipment has the characteristics of transitioning to a high-risk direction, the system will automatically calculate the path risk score and generate a control intent vector based on the combination of state factors. This control intent vector is further combined with the current hardware resource status of the circuit breaker and the control execution conditions to filter out executable control actions and arrange them according to priority. Finally, a control command is generated and sent to the control execution mechanism to complete the corresponding local control response such as circuit breaking, current limiting, load switching, or alarm prompts.

[0030] This method was continuously operated for two weeks at the power supply monitoring site of a manufacturing enterprise in southern China in June 2025. The system achieved continuous modeling and dynamic adjustment of the equipment operating status, successfully identified multiple abnormal energy consumption trends and completed early warning and disposal in a timely manner. By comparing and analyzing historical data records and response links during this deployment, the edge deployment, state factor modeling and attention-guided state phase diagram construction mechanism adopted by this invention are significantly better than the original centralized processing-based control mechanism in terms of reducing prediction lag, improving the accuracy of anomaly identification and control response efficiency.

[0031] To verify the performance of the present invention in practice, it was compared with traditional methods, and the results are shown in Table 1.

[0032] Table 1. Comparison of Key Performance of Edge Computing-Based Circuit Breaker Energy Consumption Monitoring Method and Traditional Method

[0033] A thorough analysis of Table 1 can comprehensively explain the performance advantages of this invention and their causes from three dimensions: anomaly identification capability, response efficiency, and control execution effect.

[0034] In terms of the accuracy of identifying abnormal energy consumption states, this invention achieves a high recognition rate of 91.3%, which is about 6.6 percentage points higher than the 84.7% of the traditional method. This improvement is due to the state factor extraction mechanism and state phase diagram construction method introduced in the state modeling process of this invention. In particular, the addition of high-dimensional behavioral feature factors such as trend slope and oscillation period to the state factors makes the system more sensitive and precise in characterizing the boundary of abnormal energy consumption states, reducing the probability of misjudgment and missed judgment. In addition, by deploying the computing model on the edge side, dynamic tracking of state evolution can be achieved, which enhances the adaptive recognition capability of the model.

[0035] Regarding response latency, the average latency of the energy consumption status response of this invention is 67ms, while that of the traditional method is 131ms. The core of this performance improvement lies in the introduction of edge computing architecture. By moving the data processing and judgment logic forward to the circuit breaker itself, the network round trip and queuing delays that the traditional method relies on remote server processing are avoided. At the same time, this invention uses a sparse directed edge strategy in the construction of the state graph, which effectively reduces the computational cost of irrelevant path search and further accelerates the process of state judgment and instruction generation.

[0036] In terms of the lead time for identifying high-risk trends, this invention achieves 18.5 seconds, while the traditional method only achieves 4.2 seconds, showing a significant improvement. This advantage mainly stems from the continuous modeling of the evolution of state factors and the combination of a path trend scoring mechanism. This allows for earlier identification of risk paths pointing to the intervention prediction zone on the state phase diagram. By calculating the predefined evolution direction vector and the trend consistency coefficient, early signals of the state evolving towards the high-risk zone are effectively captured.

[0037] In terms of the success rate of control command issuance, the present invention achieved 95.1%, while the traditional method achieved 91.3%, an improvement of 3.8 percentage points. This indicates that the present invention has stronger stability in the control execution link. This is due, on the one hand, to the local direct connection between the edge unit and the control execution mechanism, which reduces the probability of failure in the intermediate communication links. On the other hand, it benefits from the dual screening strategy of control resource constraints and control intent strength, which ensures that every issued control command has executability and high priority.

[0038] Regarding the average execution time of control commands, this invention achieves 45ms, while the traditional method achieves 97ms. This is because the local decision-making mechanism adopted in this invention eliminates the need for control commands to rely on the central server for judgment, thus avoiding bottlenecks in data backhaul and centralized computation. Furthermore, a priority sorting mechanism is introduced in the command generation stage, which can quickly locate the optimal control path, reduce redundant scheduling, and improve execution efficiency.

[0039] In summary, this invention improves the response speed, identification accuracy, and execution reliability of energy consumption status monitoring and control through edge computing architecture, state phase graph modeling, trend scoring mechanism, and local resource awareness control strategy, which is superior to traditional centralized energy consumption management solutions.

[0040] The above description is only a preferred embodiment 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 method for monitoring and controlling the energy consumption of circuit breakers based on edge computing, characterized in that, Includes the following steps: S1. Deploy edge computing units inside the circuit breaker; S2. Collect the electrical parameters of the circuit breaker, construct the electrical parameter sequence, and preprocess the electrical parameter sequence; S3. Extract energy consumption state factors based on the preprocessed electrical parameter sequence and construct a state factor sequence; S4. Map the state factor sequence to nodes in the state phase graph. By calculating the evolutionary similarity and trend direction between state factors, establish directed edges between nodes to form the state phase graph. S5. Based on the trend consistency and weight of the adjacent edges of the current node in the state phase diagram, determine whether the current energy consumption state has an evolution trend toward the intervention prediction zone. S6. When the judgment result indicates that there is a trend to enter the intervention prediction zone, the edge computing unit generates control commands based on the current node's status, evolution path, and local control resource status. S7. The edge computing unit executes corresponding control actions according to the control instructions.

2. The circuit breaker energy consumption monitoring and control method based on edge computing according to claim 1, characterized in that, The electrical parameters include current, voltage, active power, reactive power, power factor, and frequency. The preprocessing includes outlier removal, moving average filtering, data completion, time alignment, unit unification, and normalization.

3. The circuit breaker energy consumption monitoring and control method based on edge computing according to claim 1, characterized in that, S3 specifically includes: S31. Input the preprocessed electrical parameter sequence into the state factor extraction process of the edge computing unit, set the sliding time window length and sliding step size, and construct the electrical parameter sampling subsequence in each window; S32. Within each sliding time window, extract energy consumption state factors based on the electrical parameter sampling subsequence to form a state factor vector, including: The fluctuation amplitude factor is used to reflect the overall fluctuation of each electrical parameter within the current window; The rate of change factor is used to characterize how fast electrical parameters change. The oscillation period factor represents the frequency of reversal in the direction of change of electrical parameters per unit time; The trend slope factor indicates the strength of the directional trend of the overall electrical parameter change. S33. Organize the state factor vectors extracted from all sliding windows in chronological order to construct a state factor sequence; S34. All state factor vectors will be cached in the temporal structure of the edge computing unit in chronological order and timestamped synchronously.

4. The circuit breaker energy consumption monitoring and control method based on edge computing according to claim 1, characterized in that, S4 specifically includes: S41. Extract the state factor vector from the state factor sequence; S42. Map each state factor vector to a node in the state phase diagram and construct a node set, wherein each node in the node set uniquely corresponds to a state factor vector. S43. For any two nodes in the node set, calculate the state factor difference vector, which is used to measure the magnitude and direction of state evolution. S44. Introduce an edge-side attention weight mechanism to normalize the sensitivity model of the state factor difference vector and obtain the attention weight; S45. Calculate the Euclidean distance between state factor vectors and calculate the trend consistency coefficient, which represents the cosine of the angle between the state factor difference vector and the predefined evolution direction vector. The circuit breaker load is stable and energy consumption is stable during the historical operation data. During these periods, the corresponding state factor vectors are extracted, and the average adjacent difference vector of the state factor vectors is calculated to obtain the predefined evolution direction vector. S46. Combine state similarity, trend consistency, and attention weights to calculate directed edges. Directed edge weights: ； in, Indicates from node To the node The weight of the directed edge. Represents an exponential function. The squared Euclidean distance represents the vector of state factor differences. This indicates the distance scale adjustment parameter. Represents a node and The state factor difference vector between them Represents a node and The trend consistency coefficient between them; S47. Each node retains only the edge with the largest weight among its outgoing edges. Edges serve as effective connections, forming a sparse directed graph structure, resulting in a state-phase graph. ,in Represents the state phase diagram. Represents a set of nodes. Represents a set of directed edges. Let represent the set of edge weights.

5. The circuit breaker energy consumption monitoring and control method based on edge computing according to claim 1, characterized in that, S5 specifically includes: S51. For each edge in the state phase diagram, determine whether the target node belongs to the intervention prediction zone. If it does, record it as a potential intervention edge set, and record the trend consistency coefficient and edge weight of this edge. The determination of whether a target node belongs to the intervention prediction zone includes: Set an early warning threshold for the energy consumption state factor, and mark any state node that exceeds the early warning threshold as an intervention prediction node; S52. Define the evolutionary intervention trend score; S53. Set dynamic thresholds based on evolutionary intervention trend scores; S54. Compare the evolutionary intervention trend score with the dynamic threshold. If the evolutionary intervention trend score is greater than or equal to the dynamic threshold, it is determined that the current energy consumption state has a significant evolutionary trend toward the intervention prediction zone. Otherwise, maintain state monitoring and update the state phase map window. The intervention prediction zone is a predefined set of high-risk states that reflects potential abnormal energy consumption patterns.

6. The circuit breaker energy consumption monitoring and control method based on edge computing according to claim 1, characterized in that, S6 specifically includes: S61. When the edge computing unit determines that the current energy consumption state node has an evolution trend toward the intervention prediction area, it first obtains the state factor vector corresponding to this node. S62. Based on the position of the current node in the state phase diagram, use the Dijkstra method to extract all intermediate nodes on the shortest path from the node to the intervention prediction area, and accumulate the edge weights between each adjacent node in the path to obtain the path risk score of the evolution path, thereby quantifying the path risk level of the current state leading to the high-risk area. S63. The edge computing unit constructs a control intent vector based on the current node's state factor vector and path risk score. The control intent vector includes circuit breaker control intent, current limiting control intent, load switching intent, and alarm prompt diagram, which are used to represent the necessity of triggering various control operations. The generation of the control intent vector is calculated based on a linear combination of the weighted score of the state factor vector and the path risk score. S64. Obtain the local control resources available to the circuit breaker at the current moment. The local control resources include hardware channel status and electrical control allowable conditions. Based on the local control resources, eliminate control items that cannot be executed immediately, and retain only control items that currently have the conditions to be executed and whose control intent strength exceeds a set threshold as candidate control operations. The control intent strength is an element component of the control intent vector. S65. The candidate control operations are prioritized according to the intensity of the control intent to form the final control instruction sequence. The control instructions include circuit breaker control instructions, current limiting control instructions, load switching control instructions, and alarm prompt control instructions.

Images