Terminal low power consumption operation control method based on pole-mounted circuit breaker feeder

Abstract

The present application relates to the technical field of terminal control of power distribution network, and discloses a terminal low-power operation control method based on pole-mounted circuit breaker feeder.The method comprises obtaining real-time operation parameters and environmental state parameters of the pole-mounted circuit breaker feeder, wherein the real-time operation parameters include line current, voltage, power factor, etc., and the environmental state parameters include temperature, humidity, etc.; the current operation scene is determined by matching parameter characteristics in combination with a preset feeder operation knowledge graph; according to the current scene and the corresponding operation specification file, the real-time operation parameters are divided into global operation indexes reflecting the overall system condition and local operation indexes embodying the local line or equipment condition; and the two types of indexes are used to implement low-power control on the feeder, so as to realize fine and dynamic regulation and control, adapt to the load and environmental changes of the power distribution network, reduce energy waste, and improve the operation efficiency and stability of the feeder terminal.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of terminal control of power distribution network, in particular to a terminal low-power operation control method based on pole-mounted circuit breaker feeder. BACKGROUND

[0002] With the rapid development of power systems towards intelligence and distribution, as a key device in the power distribution network, the operation state of the pole-mounted circuit breaker feeder terminal directly relates to the power supply reliability and energy utilization efficiency of the power distribution network. In the current power distribution network operation system, the pole-mounted circuit breaker feeder terminal is usually controlled in a fixed mode, that is, regardless of the changes of external environmental conditions and actual load conditions of the feeder, the terminal maintains the preset fixed operation parameters. This traditional control method has obvious limitations and is difficult to adapt to the complex and variable operation requirements of the power distribution network.

[0003] In actual operation, the environmental state of the pole-mounted circuit breaker feeder varies significantly, such as temperature and humidity changes in different seasons, wind and light conditions fluctuations in different regions, etc. These environmental factors will directly affect the energy consumption level and operation stability of the feeder terminal. At the same time, the real-time operation parameters of the feeder, such as line current, voltage, power factor, etc., will also dynamically fluctuate with the changes of user power demand. However, the traditional control method lacks effective perception and integration of these dynamic parameters, and cannot adjust the operation strategy according to the actual scene, resulting in excessive energy consumption of the feeder terminal in some periods, which not only increases the operation cost of the power system, but also may reduce the service life of the equipment.

[0004] The existing technology for dividing the operation state of the feeder is usually simple, usually only based on a single parameter or fixed threshold for judgment, which is difficult to fully reflect the actual operation scene of the feeder. This extensive scene division method makes the operation control strategy lack of pertinence, and cannot realize the accurate distinction and coordinated control of global and local operation indicators. For example, in the period of low load of the feeder, if the parameters are still operated according to the high load scene, it will cause unnecessary energy waste; when local line appears abnormal, if local operation indicators cannot be identified and adjusted in time, it may cause the expansion of fault range and affect the power supply reliability.

[0005] With the continuous improvement of the country's requirements for energy saving and intelligentization of power systems, the drawbacks of the traditional pole-mounted circuit breaker feeder terminal operation control method are increasingly prominent, and a technical solution is needed that can combine real-time operation parameters and environmental state parameters to realize accurate scene identification and low-power adaptive control, to meet the needs of efficient and stable operation of the power distribution network. SUMMARY

[0006] The present application aims to provide a terminal low-power operation control method based on pole-mounted circuit breaker feeder to solve the problems in the background art.

[0007] To achieve the above-mentioned purpose, the present application provides a terminal low-power operation control method based on pole-mounted circuit breaker feeder, which comprises:

[0008] Obtaining real-time operation parameters and environmental state parameters of the pole-mounted circuit breaker feeder;

[0009] According to the real-time operation parameters and environmental state parameters, and in combination with a preset feeder operation knowledge graph, determining a current operation scenario;

[0010] Based on the current operation scenario and a corresponding operation specification file, dividing the real-time operation parameters into global operation indicators and local operation indicators;

[0011] According to the global operation indicators and the local operation indicators, performing low-power operation control on the pole-mounted circuit breaker feeder.

[0012] Preferably, the process of determining the current operation scenario comprises:

[0013] Inputting the real-time operation parameters and environmental state parameters into a multi-parameter identification model to output a plurality of key operation elements;

[0014] Inputting the key operation elements into the preset feeder operation knowledge graph, and determining a current operation scenario containing all key operation elements according to the association relationship in the knowledge graph.

[0015] Preferably, the process of dividing the global operation indicators and the local operation indicators comprises:

[0016] According to the current operation scenario, calling at least one corresponding operation specification file;

[0017] Performing semantic analysis on the operation specification file by using an intelligent model to output a plurality of specification constraint conditions;

[0018] Based on the specification constraint conditions, dividing the real-time operation parameters into global operation indicators and local operation indicators.

[0019] Preferably, the process of performing low-power operation control on the pole-mounted circuit breaker feeder comprises:

[0020] According to the divided global operation indicators and local operation indicators, splitting the preprocessed operation data into a global operation data set and a local operation data set;

[0021] inputting the global running data set and the global running index into a first analysis model corresponding to the running scene, and inputting the local running data set into a second analysis model corresponding to the running scene;

[0022] fusing output results of the first analysis model and the second analysis model to generate a low-power consumption control instruction, and recording each control process.

[0023] Preferably, the first analysis model is established based on a trend prediction method, and performs trend identification on the global running data set to obtain running state boundary information, and judges whether the global running requirement is met by comparing the fitting degree of the running state boundary information and the global running index.

[0024] The second analysis model is used for fine fluctuation identification to accurately capture and analyze high-frequency change characteristics in the local running data set.

[0025] Preferably, when identifying the running state boundary information, the global running data set performs dynamic verification on the divided global running data set, and the specific steps of the dynamic verification include:

[0026] acquiring the global running data set, calculating fluctuation amplitude and change trend index of each data section, and determining that there is potential abnormality in any data section that needs to be finely monitored if the fluctuation amplitude or the change trend index of any data section exceeds a preset threshold value.

[0027] According to the determined data section with potential abnormality, a corresponding local running subset is extracted from the global running data set.

[0028] According to abnormal characteristics under different running scenes, data screening rules associated with the abnormal characteristics are established.

[0029] Preferably, the specific steps of the dynamic verification further include:

[0030] After the local running subset is extracted from the global running data set, a real-time data exchange mechanism between the first analysis model and the second analysis model is established.

[0031] The first analysis model is used to preliminarily analyze the extracted data section, and the position, size and preliminary analysis result of the data section with potential abnormality are transmitted to the second analysis model.

[0032] After receiving the preliminary analysis result, the second analysis model performs in-depth analysis on the local running subset to identify specific abnormal mode information, and the abnormal mode information includes abnormal occurrence position, change law and association state with the global running index.

[0033] Preferably, the specific steps of the dynamic verification further include:

[0034] feedback the identified abnormal pattern information to the first analysis model;

[0035] During the multiple data extraction and model analysis processes, the verification threshold and model configuration parameters are dynamically adjusted according to the difference between each verification result and the actual operating condition;

[0036] The abnormal information, the occurrence position and the influence degree obtained through the final analysis are converted into control instructions, and the complete analysis process is recorded.

[0037] Preferably, the specific process of extracting the corresponding local operation subset from the global operation data set comprises:

[0038] determining a data section with potential abnormalities in the global operation data set, the data section with potential abnormalities being framed by a numerical range composed of multiple data point indexes;

[0039] According to the determined abnormal data section, the parameters required for extracting the local operation subset are calculated, including the starting index, data length and sampling interval of the subset;

[0040] Using the extraction parameters, the corresponding local operation subset is extracted from the global operation data set through data slicing operation;

[0041] Verify whether the extracted local operation subset completely contains the abnormal data section and is not mixed with irrelevant data.

[0042] Preferably, when the output results of the first analysis model and the second analysis model are fused, a weighted fusion method based on model confidence is adopted, and corresponding weights are allocated according to the historical accuracy of the first analysis model and the second analysis model.

[0043] Compared with the prior art, the present application has the following beneficial effects:

[0044] By obtaining the real-time operation parameters and environmental state parameters of the pole-mounted circuit breaker feeder, comprehensive and dynamic data support is provided for subsequent operation scene judgment and control strategy formulation, breaking the limitation of traditional control methods relying on fixed parameters. Real-time operation parameters can reflect the current load condition, power quality and other core information of the feeder, and environmental state parameters can reflect the influence of temperature, humidity and other external factors on terminal operation. The combination of the two makes the perception of the feeder operation state more comprehensive and accurate, avoiding control deviation caused by single parameter judgment.

[0045] The current running scene is determined in combination with the preset feeder running knowledge graph, historical running data and industry experience can be fully utilized, and accurate classification of complex running scenes is realized. The feeder running knowledge graph integrates parameter characteristics, environmental influence factors and corresponding optimization running modes in different scenes. By matching the real-time collected parameters with the scene characteristics in the knowledge graph, the running scene of the current feeder can be quickly and accurately identified, such as high-load stable running scene, low-load energy-saving scene, special environment adaptation scene and the like, laying a foundation for subsequent implementation of differentiated control strategies and solving the problem of extensive and insufficiently targeted traditional scene division.

[0046] Based on the current running scene and the corresponding running specification file, the real-time running parameters are divided into global running indicators and local running indicators, and fine management of the running state of the feeder is realized. The global running indicators can reflect the overall running condition of the entire feeder system, such as system total power and overall voltage level, and the local running indicators can reflect the running condition of a specific line section or equipment, such as local line current and branch node voltage. This division method allows the control strategy to consider both global stability and local optimization, avoiding the drawbacks of the traditional "one-size-fits-all" control. While ensuring stable operation of the entire feeder system, individualized control schemes can be developed for special situations in local areas, improving the flexibility and accuracy of control.

[0047] According to the global running indicators and the local running indicators, low-power running control is performed on the pole-mounted circuit breaker feeder, which can realize accurate regulation of energy consumption and optimal allocation of resources. In different running scenes, by regulating the global running indicators, the entire feeder system can maintain low overall energy consumption while meeting power supply demands. At the same time, optimization of local running indicators can solve the problem of excessive energy consumption in local lines or equipment, avoiding energy waste. For example, in a low-load running scene, the system's overall running power can be reduced according to the global running indicators, and the running parameters of branch lines can be adjusted according to the local running indicators to further reduce local energy consumption. In a special environment scene, local running indicators can be adjusted according to environmental state parameters to ensure that equipment adapts to environmental conditions while avoiding additional energy consumption increases due to environmental factors.

[0048] This method can realize dynamic adaptive adjustment of running control strategies. As real-time running parameters and environmental state parameters change, the running scene can be updated in real time, and the corresponding global and local running indicator division and control strategies will also be adjusted to ensure that the feeder terminal always runs in the optimal low-power state. This dynamic adaptive capability not only adapts to the dynamic changes of power distribution network load and environment, but also reduces the need for manual intervention, reduces operation and maintenance costs, prolongs the service life of equipment, and improves the running efficiency and economy of the entire power distribution network. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 The working principle diagram of the terminal low-power operation control method based on the pole-mounted circuit breaker feeder of the present application;

[0050] Figure 2 The flowchart for dividing the global operation index and the local operation index;

[0051] Figure 3 The flowchart for the low-power operation control execution;

[0052] Figure 4 The flowchart for real-time data exchange and deep analysis. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0054] Please refer to Figure 1 The present application provides a terminal low-power operation control method based on a pole-mounted circuit breaker feeder, which comprises the following steps: acquiring real-time operation parameters and environmental state parameters of the pole-mounted circuit breaker feeder, wherein the real-time operation parameters include but are not limited to voltage, current, power factor and load rate, and the environmental state parameters include temperature, humidity and wind speed; determining a current operation scene according to the real-time operation parameters and the environmental state parameters, in combination with a preset feeder operation knowledge graph, wherein the knowledge graph stores the correlation of events, states and rules in the feeder operation in the form of a graph structure; dividing the real-time operation parameters into global operation indexes and local operation indexes based on the current operation scene and a corresponding operation specification file, wherein the global operation indexes involve system-level stability parameters, and the local operation indexes focus on device-level subtle changes; and performing low-power operation control on the pole-mounted circuit breaker feeder according to the global operation indexes and the local operation indexes, so as to realize power consumption reduction by adjusting the circuit breaker operation mode or optimizing energy distribution.

[0055] Embodiment 1: Please refer to Figure 2, determining the current operating scenario first requires inputting real-time operating parameters and environmental state parameters into a multi-parameter identification model. This model uses trained machine learning algorithms to process input data, including electrical quantity measurements such as voltage, current, power factor, and load rate, as well as external condition data such as temperature, humidity, and wind speed. The multi-parameter identification model analyzes these input data through a feature extraction layer to identify key operating elements, which may include overload indicators, temperature anomaly indices, and environmental disturbance levels. Each key operating element represents an important aspect of system operation. After inputting the key operating elements into a pre-defined feeder operation knowledge graph, the system begins scenario matching. The feeder operation knowledge graph is stored in a graph structure, with nodes representing various operating states and device components, and edges representing their relationships. The knowledge graph contains multiple pre-defined operating scenario patterns, each composed of a set of key operating elements. The system searches for the most matching scenario pattern to the input key operating elements in the knowledge graph using a graph traversal algorithm. During the matching process, the correlation between elements and scenarios is calculated. When all key operating elements are highly matched with a certain scenario pattern, the scenario is determined as the current operating scenario. For example, when high load rate and high temperature indicators exist simultaneously, the "high temperature overload" operating scenario may be matched.

[0056] The division process of global and local operating indicators begins with the invocation of the operating specification file. According to the determined current operating scenario, the system retrieves the corresponding operating specification file from the specification file library. These files are stored in a structured format and contain various operating constraints and standards that need to be followed in this scenario. The intelligent model then performs semantic analysis on these operating specification files. This model uses natural language processing techniques to understand technical requirements and control conditions in the specification text. During the analysis process, it identifies constraint clauses and indicator requirements in the file and converts them into executable specification constraint conditions. Based on the parsed specification constraint conditions, the system begins to divide real-time operating parameters. Specification constraint conditions usually include system-level requirements and device-level requirements. System-level requirements often involve stable operating indicators for the entire feeder system, while device-level requirements focus on the operating state of individual components. During the division process, the system compares each real-time operating parameter with the specification constraint conditions to determine whether the parameter belongs to a global or local indicator. For example, system total load and overall voltage stability are usually classified as global operating indicators, while current fluctuations in a single circuit breaker and local temperature changes are classified as local operating indicators. After the division is complete, the system establishes an indicator mapping table, which records the indicator type and related specification constraint conditions for each operating parameter, providing a basis for subsequent low-power control.

[0057] During the entire implementation process, the data flow and processing steps follow a strict logical order, the output of the multi-parameter identification model serves as the input of the knowledge graph, and the matching result of the knowledge graph triggers the calling of the operation specification file, forming a complete processing chain; each processing link is provided with a data checking mechanism to ensure the accuracy and reliability of the intermediate results, for example, in the key operation element extraction stage, the rationality and effectiveness of the element value are checked, in the scene matching stage, the confidence of the matching result is verified, and in the specification analysis stage, the consistency of the analysis result with the original specification is checked; all these processing is based on the actual operation data and standard specification requirements to ensure the accuracy and practicality of the division result. The system also designs an exception handling mechanism, when the multi-parameter identification model cannot extract enough key operation elements, a backup identification scheme is started, and a rule-based method is used to supplement the key operation elements; when multiple similar scenes appear in the knowledge graph matching, a scene arbitration algorithm is started to select the scene that best matches the current operation state; when the specification file analysis encounters ambiguity, an expert knowledge base is called to assist decision-making, these mechanisms ensure the stable operation of the system under various conditions. The data storage and transmission in the implementation process all adopt standardized formats, real-time operation parameters and environmental state parameters are stored in time series format, key operation elements are transmitted in feature vector form, operation scenes are represented in scene coding, and specification constraint conditions are saved in structured data; such standardized processing is conducive to data exchange and processing efficiency between system modules, and also facilitates subsequent maintenance and expansion, the entire implementation method focuses on the feasibility and reliability of practical engineering applications.

[0058] Example 2: see Figure 3 Low-power operation control of the pole-mounted circuit breaker feeder is carried out based on the divided global operation indicators and local operation indicators. The preprocessed operation data is first split into a global operation data set and a local operation data set, the preprocessing operations include data cleaning, missing value filling and numerical normalization, the data splitting is realized by data routing logic according to the indicator type, the global operation data set contains system-level operation parameters such as total load rate and overall voltage stability, and the local operation data set contains device-level parameters such as branch current value and local temperature reading; during the splitting process, a timestamp alignment mechanism is adopted to ensure that the two data sets remain synchronized in the time dimension, laying a foundation for subsequent collaborative analysis.

[0059] The global running data set and the global running index are input into a first analysis model corresponding to the running scene. The first analysis model is established based on a trend prediction method and uses time series analysis technology to process the input data. The model identifies the trend of the global running data set, calculates the boundary information of the running state through sliding window analysis, and the boundary information includes the load trend line and the voltage stability threshold range. The model compares the identified running state boundary information with the input global running index, evaluates the degree of coincidence between the current system running state and the standard requirement, and determines whether the global running requirement is met. For example, when the identified load trend line exceeds the safety range specified by the index, it is determined that the running requirement is not met. The local running data set is input into a second analysis model corresponding to the running scene. The second analysis model is used to identify subtle fluctuations. The model uses high-frequency signal processing algorithms to analyze the input data, accurately captures and analyzes the high-frequency change characteristics in the local running data set, and the high-frequency change characteristics include instantaneous current spikes and rapid temperature fluctuations. The model detects abnormal patterns in the local running data through multi-scale analysis technology from different time dimensions. The identified subtle fluctuation characteristics are quantified as specific parameter indexes, which are used to evaluate the running state of the equipment level.

[0060] The output results of the first analysis model and the second analysis model need to be fused. A weighted fusion method based on model confidence is used to complete this process. The method assigns corresponding weights according to the historical accuracy of the two analysis models. The weight distribution is calculated based on the performance record of the model in the historical running. The fusion process first standardizes the output results of the two models to make them comparable, then performs weighted calculation according to the weight coefficients, and generates the final low-power control instruction. The control instruction includes adjusting the circuit breaker operation mode and optimizing the energy distribution strategy. The execution results of the entire control process are recorded and saved by the system. The record content includes the generation time, specific content and execution effect of each control instruction. These record data are stored in a special log database to form a complete historical running file. A structured storage format is used in the recording process to facilitate subsequent query and analysis. A data indexing mechanism is established to improve data retrieval efficiency.

[0061] In the implementation process, multiple safeguard mechanisms are designed. Data integrity checks are set in the data splitting stage to ensure that the split data set does not lose key information; abnormality detection is set in the model analysis stage to start the review process when the model output is abnormal; safety checks are set in the instruction generation stage to prevent the generation of control instructions that do not meet safety specifications; these safeguard mechanisms are implemented through automated processes and do not require manual intervention to complete the entire control process. The system also establishes a feedback optimization mechanism, and the recorded control results are used to analyze the model performance, regularly assess the accuracy changes of the first analysis model and the second analysis model, and dynamically adjust the model parameters and weight distribution scheme according to the evaluation results; at the same time, the execution effect of the control instruction is also included in the evaluation system to optimize the entire low-power control strategy, and this closed-loop optimization mechanism enables the system to continuously improve the operation effect. Real-time stream processing technology is used for data processing, and the splitting and analysis operations of global and local running data sets are completed immediately when data flows in to ensure the timeliness of the control instructions; the system uses a distributed computing architecture to distribute data processing tasks to multiple computing nodes for parallel execution, improving overall processing efficiency; all data processing links have monitoring points to track processing status and data quality in real time, ensuring reliable operation of the entire control process. The model fusion link pays special attention to the balance between timeliness and accuracy, and the weighted fusion method not only considers the historical accuracy of the model, but also introduces real-time confidence evaluation to dynamically adjust the weight distribution according to the characteristics of the current input data; the fusion algorithm uses a gradual optimization strategy to gradually improve the fusion accuracy while ensuring real-time, ensuring that the generated control instructions are both timely and accurate. The execution of the control instruction uses a hierarchical processing method, which divides different priorities according to the importance and urgency of the instruction, and high-priority instructions are executed immediately, while low-priority instructions are queued for processing; a state tracking mechanism is provided during instruction execution to monitor the execution effect in real time, and adjustments are made in a timely manner when the execution effect deviates from the expected value, which ensures the timeliness of key controls and avoids system overload. The entire implementation method focuses on the feasibility of practical engineering applications, and all processing steps are based on existing technology platforms and do not require special hardware support; the system design considers various abnormal situation handling schemes, including data abnormalities, model abnormalities, and execution abnormalities, to ensure that basic operational functions are maintained under adverse conditions; the implementation method also provides detailed log recording and state monitoring functions to facilitate system management and troubleshooting by operation and maintenance personnel.

[0062] Embodiment 3: The first step of dynamic verification is to obtain the divided global running dataset, which contains time series data of system-level running parameters, and each data section represents the running state within a specific time window; calculate the fluctuation amplitude and change trend indicators of each data section in the dataset, the fluctuation amplitude is represented by calculating the dispersion of data within the section, and the change trend indicator reflects the direction and strength of data change, the calculation of the two indicators uses the following formula:

[0063]

[0064] Wherein: represents the comprehensive fluctuation indicator, represents the number of sampling points in the data section, is the value of the sampling point, is the average value of the data section, represents the data change rate. If the fluctuation amplitude or change trend indicator of any data section exceeds the preset threshold value, which is determined according to historical running data and system safety requirements, it is determined that the data section has potential abnormalities that need to be finely monitored, and the setting of the threshold value considers the safety margin under different running scenarios.

[0065] According to the determined data section with potential abnormalities, the system extracts the corresponding local running subset from the global running dataset, and the extraction process is based on the timestamp range and parameter type of the abnormal data section; the extraction of the local running subset needs to ensure that it contains complete abnormal feature information, while avoiding the introduction of too much normal data to maintain the pertinence of analysis. The extraction operation uses the sliding window technique, which expands appropriately on both sides of the abnormal section, ensuring that the complete evolution process before and after the abnormality occurs is captured. According to the abnormal features under different running scenarios, the system establishes data filtering rules associated with abnormal features, which are based on historical abnormal cases and expert experience; data filtering rules include feature frequency range, amplitude threshold, and duration, etc. parameters, which are used to further filter out the data segments that need to be focused on from the local running subset. The rule library is stored according to the classification of running scenarios, and each scenario corresponds to a set of filtering rules to ensure the accuracy of rule application.

[0066] Multiple verification mechanisms are set in the dynamic verification process. The calculated fluctuation indicators are cross-verified, and different time scale analysis methods are used to confirm each other. The abnormal judgment results are reviewed, and the authenticity of the anomaly is confirmed by comparing the index changes of adjacent data segments. The quality of the extracted local operation subset is checked to ensure data integrity and accuracy. These verification mechanisms are implemented through an automated process to ensure the reliability of the verification results. The system also designs a dynamic threshold adjustment mechanism. According to the changes in the running state, the judgment threshold is automatically adjusted. When the system is in a stable running state, the threshold value is appropriately increased to avoid false positives. When the system is disturbed, the threshold value is correspondingly reduced to improve the detection sensitivity. The threshold adjustment is based on real-time running state evaluation, and a gradual adjustment strategy is adopted to avoid sudden changes in the threshold value affecting system stability. The positioning of abnormal data segments uses a multi-dimensional analysis method, considering time dimension, space dimension, and parameter dimension. Time dimension analysis of abnormal time regularity, spatial dimension analysis of abnormal distribution characteristics in the system, and parameter dimension analysis of specific operation parameters involved in the anomaly. Through multi-dimensional analysis, the specific location and impact range of the anomaly are accurately located. The generation and application of data screening rules use machine learning technology. Rule generation is based on historical abnormal data training, and common features of abnormal data are extracted through feature engineering. Rule application uses real-time matching method to calculate the similarity between current data features and feature patterns in the rule library. When the similarity exceeds the threshold, the corresponding screening rule is applied. The rule library is updated regularly, incorporating new abnormal cases and experience to maintain the timeliness of the rules. The extraction process of local operation subset focuses on the balance between efficiency and accuracy, and uses an adaptive window size adjustment algorithm to automatically determine the extraction range according to the abnormal characteristics. For fast-changing transient anomalies, a smaller time window is used, and for slowly developing trend anomalies, a larger time window is used. The extraction algorithm also considers data storage and transmission efficiency, minimizing data volume while ensuring analysis accuracy. The intermediate results and final conclusions generated during the dynamic verification process are recorded in detail, including the calculation indicators, judgment results, extracted parameters, and analysis process of each data segment. These records are used for subsequent traceability analysis and system optimization. The records are stored in a structured format, with a complete indexing system to support multi-condition queries and statistical analysis.

[0067] The whole implementation process adopts a pipeline architecture, and each processing link is carried out in turn. The links are connected through data buffer to ensure the continuity of processing. At the same time, parallel processing channels are set up to analyze multiple data segments at the same time, improving processing efficiency. The system resource allocation adopts a dynamic scheduling strategy, which automatically adjusts the computing resources according to the processing load to ensure real-time requirements. The visualization of the abnormal judgment result helps the operator understand the system state. The visualization interface displays the index curve of the data segment, the threshold line mark and the abnormal area highlight, and provides detailed parameter description of the abnormal characteristics. The visualization tool supports multi-dimensional data navigation, allowing the operator to view and analyze abnormal data from different angles. Various boundary conditions are considered in the implementation process, including data missing, noise interference and equipment failure, etc. For each case, a corresponding processing strategy is developed. When data is missing, interpolation algorithm is used to supplement, when noise interference occurs, filtering processing is used, and when equipment failure occurs, standby processing flow is started. These measures ensure the robustness and reliability of the system. The final output of the abnormal information adopts a standardized format, including fields such as abnormal type, occurrence time, impact degree and suggested treatment measures, which can be directly used for subsequent control instruction generation. The abnormal information format is compatible with the control system interface, ensuring the accuracy and efficiency of information transmission.

[0068] Example 4: refer to Figure 4 The cooperative working mechanism between the first analysis model and the second analysis model in the dynamic verification process is started after extracting the local running subset from the global running data set. The real-time data exchange mechanism between the two analysis models is realized by using a message middleware, which is built based on the publish-subscribe mode, allowing asynchronous transmission of data messages between models. The data messages are serialized in ProtocolBuffers format, containing timestamp, data identifier and pre-processing results, etc. The message queue sets up a priority processing channel to ensure that the transmission delay of abnormal data is controlled within milliseconds, and a retransmission mechanism is provided to ensure data integrity. The whole exchange process follows the pre-defined data contract specification.

[0069] When the first analysis model performs preliminary research on the extracted data segment, the model loads the feature template of the current running scene, which contains the matching rules of typical abnormal patterns in this scene. The preliminary research process uses a rule-based reasoning method to calculate the matching degree of the data segment and the feature template, output the abnormal probability score and the preliminary classification result. For the abnormal data segment numbered A-001 in Table 1, the first analysis model detects that the load rate jumps from 78% to 92% in a short time, and the temperature rising rate exceeds the conventional threshold, and preliminarily judges it as "overload risk-high probability". The location, size and preliminary research result of the data segment with potential abnormalities are transmitted to the second analysis model. The location information is represented by hierarchical coding, including feeder segment identification, device number and data point index. The size information contains the number of data points and the time span. The preliminary research result is packaged in the form of a structure, including abnormal type, confidence and feature vector. The transmission process uses an encrypted channel to ensure data security, and adds a digital signature to prevent tampering. The receiving end performs integrity check before triggering subsequent processing.

[0070] After receiving the preliminary research result, the second analysis model performs in-depth analysis on the local running subset. This process uses a multi-modal fusion analysis method, combining time domain analysis, frequency domain analysis and pattern recognition technology. The in-depth analysis first reconstructs the signal of the local running subset, extracts transient features and steady-state features, and then establishes a feature space mapping relationship to identify specific abnormal pattern information. For the A-001 segment in Table 1, the second analysis model identifies that the load fluctuation has periodic characteristics, and the temperature change has a correlation coefficient of 0.85 with the load change, and the abnormal occurrence position is accurate to the in-out line interface of the circuit breaker on the 3rd column. The feedback of abnormal pattern information to the first analysis model uses an incremental learning mechanism. The first analysis model adjusts the weight parameters and matching threshold of the feature template according to the feedback information. The feedback data includes abnormal confirmation flag, mode correction parameter and feature importance score. The first analysis model updates the internal knowledge base using these data to improve the accuracy of subsequent research, for example, according to the learning results of the A-001 case, the first analysis model increases the detection sensitivity of load rate mutation combined with temperature change.

[0071] During multiple data extraction and model analysis processes, the system dynamically adjusts the verification threshold and model configuration parameters according to the difference between each verification result and the actual operating condition. The adjustment process uses a self-adaptive algorithm based on reinforcement learning. The adjustment of the verification threshold takes into account the historical false positive rate and false negative rate indicators. The optimization of the model configuration parameters is based on the principle of minimizing prediction error. All adjustment operations are recorded through version control to record the change history and support the rollback mechanism. The abnormal information, occurrence location, and impact degree obtained through the final analysis are converted into control instructions. The conversion process uses a rule engine to achieve this. The rule base contains various abnormal mode corresponding processing strategies. After the control instructions are generated, they are verified by a security verification logic, including instruction conflict detection and execution consequence evaluation. For the A-001 anomaly in Table 1, the generated control instructions include reducing the load distribution of the third circuit breaker and starting the auxiliary cooling device. The structured log format is used to record the complete analysis process. The log content includes input data fingerprint, processing time node, model version information, intermediate result, and final decision basis. The log system uses a hierarchical storage strategy. Recent data is saved in high-speed storage media, and historical data is compressed and archived. A full-text search index is established to support post-analysis. Refer to Table 1.

[0072] Table 1: Abnormal data section analysis

[0073]

[0074] The entire implementation process adopts a distributed architecture deployment, the first analysis model and the second analysis model run on independent computing nodes, and are interconnected through a high-speed network; the model version management adopts a blue-green deployment strategy, the new version model is switched to the production environment only after being verified by the shadow mode, ensuring the stability of the system operation. The data processing pipeline sets monitoring points, real-time collects performance indicators such as throughput, delay and error rate, and triggers an alarm mechanism when the indicators are abnormal. The visualization display of the abnormal analysis result adopts a multi-view collaborative interface, the main view displays the original curve and feature label of the abnormal data section, the auxiliary view displays the decision path of the model analysis process, and the detail view provides parameterized description of the abnormal mode. The visualization tool supports time axis navigation, allowing the operator to backtrack the analysis process and view the intermediate results of each processing link. The system also establishes a case library management mechanism, classifies and stores typical abnormal analysis cases, including cases A-001 to A-003 recorded in Table 1, and labels them; the case library supports similarity search function, when new abnormal data appears, the system automatically searches for similar historical cases to provide reference for the analysis process. The case library is regularly reviewed and supplemented by experts to maintain the accuracy and integrity of the cases. Data privacy and compliance requirements are considered during implementation, all abnormal data are anonymized before storage, removing personal identification information; data access adopts role-based permission control, operation records are audited throughout the process, meeting the data security management specifications of the power industry. The system interface meets the security communication requirements of international standard IEC62351, ensuring the confidentiality and integrity of data transmission. The finally generated control instruction is issued to the execution equipment through the standard power communication protocol, and the last validity check is performed before the instruction is issued to prevent incorrect operation; the instruction execution state is monitored in real time, and the execution result is fed back to the analysis system to form a closed-loop control. The entire implementation mode embodies the high integration of analysis model and control system, realizing the automatic processing from abnormal detection to control execution.

[0075] Example 5: Determine the data segment with potential abnormal data in the global running data set needs to analyze the numerical range composed of data point indexes, for example, in the running data on August 15, 2023, the system detects that the data segment with timestamp from 14:05 to 14:15 shows that the load rate rises rapidly from 78% to 92%, while the temperature reading rises from 65°C to 72°C, this numerical range composed of 1200 consecutive data point indexes is framed as an abnormal area; the data point indexes are generated based on the millisecond-level timestamp sequence, each index corresponds to a sampling time, the upper and lower bounds of the numerical range are determined by analyzing the starting point and ending point of the abnormal feature, forming the complete abnormal data boundary. According to the determined abnormal data segment, the system calculates the parameters required to extract the local running subset, these parameters include the starting index of the subset, data length and sampling interval; the starting index is determined by locating the first data point timestamp of the abnormal data segment, for example, the starting index of the above abnormal segment corresponds to the timestamp code of 14:05:00.000, the data length is calculated according to the duration of the abnormality, and the sampling interval is determined according to the collection frequency of the original data and the analysis requirements, for example, 1 millisecond interval for high-frequency sampled current data, and 100 millisecond interval for low-frequency collected temperature data; the parameter calculation process uses a sliding window algorithm to ensure that the complete evolution period of the abnormal feature is covered. Using the calculated extraction parameters, the system extracts the corresponding local running subset from the global running data set through data slicing operation, the data slicing is directly operated in memory to avoid frequent disk read and write; the extraction process uses pointer positioning technology to quickly locate the target data segment according to the starting index, then according to the data length parameter to cut the continuous data block, and according to the sampling interval parameter to resample the data, to ensure that the extracted subset meets the analysis requirements; for the abnormal segment in the above example, the extraction operation obtains a local running subset containing 1200 data points, which completely retains the abnormal change characteristics of load rate and temperature.

[0076] The extracted local running subset is checked for completeness of the abnormal data section and non-mixed irrelevant data using multiple verification mechanisms. First, the timestamp continuity of the subset is checked to ensure that no data points are missing or repeated. Then, the subset data range is compared with the numerical range of the original abnormal section to confirm that the key feature values are completely retained. Finally, the subset is verified by an abnormal feature matching algorithm to confirm that it contains the expected abnormal pattern. Any incomplete or mixed data found during the verification process will trigger a re-extraction process until a local running subset that meets the requirements is obtained. The entire extraction process uses a transaction processing mechanism to ensure the atomicity and consistency of data operations. Each extraction operation is recorded in the transaction log, including extraction parameters, timestamps, and operation results. The system sets a timeout limit to prevent operation stagnation due to large data volume or system abnormalities. The timeout limit is dynamically adjusted based on data size to ensure extraction efficiency and avoid premature termination of legitimate operations. The positioning accuracy of abnormal data sections directly affects the extraction quality. The system uses a multi-level index structure to improve positioning accuracy. The first level index is based on time range to quickly locate possible abnormal sections. The second level index is based on feature value screening to determine the exact abnormal boundaries. The third level index is based on pattern matching to verify the authenticity of abnormalities. This multi-level index mechanism effectively reduces mispositioning and missed positioning, ensuring that the extracted local running subset accurately corresponds to the real abnormal area.

[0077] The parameter calculation algorithm takes into account the feature differences of different abnormal types. For transient abnormalities, a shorter data length and higher sampling interval are used to capture rapid changes in detail features. For sustained abnormalities, a longer data length and lower sampling interval are used to analyze overall trends. The algorithm automatically adjusts the parameter calculation strategy based on the abnormal classification results, making the extracted subset most suitable for subsequent deep analysis needs. Data slicing optimizes memory usage efficiency. The lazy loading technique is used to load related data blocks only when needed, reducing memory usage. For large global running data sets, a block processing strategy is used to divide the data set into multiple logical blocks for parallel processing, improving extraction speed. Memory protection mechanisms are implemented during slicing to prevent buffer overflow and data corruption. The verification mechanism introduces digital fingerprint technology to generate a unique hash value for each local running subset. This hash value is calculated based on the subset data content and is used for subsequent integrity verification. The verification process also includes data quality assessment, calculating the signal-to-noise ratio and integrity index of the subset. Only subsets that meet the quality threshold will be sent to the next stage of analysis.

[0078] The system provides a visual monitoring interface for the extraction process, displaying the extraction progress, parameter calculation results and verification status in real time. The running personnel can view the extraction details of each abnormal data section, including data point distribution graph, characteristic value curve and verification report. The monitoring interface also supports manual intervention function, allowing the running personnel to adjust the extraction parameters or re-execute the extraction operation. The management of abnormal data adopts version control method, saving each local running subset of extraction as an independent data version, labeled with extraction time and parameter configuration, facilitating subsequent tracing and comparison. The version management system records the metadata information of each subset, including source data identification, extraction parameters and verification results, supporting fast retrieval and batch processing based on conditions. During implementation, edge case handling schemes are considered. When the abnormal data section spans multiple data files, the cross-file extraction mode is started to ensure data continuity. When data damage or loss is encountered, the data repair program is started to supplement the missing values using interpolation or prediction methods. When system resources are insufficient, the priority scheduling mechanism is started to prioritize the data extraction tasks of high-risk abnormalities. The local running subset after extraction is stored in a standardized format, using the industry-standard data container format, including data array, timestamp sequence and metadata header file. The storage structure design supports fast reading and random access, facilitating efficient processing by subsequent analysis models, while preserving the mapping relationship with the original global data set, supporting reverse tracing and data restoration. The entire implementation process focuses on balancing extraction efficiency and analysis quality, continuously improving extraction performance through parameter optimization and algorithm adjustment, regularly evaluating the accuracy and integrity of extraction results, and improving extraction strategies based on evaluation results. The system maintains quality statistical indicators of extraction operations, including average extraction time, success rate and data integrity, for continuous optimization of implementation effects.

[0079] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another, without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0080] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A terminal low-power operation control method based on pole-mounted circuit breaker feeder, characterized in that, include: Obtain real-time operating parameters and environmental status parameters of the pole-mounted circuit breaker feeder; Based on the real-time operating parameters and environmental status parameters, and combined with the preset feeder operation knowledge graph, the current operating scenario is determined; Based on the current operating scenario and the corresponding operating specification document, the real-time operating parameters are divided into global operating indicators and local operating indicators. Based on the global and local operating indicators, low-power operation control is performed on the pole-mounted circuit breaker feeder. The process of performing low-power operation control on the pole-mounted circuit breaker feeder includes: Based on the global and local operational metrics, the preprocessed operational data is split into a global operational dataset and a local operational dataset. The global runtime dataset and global runtime metrics are input into the first analysis model corresponding to the runtime scenario, and the local runtime dataset is input into the second analysis model corresponding to the runtime scenario. The first analysis model is established based on the trend prediction method. It uses time series analysis technology to process the input data, identifies the trend of the global operation dataset, calculates the boundary information of the operation status through sliding window analysis, compares the identified boundary information of the operation status with the input global operation indicators, evaluates the degree of conformity between the current system operation status and the standard requirements, and determines whether the global operation requirements are met. The local operation dataset is input into the second analysis model of the corresponding operation scenario. The boundary information includes load change trend lines and voltage stability threshold ranges; The second analysis model is used to identify subtle fluctuations. It uses a high-frequency signal processing algorithm to analyze the input data and achieve accurate capture and analysis of high-frequency change features in the local operating data. The model uses multi-scale analysis technology to detect abnormal patterns in the local operating data from different time dimensions. The identified subtle fluctuation features are quantified into specific parameter indicators to evaluate the operating status of the equipment. The high-frequency variation characteristics include instantaneous current spikes and rapid temperature fluctuations; By combining the outputs of the first analysis model and the second analysis model, low-power control instructions are generated, and each control process is recorded.

2. The terminal low-power operation control method based on pole-mounted circuit breaker feeder as described in claim 1, characterized in that, The process of determining the current operating scenario includes: The real-time operating parameters and environmental status parameters are input into the multi-parameter recognition model, which outputs multiple key operating elements. The key operational elements are input into the preset feeder operation knowledge graph, and the current operation scenario containing all key operational elements is determined based on the relationships in the knowledge graph.

3. The terminal low-power operation control method based on pole-mounted circuit breaker feeder as described in claim 1, characterized in that, The process of dividing global and local operational metrics includes: Based on the current operating scenario, call at least one corresponding operating specification file; The intelligent model is used to perform semantic parsing on the operation specification file, and multiple specification constraints are output. Based on the aforementioned standard constraints, the real-time operating parameters are divided into global operating indicators and local operating indicators.

4. The terminal low-power operation control method based on pole-mounted circuit breaker feeder as described in claim 1, characterized in that, When identifying runtime state boundary information, the global runtime dataset undergoes dynamic verification of the divided global runtime dataset. The specific steps of the dynamic verification include: Obtain the global running dataset, calculate the fluctuation amplitude and trend index of each data segment, and determine that any data segment has potential anomalies that need to be monitored in detail if the fluctuation amplitude or trend index of any data segment exceeds the preset threshold. Based on the identified data segments with potential anomalies, extract the corresponding local runtime subsets from the global runtime dataset; Based on the abnormal characteristics under different operating scenarios, establish data filtering rules associated with the abnormal characteristics.

5. The terminal low-power operation control method based on pole-mounted circuit breaker feeder as described in claim 4, characterized in that, The specific steps of the dynamic verification also include: After extracting a local runtime subset from the global runtime dataset, a real-time data exchange mechanism is established between the first analysis model and the second analysis model. The first analysis model is used to make a preliminary judgment on the extracted data segments, and the location, size and preliminary judgment results of the data segments with potential anomalies are transmitted to the second analysis model. After receiving the preliminary judgment results, the second analysis model performs in-depth analysis on the local operating subset to identify specific abnormal pattern information. The abnormal pattern information includes the location of the abnormality, the pattern of change, and its correlation with the global operating indicators.

6. The terminal low-power operation control method based on pole-mounted circuit breaker feeder as described in claim 5, characterized in that, The specific steps of the dynamic verification also include: The identified abnormal pattern information is fed back to the first analysis model; During multiple data extraction and model analysis processes, the verification threshold and model configuration parameters are dynamically adjusted based on the differences between each verification result and the actual operating conditions. The anomaly information, location, and impact level obtained from the final analysis are converted into control commands, and the entire analysis process is recorded.

7. The terminal low-power operation control method based on pole-mounted circuit breaker feeder as described in claim 4, characterized in that, The specific process of extracting the corresponding local runtime subset from the global runtime dataset includes: Identify data segments in the global running dataset that have potential anomalies, wherein the data segments with potential anomalies are defined by a numerical range consisting of multiple data point indices; Based on the identified abnormal data segments, calculate the parameters required to extract the local running subset, including the starting index of the subset, the data length, and the sampling interval; By utilizing the extracted parameters, the corresponding local runtime subsets are extracted from the global runtime dataset through data slicing operations; Verify whether the extracted local running subset completely contains the abnormal data segment and is not mixed with irrelevant data.

8. The terminal low-power operation control method based on pole-mounted circuit breaker feeder as described in claim 1, characterized in that, When fusing the outputs of the first analysis model and the second analysis model, a weighted fusion method based on model confidence is adopted, and corresponding weights are assigned according to the historical accuracy of the first analysis model and the second analysis model.

Images