Multi-level power equipment state perception and collaborative analysis method and system

Abstract

The invention discloses a multi-level power equipment state perception and collaborative analysis method and system, and relates to the technical field of power equipment states, the operation state of multi-level power equipment is marked in a virtual space, a plurality of state components are determined according to a plurality of data combinations corresponding to the operation state and multi-source heterogeneous data, and the state components of the multi-level power equipment are analyzed. And the current equipment state of the multi-level power is sensed in combination with the current working condition of the multi-level power equipment, so that the accuracy of the current equipment state of the multi-level power is improved. Determining a corresponding equipment operation trend based on the identification of the current equipment state of the multi-level power, outputting implicit association features among all levels, and constructing a collaborative analysis system based on the implicit association features, a power grid butted with the multi-level power equipment and the corresponding sub-working modules; and according to the plurality of abnormal data, the workflow relationship of the corresponding sub-working modules and the mapping relationship of the autonomous regulation measures, determining the autonomous regulation measures of the multi-level power equipment.

Description

Technical Field

[0001] This invention relates to the technical field of power equipment status, and in particular to a multi-level power equipment status perception and collaborative analysis method and system. Background Technology

[0002] As the core hub of the power system, the reliability and safety of the operation of multi-level power equipment directly affect the power supply quality and stability of the power grid. With the advancement of smart grid construction, status perception and fault diagnosis technologies for multi-level power equipment (such as hierarchical systems including high-voltage incoming lines, main transformers, and medium-voltage feeders) have become a research hotspot in the industry.

[0003] Most existing state-aware technologies rely on data acquisition from a single source or simple data overlay, lacking a deep correlation mechanism for data from different levels and with different physical attributes. In complex scenarios involving multi-level power equipment, existing technologies often ignore the constraints of power transmission direction on data logic, failing to construct a heterogeneous data network based on energy flow direction. This affects the accuracy of each state component, resulting in low accuracy of the current equipment state in multi-level power systems, and consequently impacting the accuracy of autonomous control measures for multi-level power equipment. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a multi-level power equipment status perception and collaborative analysis method and system.

[0005] This invention provides a multi-level power equipment state perception and collaborative analysis method, including: Multiple data regions are determined based on the detection of multi-level power equipment, and a heterogeneous data network is determined along the power transmission direction of the multiple data regions and the multi-level power equipment. This heterogeneous data network acquires multi-source heterogeneous data of the multi-level power equipment in real time. The physical model of multi-level power equipment is mapped to the corresponding virtual space, and the operating status of multi-level power equipment is marked in the virtual space. Based on the operating status and multiple data combinations corresponding to the multi-source heterogeneous data, multiple state components are determined, and the current equipment status of multi-level power equipment is perceived in combination with the current operating condition of multi-level power equipment. Based on the identification of the current equipment status of multi-level power, the corresponding equipment operation trend is determined, the equipment operation trend is deeply mined, and the implicit correlation characteristics between each level are output. Based on the implicit correlation characteristics, the power grid connected to the multi-level power equipment, and the corresponding sub-working modules, a collaborative analysis system is constructed. In this collaborative analysis system, multiple sub-fault contents are traced back to their source, and the abnormal data of the corresponding sub-working modules are marked. Based on the mapping relationship between multiple abnormal data, the workflow relationship of the corresponding sub-working modules, and the autonomous control measures, the autonomous control measures of multi-level power equipment are determined.

[0006] This invention provides a multi-level power equipment state perception and collaborative analysis system, which is applied to the aforementioned multi-level power equipment state perception and collaborative analysis method; the multi-level power equipment state perception and collaborative analysis system includes: The multi-source heterogeneous data module is used to determine multiple data regions based on the detection of multi-level power equipment, and to determine a heterogeneous data network along the multiple data regions and the power transmission direction of the multi-level power equipment. This heterogeneous data network acquires multi-source heterogeneous data of the multi-level power equipment in real time. The state perception module is used to map the physical model of multi-level power equipment to the corresponding virtual space, mark the operating state of multi-level power equipment in the virtual space, determine multiple state components based on the operating state and multiple data combinations corresponding to multi-source heterogeneous data, and perceive the current equipment state of multi-level power equipment in combination with the current operating condition of multi-level power equipment. The collaborative analysis system module is used to determine the corresponding equipment operation trend based on the identification of the current equipment status of multi-level power, to deeply mine the equipment operation trend, and to output the implicit correlation features between each level. Based on the implicit correlation features, the power grid connected to the multi-level power equipment, and the corresponding sub-working modules, a collaborative analysis system is constructed. The autonomous control measures module is used in this collaborative analysis system to trace the source of multiple sub-faults and mark the abnormal data of the corresponding sub-working modules. Based on the mapping relationship between multiple abnormal data, the workflow relationship of the corresponding sub-working modules and the autonomous control measures, the autonomous control measures of multi-level power equipment are determined.

[0007] Compared with the prior art, the beneficial effects of the present invention are: (1) Based on the detection of multi-level power equipment, multiple data regions are determined, and a heterogeneous data network is determined along the multiple data regions and the power transmission direction of multi-level power equipment. The heterogeneous data network acquires multi-source heterogeneous data of multi-level power equipment in real time. The physical model of multi-level power equipment is mapped to the corresponding virtual space, and the operating status of multi-level power equipment is marked in the virtual space. Multiple state components are determined according to the operating status and the multiple data combinations corresponding to the multi-source heterogeneous data. The current equipment status of multi-level power is perceived in combination with the current operating condition of multi-level power equipment. A heterogeneous data network is introduced, and the multi-source heterogeneous data and the operating status are fully utilized to improve the accuracy of each state component, thereby improving the accuracy of the current equipment status of multi-level power.

[0008] (2) Based on the identification of the current equipment status of multi-level power, the corresponding equipment operation trend is determined, the equipment operation trend is deeply mined, and the implicit correlation features between each level are output. Based on the implicit correlation features, the power grid connected to the multi-level power equipment and the corresponding sub-working modules, a collaborative analysis system is constructed. In this collaborative analysis system, multiple sub-fault contents are traced, and the abnormal data of the corresponding sub-working modules are marked. Based on the mapping relationship of multiple abnormal data, the workflow relationship of the corresponding sub-working modules and the autonomous control measures, the autonomous control measures of multi-level power equipment are determined. The collaborative analysis system is introduced to control the implicit correlation features and to consider the mapping relationship of multiple abnormal data, the workflow relationship of the corresponding sub-working modules and the autonomous control measures as a whole, thereby improving the accuracy of the autonomous control measures of multi-level power equipment. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating the multi-level power equipment status perception and collaborative analysis method in this embodiment of the invention. Figure 2 This is a flowchart illustrating step S11 in the multi-level power equipment status perception and collaborative analysis method in this embodiment of the invention. Figure 3 This is a flowchart illustrating step S12 in the multi-level power equipment status perception and collaborative analysis method in this embodiment of the invention. Figure 4 This is a flowchart illustrating step S13 in the multi-level power equipment status perception and collaborative analysis method in this embodiment of the invention. Figure 5 This is a flowchart illustrating step S14 in the multi-level power equipment status perception and collaborative analysis method in this embodiment of the invention. Figure 6 This is a schematic diagram of the structural composition of the multi-level power equipment state perception and collaborative analysis system in this embodiment of the invention. Detailed Implementation

[0010] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0011] Please see Figures 1 to 6 A multi-level power equipment state perception and collaborative analysis method is proposed and applied to power equipment state scenarios. The multi-level power equipment state perception and collaborative analysis method includes: Step S11: Based on the detection of multi-level power equipment, multiple data regions are determined, and a heterogeneous data network is determined along the multiple data regions and the power transmission direction of multi-level power equipment. This heterogeneous data network acquires multi-source heterogeneous data of multi-level power equipment in real time. Step S12: Map the physical model of the multi-level power equipment to the corresponding virtual space, mark the operating status of the multi-level power equipment in the virtual space, determine multiple state components based on the operating status and multiple data combinations corresponding to the multi-source heterogeneous data, and perceive the current equipment status of the multi-level power equipment in combination with the current operating condition of the multi-level power equipment. Step S13: Based on the identification of the current equipment status of the multi-level power system, determine the corresponding equipment operation trend, perform in-depth mining of the equipment operation trend, and output the implicit correlation features between each level. Based on the implicit correlation features, the power grid connected to the multi-level power equipment, and the corresponding sub-working modules, construct a collaborative analysis system. Step S14: In this collaborative analysis system, the source of multiple sub-faults is traced, and the abnormal data of the corresponding sub-working modules is marked. Based on the mapping relationship between multiple abnormal data, the workflow relationship of the corresponding sub-working modules, and the autonomous control measures, the autonomous control measures of the multi-level power equipment are determined.

[0012] refer to Figure 2 In step S11, the specific steps are as follows: S111: Mark the topology of multi-level power equipment and divide the area according to the key operating indicators of multi-level power equipment to subdivide the multi-level power equipment into multiple data areas. Based on the identification of multiple data areas, determine the corresponding sensing data. The multiple data areas include "core state sensing area", "boundary coupling monitoring area" and "environment association auxiliary area". S112: Based on the power detection of multi-level power equipment, the corresponding power transmission direction is determined. Sensing data of different data formats are initially aligned and iterated sequentially along the power transmission direction to construct the corresponding heterogeneous data network. At the same time, the heterogeneous data network is detected in real time and multiple sub-heterogeneous data are acquired. Multi-source heterogeneous data of multi-level power equipment is constructed based on the multiple sub-heterogeneous data.

[0013] In the embodiments of this application, the topology of multi-level power equipment is marked, and the key operating indicators of multi-level power equipment are combined to divide the multi-level power equipment into multiple data regions. Based on the identification of multiple data regions, the corresponding sensing data is determined. The multiple data regions include "core state sensing region", "boundary coupling monitoring region" and "environment association auxiliary region", which is compatible with the overall consideration of the identification of multiple data regions and ensures the accuracy of the corresponding sensing data.

[0014] At this point, the topology of multi-level power equipment is marked. Topology marking is not just about drawing electrical connection diagrams, but about constructing a coupled topology that includes energy flow, information flow, and material flow (such as cooling medium). Combined with "key performance indicators" (such as power quality distortion rate, thermal stability limit, mechanical wear threshold, etc.), the goal is to screen out the critical paths that have the greatest impact on system stability from the huge topology network. Through topology marking and indicator constraints, the transformation from "blind perception of the whole domain" to "targeted perception of critical paths" is realized, providing mathematical and physical basis for subsequent regional division.

[0015] Based on the importance weights and sensitivity of topology nodes, the monitoring space is deconstructed into three heterogeneous but logically related data regions: Core State Perception Region: Corresponding to the core energy conversion nodes and control center in the equipment topology, this region's data characteristics are high frequency, high precision, and strong real-time performance, directly related to the equipment's "health vital signs"; Boundary Coupling Monitoring Region: Corresponding to the electrical interfaces, bus connection points, and hierarchical interaction interfaces between multi-level devices, this region focuses on monitoring energy interaction and logical coordination between levels, serving as a key window for identifying "cascading faults" and "coupled oscillations"; Environmental Correlation Auxiliary Region: Corresponding to the physical environment boundaries and auxiliary systems of the equipment's operation, the data in this region is mainly used to construct boundary conditions for state assessment, eliminate environmental noise interference, and improve the robustness of the diagnostic model.

[0016] Based on the defined areas, sensing resources are dynamically configured; different areas correspond to different sensor array configurations (such as high-frequency sampling sensors in the core area and temperature, humidity and vibration sensors in the environmental area), and the data transmission protocol and sampling frequency are predefined to form a deterministic mapping relationship of "area-sensor-data stream".

[0017] Specifically, in the scenario of multi-level power equipment status perception and collaborative analysis, key topological nodes such as high-voltage side circuit breakers, main transformer windings, and low-voltage side busbars are first marked; combined with key working indicators, such as setting "transformer top oil temperature threshold" and "feeder current harmonic content" as core constraint indicators; the system identifies the main transformer as the bottleneck node of energy transmission, thereby locking it as a high-priority monitoring object.

[0018] For the core status sensing area, this area is defined within the main transformer body and the high-voltage side intelligent circuit breaker; real-time acquisition of transformer oil chromatography data (DGA), winding hot spot temperature, and traveling wave current data of the high-voltage side current transformer (CT) is conducted. For the boundary coupling monitoring area, this area is defined at the connection bus bridge between the low-voltage side outgoing terminal of the main transformer and each feeder switchgear. This is the physical and logical interaction interface between the "high-voltage level" and the "medium-voltage feeder level" of multi-level power equipment; bus contact temperature, the operating sequence of tie switches, and abrupt changes in power flow are monitored. At the same time, for the environmental correlation auxiliary area, this area covers the enclosed electrical room and heat dissipation and ventilation system where multi-level power equipment is located; ambient temperature and humidity, gas leakage concentration, and the speed and vibration signals of the cooling fans are collected.

[0019] Through the above division, multi-level power equipment no longer outputs massive and disordered logs, but outputs structured datasets with clear physical meanings: a set of high-frequency thermal data from the transformer core, a set of coupled vibration data from the bus bridge, and a set of background parameter data from the environmental chamber. These three types of data complement each other and are logically consistent, providing accurate physical mapping source data for the virtual space mapping in the subsequent step S12.

[0020] Furthermore, based on the power detection of multi-level power equipment, the corresponding power transmission direction is determined. Sensing data of different data formats are initially aligned and iterated sequentially along the power transmission direction to construct a corresponding heterogeneous data network. At the same time, the heterogeneous data network is detected in real time, and multiple sub-heterogeneous data are acquired. Based on the multiple sub-heterogeneous data, multi-source heterogeneous data of multi-level power equipment is constructed, which is compatible with the overall consideration of power detection of multi-level power equipment and ensures the accuracy of the corresponding power transmission direction.

[0021] At this point, the operation of multi-level power equipment strictly follows the law of conservation of energy and Kirchhoff's laws. The flow of electrical energy from the high-voltage side to the low-voltage side has an irreversible temporal direction. This step takes the "direction of power transmission" as the main axis to establish the logical timing of data flow. This breaks the traditional flattened mode of random networking based on IP address or device ID and constructs a directed data topology with a "causal chain". The power flow direction of the physical power grid is directly mapped to the logical flow direction of the information network, which makes the data network naturally have the ability to trace the path of fault propagation, laying a logical foundation for subsequent fault tracing.

[0022] The data output frequencies, encoding formats, and communication protocols of sensors at different levels (such as optical CT, mechanical vibration sensors, and temperature probes) vary greatly. In this step, before the data is entered into the network, timestamps are unified and semantically aligned based on the time scale of power transmission (such as microsecond-level traveling waves, millisecond-level switching actions, and hourly-level temperature rises). Low-frequency data is processed by interpolation to match high-frequency sampling, or high-frequency data is compressed by feature extraction technology to ensure that data from different sources can be effectively correlated on the same time segment.

[0023] Along the direction of power transmission, starting from the source (such as the high-voltage incoming line), the sensing data is gradually incorporated into the downstream nodes. Each iteration is not only the superposition of data, but also the dynamic mounting of network topology nodes. The network edge represents the data flow path, and the node represents the specific data acquisition unit. Through iteration, the network expands from local to global, forming a complete panoramic data map.

[0024] In the constructed network, the data stream of each node is scanned in real time. The so-called "sub-heterogeneous data" is not the original data stream, but effective information fragments (such as harmonic components of specific frequency bands and instantaneous power mutation values) after preliminary cleaning, noise reduction and feature extraction. By extracting sub-heterogeneous data, massive redundant noise is eliminated, and key feature vectors reflecting the status of the equipment are retained, which are finally aggregated into a high-value "multi-source heterogeneous dataset".

[0025] Specifically, in multi-level power equipment, the power transmission direction is clearly defined as: high-voltage incoming line (220kV) → main transformer → medium-voltage feeder (10kV). When constructing the data network, it is not based on the physical location of the network switch, but strictly on this power flow direction. The logical starting point of the data network is set at the PT / CT (voltage / current transformer) at the high-voltage incoming line, and the ending point is set at the outlet of each feeder. This method of establishment ensures that when a fault occurs, the data flow can reflect whether the fault is "upstream conduction" or "downstream feedback". For example, if the harmonic data flow is from the high-voltage incoming line layer to the transformer layer, it is determined to be external interference of the system; otherwise, it is determined to be an internal equipment fault.

[0026] In multi-level power equipment, the characteristics of sensors at different levels are quite different: High-voltage incoming line level: equipped with high-frequency traveling wave sensors (sampling rate in the MHz range, data format is high-speed sampling frames); Main transformer level: equipped with oil chromatography monitoring devices (sampling rate in the hour range, data format is concentration value structure); Medium-voltage feeder level: equipped with intelligent circuit breaker contact temperature sensors (sampling rate in the second range, data format is wireless message).

[0027] The system uses the "fault triggering moment" of the high-voltage traveling wave data as the reference time anchor point, traces back the historical records of transformer oil temperature, and indexes the action sequence of the feeder switch. It interpolates and resamples data with different sampling rates on the time axis, so that within a 10ms time window before and after the fault occurs, the three types of data can be presented in a unified "time-value" sequence, eliminating the barriers of data format and frequency.

[0028] The first iteration (high-voltage incoming line level): The network first incorporates the voltage and current data from the high-voltage side to establish the backbone nodes of the network. At this time, the network model represents the characteristics of the high-voltage source end. The second iteration (main transformer level): The data flow extends along the power transmission direction to the transformer. At this time, the transformer oil temperature and winding deformation data are attached to the backbone network as sub-nodes and logically associated with the high-voltage side data in terms of electrical parameters (such as power balance). The third iteration (medium-voltage feeder level): The network continues to extend downwards, incorporating the load data and switch status of the feeder side into the network. Through these three iterations, a hierarchical and logically clear heterogeneous data network is constructed, which completely replicates the entire process of multi-level power equipment from power reception to power distribution.

[0029] The network detected a transient overvoltage signal on the high-voltage side, and at the same time detected a small jitter signal from the oil flow relay at the transformer level. Instead of transmitting all the original waveforms, the system extracted the "overvoltage amplitude", "duration" and "oil flow jitter frequency" as sub-heterogeneous data.

[0030] These extracted feature values ​​are encapsulated based on their physical correlation (overvoltage caused oil flow disturbance) to form a complete "fault suspected chain" multi-source data packet. This data packet contains both electrical quantities (voltage) and non-electrical quantities (oil flow), and establishes a causal order through the direction of power transmission, providing an accurate input source for the virtual space mapping in the subsequent step S12.

[0031] refer to Figure 3 In step S12, the specific steps are as follows: S121: Collect the physical model of the multi-level power equipment, and mark the geometric attributes, physical attributes and topology of the multi-level power equipment. Map the geometric attributes, physical attributes and topology of the multi-level power equipment to the corresponding virtual space. In the virtual space, perform dynamic semantic marking on the equipment operation status based on the real-time data stream of the multi-level power equipment, and determine the operation status of the multi-level power equipment. S122: Match the operating status and multiple data combinations corresponding to the multi-source heterogeneous data, and construct the corresponding association combination. Based on the identification of each association combination, determine the corresponding state component. Trigger the state control of the multi-level power according to the priority of multiple state components and the current operating condition of the multi-level power equipment, and perceive the current equipment status of the multi-level power.

[0032] In the embodiments of this application, a physical model of a multi-level power equipment is acquired, and the geometric attributes, physical attributes, and topology of the multi-level power equipment are marked. The geometric attributes, physical attributes, and topology of the multi-level power equipment are then mapped to the corresponding virtual space. In the virtual space, the operating status of the equipment is dynamically semantically marked based on the real-time data stream of the multi-level power equipment, and the operating status of the multi-level power equipment is determined. This approach is compatible with the overall consideration of dynamically semantically marking the operating status of the equipment based on the real-time data stream of the multi-level power equipment, ensuring the accuracy of the operating status of the multi-level power equipment.

[0033] At this point, the acquisition of the physical model is not only three-dimensional geometric modeling, but also includes in-depth parameterization of physical properties (such as material density, thermal conductivity, and dielectric constant) and topological structure (such as electrical connection logic and hierarchical membership). Traditional mapping often separates geometry and physical properties, while this method integrates the three into a single mapping, so that the model in the virtual space is not only "similar in appearance" but also has "physical compliance". This means that when a voltage load is applied in the virtual space, the model can generate a response that conforms to the mechanism based on its physical properties.

[0034] Real-time data streams (from the heterogeneous data network built by S112) are injected into the virtual space. The system does not simply display numerical values, but rather uses a preset state semantic dictionary to semantically annotate the current operating segment of the device. For example, when the current data stream exceeds the threshold and the temperature rises, the system automatically marks it with the semantic tag "overload operation" instead of just displaying the current value. This label includes multi-dimensional information such as state type, danger level, and duration. Dynamic semantic labeling transforms the underlying heterogeneous data into high-level state knowledge, greatly reducing the cognitive load of maintenance personnel and making complex operating states clear at a glance.

[0035] By combining virtual space simulation and real-time data feedback, the current operating status of the equipment is determined comprehensively. This is not just about reading sensor readings, but about estimating the status through a virtual model, correcting deviations caused by sensor errors or noise, and outputting a high-confidence operating status conclusion.

[0036] Specifically, for the physical model of multi-level power equipment, geometric attribute mapping is performed as follows: For the main transformer level, a high-precision three-dimensional geometric model including the oil tank, windings, core, bushings, and cooling devices is constructed in the virtual space; not only the external dimensions, but also the distance between winding discs and the diameter of oil pipelines are accurately mapped; physical attribute mapping is performed: the transformer windings are given the physical properties of "copper conductor" (resistivity, thermal conductivity), and the insulating oil is given "fluid thermophysical properties" (viscosity, specific heat capacity); at the high-voltage incoming line level, the GIS (gas-insulated metal-enclosed switchgear) is given the insulation characteristic parameters of SF6 gas; topology mapping is performed: in the virtual space, the logical architecture of "high-voltage incoming line circuit breaker" being connected to "main transformer high-voltage bushing" through electrical topology, and then connected to "medium-voltage feeder switch" is established. This topology mapping ensures that the virtual model is completely consistent with the physical site in terms of logical connection, and the electrical relationship remains accurate even if the physical locations are far apart.

[0037] The multi-source heterogeneous data stream output from step S112 is continuously injected into the virtual space: The marking process of the main transformer level: When the real-time data stream shows: ambient temperature 35℃ (from the environment-related auxiliary area), top oil temperature 85℃ and load rate 95% (from the core status perception area); the transformer object in the virtual model is automatically marked as "heavy load - high temperature warning" semantic state; the system not only renders a warning red on the model surface, but also marks the insulation aging acceleration rate of the transformer in the current state in the background logic, and the semantic label includes the operation prompt "suggest starting auxiliary cooling".

[0038] The marking process of the high-voltage incoming line level: When the real-time data stream shows that the current harmonic distortion rate (THD) suddenly increases to 8% and is accompanied by high-frequency oscillation signals, the virtual model of the high-voltage incoming line is marked as "power quality abnormal - harmonic pollution" state. This semantic label is activated instantly and automatically correlated with whether a large-capacity nonlinear load is connected downstream according to the power transmission direction logic.

[0039] The marking process for medium-voltage feeder levels: When the contact temperature data stream of a feeder switch is stable and the load current is within the rated range, the feeder segment is marked as "healthy operation" and the semantic label is displayed in green, indicating that this level does not currently require attention.

[0040] Based on the combined marking results of the above levels, the overall operating status of the multi-level power equipment in the virtual space is determined to be: "coexistence of local heavy load warning and power quality disturbance"; the system determines that the main transformer level is in a critical state, although it has not yet tripped, but the thermal stability margin is extremely small; the system determines that there is a risk of harmonic source injection on the high-voltage side, and attention should be paid to the cumulative damage effect on the transformer life.

[0041] Through the implementation of S121, the originally scattered and isolated sensor values ​​of multi-level power equipment are transformed into "state semantics" with clear physical meaning and operational guidance value in the virtual space. This provides accurate input basis for state component decoupling and fault diagnosis in the subsequent step S122. This process fully demonstrates the excellent technical concept of applying digital twin technology to multi-level complex power systems.

[0042] Furthermore, the system matches the operating status and multiple data combinations corresponding to the multi-source heterogeneous data, and constructs corresponding association combinations. Based on the identification of each association combination, the corresponding state component is determined. The system triggers the state control of the multi-level power system according to the priority of multiple state components and the current operating condition of the multi-level power equipment, and perceives the current equipment status of the multi-level power system. This system takes into account the overall consideration of the identification of each association combination, ensuring the accuracy of the corresponding state components. At the same time, a heterogeneous data network is introduced, and the system makes full use of multi-source heterogeneous data and the operating status, improving the accuracy of each state component, thereby improving the accuracy of the current equipment status of the multi-level power system.

[0043] At this point, the “operating status” label generated in S121 serves as an index key to retrieve strongly correlated data combinations in the heterogeneous database. This is not a simple data query, but a feature matching process based on Bayesian networks or correlation analysis models. The discrete “data combinations” are bidirectionally anchored to the higher-level “operating status”. Through matching, it is confirmed which data segments (such as harmonics and partial discharge pulses in specific frequency bands) are the core driving factors causing the current state (such as “insulation abnormality”), and invalid background noise data is eliminated.

[0044] The overall operating state of multi-level equipment is often a superposition of multiple concurrent sub-states. By constructing associated combinations, the complex overall state can be deconstructed into several independent "state components" with clear physical meanings. Each associated combination corresponds to a specific physical mechanism. For example, the "current-temperature" combination is mapped to the "thermal stability component", and the "voltage-partial discharge" combination is mapped to the "insulation health component". In this way, the system can identify whether the equipment is "overheating" or "insulation aging", or both, and achieve refined diagnosis of complex operating conditions.

[0045] The identified multiple state components have varying degrees of impact on system safety. Based on a preset risk weight matrix and the current operating condition (such as "heavy load operation" or "maintenance status"), the system dynamically calculates the priority of each component. When the confidence level of a high-priority component (such as "critical insulation defect") exceeds a threshold, the system automatically triggers a state control command. This process embodies the idea of ​​proactive defense, that is, intervening by adjusting the operating mode (such as switching to backup power or reducing load) before a fault evolves into an accident. After the control is triggered, the system rescans the data stream to sense the response of the equipment. This forms a closed loop of "perception-decision-execution-feedback" to ensure that the judgment of the current equipment status is real-time, dynamic, and verified.

[0046] Specifically, step S121 determines that the main transformer level is in a "heavy load-high temperature warning" state, and the high-voltage incoming line level is in a "power quality anomaly" state. Using "heavy load-high temperature warning" as an index, the system accurately extracts data combinations related to thermal effects from the heterogeneous data network: {top oil temperature data, winding hot spot temperature data, cooler fan speed data, load rate data}. Simultaneously, using "power quality anomaly" as an index, another set of data is extracted: {high-voltage side three-phase current waveform data, harmonic spectrum data, zero-sequence current component data}. This process eliminates irrelevant redundant data such as ambient humidity and cabinet door opening / closing status, achieving precise data focus.

[0047] The system performs in-depth analysis on the successfully matched data combinations, decoupling two key state components of the multi-level power equipment: the first state component (thermal stability component): calculated from the correlation combination constructed by {winding temperature, oil temperature, load rate}; the analysis shows that although the load rate is high, the oil temperature rise trend does not match the fan speed; the specific state component of "insufficient heat dissipation efficiency" is identified; the second state component (power quality component): analyzed from the correlation combination constructed by {harmonic spectrum, current waveform}; it is found that the 5th and 7th harmonic contents exceed the standard, and coincide with the rectifier load on the medium voltage feeder side; the state component of "harmonic pollution conduction" is identified.

[0048] The system, considering the current operating conditions of multi-level power equipment (currently during the "summer peak power supply period," with the primary task being to ensure power supply continuity), prioritizes each component: "Insufficient heat dissipation efficiency" (first state component) directly threatens the insulation life of the main transformer, with a risk level of "high"; "Harmonic pollution conduction" (second state component) has not yet affected protection actions, with a risk level of "medium"; therefore, the first state component has the highest priority; the system automatically triggers a control command for the first state component—"Start auxiliary cooling forced operation mode"; the virtual space issues a command to the cooling system control unit of the physical entity, forcibly starting all standby fans and increasing the oil pump flow rate; simultaneously, an "alarm log" is generated for the second state component, prompting maintenance personnel to pay attention to harmonic mitigation.

[0049] After the wind turbine forced start command is executed, the heterogeneous network constructed by S112 transmits new sub-heterogeneous data back in real time; the system senses that the rising trend of winding temperature has been curbed and has begun to slowly decrease; the system confirms that the control measures are effective and updates the status of the main transformer level from "heavy load - high temperature warning" to "heavy load - temperature controlled".

[0050] refer to Figure 4 In step S13, the specific steps are as follows: S131: Dynamically identify the current equipment status of multi-level power systems, output multiple status nodes during the identification process, perform time-domain simulation of multiple status nodes to output the corresponding equipment operation trend, and present the dynamic coupling relationship of multi-level power systems in the equipment operation trend. S132: Based on the tracing of the dynamic coupling relationship, determine the state change content, and perform in-depth mining of the state change content to output the potential dynamic association chain of multi-level devices, and match the corresponding implicit association features. Based on the multi-factor fusion of each implicit association feature, construct the corresponding hierarchical association map. Step S133: Mark the power grid and corresponding sub-working modules connected to the multi-level power equipment, determine the collaborative framework based on the hierarchical association map and the corresponding sub-working modules of the power equipment, determine multiple sub-collaborative combinations based on the power grid connected to the multi-level power equipment, and construct a collaborative analysis system based on the multiple sub-collaborative combinations and the current operating conditions of the multi-level power equipment.

[0051] In the embodiments of this application, the current equipment status of multi-level power is dynamically identified, and multiple status nodes are output during the identification process. The multiple status nodes are pre-simulated according to the time domain to output the corresponding equipment operation trend. The dynamic coupling relationship of multi-level power is presented in the equipment operation trend, thus introducing the dynamic coupling relationship of multi-level power in the equipment operation trend.

[0052] At this point, the system performs high-frequency scanning and feature extraction on the current device status, discretizing the continuous state flow into a series of key "state nodes". The state node is not a simple data snapshot, but a structured data unit containing information such as state attributes, confidence level, and duration. Through dynamic identification, the system can capture the tiny leaps in the device status from "normal" to "abnormal" and identify the key turning point in the early stage of the fault.

[0053] After acquiring a series of state nodes, time-domain pre-simulation is performed in virtual space using time series analysis methods (such as Long Short-Term Memory Network LSTM or state-space model), which is similar to "fast-forwarding" the future operation process in a digital twin. Based on the trajectory of historical state nodes, the possible sequence of future state nodes is deduced to form the "equipment operation trend". This trend not only predicts the change of a single parameter (such as the temperature will continue to rise), but also predicts the evolution direction of the overall system performance (such as sliding from a stable state to an unstable state).

[0054] The operating trends of multi-level power equipment are not the superposition of isolated curves, but the result of interactions between levels. In the trend deduction, the system is deconstructed and presents a "dynamic coupling relationship". It reveals how fluctuations in the upstream level (such as a sudden drop in voltage on the high-voltage side) trigger responses in the downstream level (such as the operation of protection on the medium-voltage side) through electrical connections, and how faults in the downstream level (such as feeder short circuits) impact the upstream level in reverse (such as increased stress in transformer windings). This coupling relationship is dynamically adjusted as the trend evolves, providing a causal chain for subsequent collaborative analysis.

[0055] Specifically, S12 has determined that the main transformer is in a "heavy load - temperature controlled" state, and there is "harmonic pollution" on the high-voltage side; the system continuously monitors the transformer oil temperature and harmonic data at a millisecond frequency; at time T1, it is found that although the oil temperature is stable, the winding vibration has increased slightly; at time T2, it is found that the harmonic distortion rate on the high-voltage side jumps from 5% to 7%.

[0056] The system outputs two key status nodes: Node N1: marked at time T1, with the attribute "slight increase in mechanical vibration", belonging to the main transformer level; Node N2: marked at time T2, with the attribute "harmonic level exceeding the standard", belonging to the high-voltage incoming line level. Simultaneously, the system uses nodes N1 and N2 as inputs, combined with a physical model of multi-level power equipment, to perform a time-domain simulation of the next hour in virtual space: Based on the harmonic exceeding the standard node N2, the model predicts that eddy current losses caused by harmonics will lead to a nonlinear increase in the temperature of the transformer's internal hotspots within the next 30 minutes, even though the cooling system is fully operational; based on the vibration node N1, it predicts that if harmonics continue, the vibration frequency will approach the natural frequency of the transformer body; it outputs a "heat accumulation overheating risk trend" curve, predicting that the winding hotspot temperature will exceed the warning threshold at T+30 minutes; it also outputs a "mechanical resonance risk trend", predicting that the vibration amplitude will reach its peak at T+45 minutes.

[0057] In the trend diagrams above, the system clearly presents the dynamic coupling relationships between the multi-level power equipment: Electrical-thermal coupling relationship: The trend diagrams show that the "harmonic level" (electrical quantity) of the high-voltage incoming line level is strongly positively correlated with the "hot spot temperature" (non-electrical quantity) of the main transformer level; the trend lines indicate that the high-order harmonic current on the high-voltage side generates additional high-frequency eddy current losses in the transformer windings, which is the fundamental driving force for the abnormal rise in the temperature of the main transformer, revealing the "electric-thermal" coupling effect across levels; Source-load coupling relationship: Further trend analysis reveals that although the medium-voltage feeder level is operating normally, the trend shows that if the main transformer trips due to overheating, the medium-voltage feeder level will instantly lose power, resulting in a large-scale power outage; the trend also reveals that there is an "electromagnetic force-mechanical vibration" coupling channel between the current waveform distortion (harmonic) on the high-voltage side and the mechanical vibration of the main transformer; the non-fundamental electromagnetic force generated by the harmonic current directly excites the vibration mode of the transformer body.

[0058] Furthermore, the state change content is determined by tracing the dynamic coupling relationship, and the state change content is deeply mined to output the potential dynamic association chain of multi-level devices. The corresponding implicit association features are matched, and the corresponding hierarchical association map is constructed by fusing multiple factors of each implicit association feature. This approach is compatible with the overall consideration of tracing the dynamic coupling relationship and ensures the accuracy of the state change content.

[0059] At this point, the system traces back along the causal chain of the dynamic coupling relationship to extract the specific state variable change (ΔX) that caused the trend to deteriorate. This not only focuses on the result of the change, but also locks in the "source factor" of the change. By tracing back, the dynamic mapping between the "input stimulus" (such as the current component of a specific frequency) and the "output response" is quantified, and the fuzzy trend is transformed into a calculable "state change content".

[0060] Based on the state change content, the system uses association rule mining methods (such as the Apriori improved method) or causal inference models to find strong correlation patterns across levels and physical quantities in massive historical data and real-time streaming data. These chains are "potential", meaning that they may not be intuitively reflected on the electrical wiring diagram; for example, a weak vibration signal may be strongly correlated with voltage flicker hundreds of meters away; the mined chains reveal the hidden path of fault transmission.

[0061] Implicit correlation features refer to those features that cannot be directly measured but can be calculated through multi-source data fusion (such as "power quality-heat loss coefficient" and "mechanical loosening-load fluctuation correlation"). The system compares the mined dynamic correlation chains with fault feature patterns in the expert knowledge base to identify the implicit features that play a decisive role in the chain. These features are the "information" that characterizes deep-seated defects in the equipment. At the same time, the scattered implicit correlation features are fused in multiple dimensions (time dimension, spatial dimension, and logical dimension) to construct a visualized graph structure. This graph clearly shows how faults penetrate and couple between levels. Nodes represent equipment units or features, and edges represent correlation strength and direction, which provides intuitive topology navigation for subsequent collaborative analysis.

[0062] Specifically, in S131, the system predicts that the main transformer has a "risk of heat accumulation overheating"; the system traces the energy source in reverse along the coupling path of "temperature rise"; it finds that although the main transformer load rate (95%) is the basis, it is insufficient to explain the slope of the temperature rise; the system locks that the current waveform distortion rate of the high-voltage incoming line level changed abruptly at time T, with significant 5th and 7th harmonic components superimposed on the fundamental wave; at the same time, the vibration acceleration of the main transformer tank showed a specific frequency modulation phenomenon at the corresponding time; the state change content is determined to be "abnormal eddy current loss increase" and "electromagnetic force surge" caused by "high-frequency harmonic current injection".

[0063] Based on the aforementioned changes, the system conducts in-depth data mining on multi-level power equipment data: the system uncovers a potential chain spanning three levels—"rectifier load initiation at the medium-voltage feeder level (source) → harmonic penetration at the high-voltage incoming line level (transmission channel) → increased hysteresis and expansion of the main transformer core (response terminal)." This is a hidden chain of "load characteristics - grid quality - equipment health." Traditional monitoring may only see the increase in medium-voltage side load, but does not directly link it to the abnormal vibration of the main transformer. This step reveals the strong spatiotemporal relationship between the two.

[0064] The system extracts and matches features from the above chain, introducing the "harmonic loss temperature rise coefficient," which means that under the current harmonic content, the extra heat energy generated per unit current far exceeds the design value; at the same time, it extracts the "vibration-harmonic coherence coefficient," finding that the vibration frequency and harmonic frequency exhibit extremely high coherence (>0.9); matching these features with the expert database, it identifies the implicit correlation features as "harmonic heating effect" and "electromagnetic forced vibration," indicating that the transformer's overheating and vibration are not due to a fault in the transformer itself, but rather to "pollution" from the external electrical environment.

[0065] The system ultimately constructs a hierarchical relationship graph of multi-level power equipment: the graph includes "medium-voltage rectifier load node", "high-voltage harmonic source node", "transformer winding hot spot node", and "cooling system node"; the edge connecting the "high-voltage harmonic source node" and the "transformer winding hot spot node" is assigned a high weight and labeled as "strong thermal coupling"; the edge connecting the "medium-voltage rectifier load node" and the "high-voltage harmonic source node" is labeled as "pollution conduction"; the graph clearly presents a multi-level, multi-dimensional fault network with "harmonic pollution" as the core, running through medium voltage to high voltage, and simultaneously causing "thermal faults" and "mechanical faults".

[0066] Therefore, the power grids connected to the multi-level power equipment and the corresponding sub-working modules are marked. Based on the hierarchical association map and the corresponding sub-working modules of the power equipment, a collaborative framework is determined. In addition, multiple sub-collaborative combinations are determined in combination with the power grids connected to the multi-level power equipment. Based on the multiple sub-collaborative combinations and the current operating conditions of the multi-level power equipment, a collaborative analysis system is constructed. This system is compatible with the overall consideration of the hierarchical association map and the corresponding sub-working modules of the power equipment, ensuring the accuracy of the collaborative framework.

[0067] At this point, the system marks the grid topology nodes (such as upstream power sources and downstream load centers) to which multi-level power equipment is connected, and at the same time, it refines the internal structure of the equipment into specific sub-working modules (such as cooling control modules, on-load tap changer modules, and protection and control modules). The sub-working modules are defined as controllable independent units, breaking away from the traditional extensive management model that treats equipment as a black box. Each module is given an independent communication address, control interface, and status feedback mechanism, laying the physical foundation for subsequent precise coordination.

[0068] Based on the hierarchical association graph constructed by S132, the system dynamically generates a collaborative framework. Strongly correlated edges in the graph (such as the high coherence of "harmonic-vibration") determine the logical connections in the collaborative framework. If the graph shows that two levels are extremely coupled, the interaction weights of these two levels in the collaborative framework are increased, forming a tightly coupled control loop. The collaborative framework defines the question of "who collaborates with whom". It is not a static fixed structure, but a flexible architecture that is dynamically adjusted according to the shape of the association graph, ensuring that collaborative resources are concentrated on the most active path of fault propagation.

[0069] By combining the operational requirements of the power grid (such as voltage qualification rate and frequency stability), relevant sub-working modules are logically bound to power grid nodes to form "sub-cooperative combinations". Each sub-cooperative combination corresponds to a specific control objective. For example, the "voltage regulation combination" includes on-load tap changers and upstream power grid voltage monitoring points; the "thermal stability control combination" includes cooling system modules and load-side monitoring modules. This division achieves decoupling and divide-and-conquer for complex problems. All sub-cooperative combinations are integrated in a unified spatiotemporal coordinate system, and cooperative strategies are set according to the current operating conditions (such as "heavy load operation" and "maintenance status"). This forms a complete cooperative analysis system with the ability to perceive in real time, make dynamic decisions, and link across levels.

[0070] Specifically, for the power grid, the high-voltage incoming line level is marked to connect to the "upper-level 220kV power grid node" (mainly focusing on short-circuit capacity and voltage support); the medium-voltage feeder level is marked to connect to the "lower-level 10kV distribution network" (mainly focusing on load characteristics and power quality).

[0071] For the sub-working modules, the main transformer level is materialized as "On-Load Tap Changer (OLTC)", "Oil Circulation Cooling (OCP)" and "Neutral Grounding Module"; the high-voltage incoming line level is materialized as "Overvoltage Protection Module" and "Power Quality Monitoring Module"; and the medium-voltage feeder level is materialized as "Intelligent Feeder Switch Module" and "Reactive Power Compensation Module".

[0072] Based on the correlation diagram output by S132 (the diagram shows that there is strong coupling between the medium-voltage side rectifier load and the main transformer vibration, and severe harmonic pollution on the high-voltage side), the system constructs a collaborative framework: traditional frameworks may only focus on "voltage-reactive power" control; however, this method constructs a collaborative framework with "power quality management and thermal stability control" as the core, based on the diagram indication; the framework particularly strengthens the logical connection between the "medium-voltage reactive power compensation module" and the "main transformer cooling module"; the system determines that simply adjusting the cooler cannot fundamentally solve the transformer overheating problem, and it is necessary to coordinate the control of reactive power compensation on the medium-voltage side to filter out the harmonic source, thus forming a cross-level "source-load-heat" collaborative framework; Under the collaborative framework, combined with the grid demand, the system is divided into two key sub-collaborative combinations: Sub-collaborative combination A (harmonic suppression combination): It consists of a "reactive power compensation module at the medium-voltage feeder level" (execution end) and a "power quality monitoring module at the high-voltage incoming line level" (feedback end). By switching the medium-voltage side filter branch, it suppresses the harmonic current detected on the high-voltage side and cuts off the source of transformer overheating.

[0073] Sub-coordination module B (thermal overload protection module): It consists of the "main transformer cooling module" (execution end), the "medium voltage intelligent feeder switch module" (execution end) and the "transformer winding temperature measurement module" (feedback end); when the cooling system is running at full speed but still cannot suppress the temperature rise, it coordinates the control of the feeder switch to reduce the load (load transfer) to ensure the safety of the main equipment.

[0074] Operating Condition Constraints: The multi-level power equipment is currently operating under "summer peak heavy load conditions," requiring extremely high power supply reliability and prohibiting easy load shedding. System Construction Strategy: The system sets the collaborative analysis system to a "supply guarantee priority, governance as a supplement" mode. Priority Ranking: Sub-collaborative combination A is activated first to eliminate harmonics and reduce transformer additional losses through refined reactive power compensation strategies. Backup Strategy: Sub-collaborative combination B is set as the last line of defense. The system will only trigger the load transfer command of the feeder switch when the winding temperature approaches the tripping threshold. The final collaborative analysis system coordinates the monitoring data on the high-voltage side, the compensation devices on the medium-voltage side, and the cooling system of the transformer body in real time, achieving the optimal solution for equipment safety under the premise of ensuring power supply.

[0075] refer to Figure 5 In step S14, the specific steps are as follows: S141: Real-time monitoring of the collaborative analysis system, which determines multiple fault markers as multi-level power equipment operates, determines the corresponding sub-working modules and corresponding sub-fault contents based on the tracing of each fault marker, performs multi-dimensional correlation on multiple sub-fault contents, and outputs the corresponding fault combinations, and determines the abnormal data of the sub-working modules by tracing the source of each fault combination. S142: In each sub-working module, the corresponding workflow relationship is determined based on the detection of the sub-working module. The corresponding state discrete factors are determined according to multiple abnormal data and the workflow relationship of the corresponding sub-working modules. The autonomous control measures of the multi-level power equipment are determined according to the matching of the mapping relationship between each state discrete factor and the autonomous control measures. The autonomous control measures present the autonomous control content of the multi-level power equipment at the time of failure and the response of the corresponding sub-working module.

[0076] In the embodiments of this application, the collaborative analysis system is monitored in real time. This collaborative analysis system determines multiple fault markers as multi-level power equipment operates. Based on the tracing of each fault marker, the corresponding sub-working module and corresponding sub-fault content are determined. The multiple sub-fault contents are correlated in multiple dimensions and the corresponding fault combinations are output. The abnormal data of the sub-working module is determined by tracing the source of each fault combination. This system is compatible with the overall consideration of tracing each fault marker and ensures the accuracy of the corresponding sub-working module and corresponding sub-fault content.

[0077] At this time, the collaborative analysis system is monitored in real time, and the logical nodes (sub-collaborative combinations) within the system are compared with the deviation between the operating status and the expected target in real time. When the deviation exceeds the dynamic threshold, the system automatically generates a "fault marker". Unlike traditional single threshold alarms, these fault markers are generated based on the "break of the collaborative relationship". For example, the alarm is not triggered simply because of "high temperature", but because of the imbalance of the collaborative relationship of "mismatch between cooling capacity and heat loss", which gives the fault marker a deeper semantic information.

[0078] The system traces back along the logical path marked by the fault in the hierarchical association graph. Using the topological connections in the graph, it quickly locates the physical entity causing the anomaly—the specific "sub-working module". Since the collaborative system has predefined the cooperation relationship between modules, the tracing process is the process of verifying the contribution of each module. The system can quickly distinguish between "execution-end fault" (such as module refusal to operate) and "source-end fault" (such as input anomaly), achieving accurate identification.

[0079] Power system faults often exhibit multiple manifestations. This step performs multi-dimensional correlation analysis on the "sub-fault contents" (such as insulation degradation, control circuit disconnection, and harmonic exceedance) corresponding to a single fault marker. The correlation dimensions include temporal correlation (sequential order), spatial correlation (different levels and locations), and logical correlation (causal chain). The scattered sub-fault contents are combined into "fault combinations" with logical chains, revealing the complex characteristics of the fault (e.g., "harmonic exceedance" + "accelerated insulation aging" combine to form a "harmonic-induced heating fault chain"). Based on the fault combinations, the system directly penetrates to the underlying data stream and extracts the original "abnormal data" that caused the fault. This completes the closed loop from "qualitative diagnosis" to "quantitative evidence collection," providing accurate data support for subsequent autonomous control.

[0080] Specifically, the "thermal overload defense combination" constructed by S133 is running; the system compares the "calculated value of the main transformer hot spot temperature" with the "upper limit of the cooling system's heat dissipation capacity" in real time; the system finds that although the cooler is running at full speed (fan speed at full load), the rate of decrease of the transformer winding hot spot temperature is much lower than the model prediction value; the collaborative system determines that the "thermal balance collaborative relationship has failed" and automatically generates the fault mark FM_01 (insufficient thermal control effectiveness). This is not just a high temperature alarm, but a high-level mark indicating "control system failure".

[0081] Based on the hierarchical association graph, the system traces back the marker FM_01: the system traces back along the "thermal balance failure" path, checking the oil flow feedback of the "cooling module" and the input data of the "harmonic source module"; the trace reveals that the oil pump operation status feedback of the cooling system is normal, but the data of the "power quality monitoring module" at the high-voltage incoming line level shows that the input side harmonic current is continuously abnormal; at the same time, the "oil flow indicator module" at the main transformer level reports that the oil flow has periodic fluctuations; finally, the problem is located in the "submersible oil pump and its drive unit (sub-working module)" at the main transformer level, and the "filter control interface (sub-working module)" at the high-voltage incoming line level.

[0082] The system performs in-depth correlation analysis on the identified module faults: First sub-fault: Abnormal vibration of the submersible pump and large pulsation of oil flow pressure (mechanical fault characteristic); Second sub-fault: The high-voltage side harmonic current exceeds the standard for a long time, causing local overheating of the transformer tank wall (electrical fault characteristic); The system analysis found that the vibration frequency generated by the second sub-fault (harmonic flux) is superimposed with the frequency of sub-fault A (submersible pump mechanical vibration), causing resonance, resulting in oil flow fluctuations and reduced heat dissipation efficiency; The system outputs fault combination FC_01: {electrical harmonic excitation + mechanical resonance + heat dissipation efficiency attenuation}, which is a composite fault combination across physical fields (electromagnetic-mechanical-fluid).

[0083] Based on fault combination FC_01, the system accurately extracts abnormal data as a chain of evidence: Abnormal data 1: "Current anomaly waveform data" and "Bearing vibration acceleration spectrum data" of the main transformer level submersible oil pump (center frequency 150Hz, amplitude exceeds the standard); Abnormal data 2: "5th and 7th harmonic current effective value time series" of the high voltage incoming line side (continuously exceeds the national standard limit); Abnormal data 3: "Temperature rise gradient data" of the transformer top oil temperature (abnormally steep).

[0084] Furthermore, within each sub-module, the corresponding workflow relationship is determined based on the detection of that sub-module. The corresponding state discrete factors are determined based on multiple abnormal data and the workflow relationships of the corresponding sub-modules. The autonomous control measures for multi-level power equipment are determined based on the matching of the mapping relationships between each state discrete factor and the autonomous control measures. These autonomous control measures present the autonomous control content of the multi-level power equipment at the time of failure and the responses of the corresponding sub-modules. They incorporate a holistic consideration of matching the mapping relationships between each state discrete factor and the autonomous control measures, ensuring the accuracy of the autonomous control measures for multi-level power equipment. Simultaneously, a collaborative analysis system is introduced to control implicit correlation characteristics and comprehensively considers the mapping relationships between multiple abnormal data, the workflow relationships of the corresponding sub-modules, and the autonomous control measures, thereby improving the accuracy of the autonomous control measures for multi-level power equipment.

[0085] At this point, the system performs in-depth testing on each sub-module, analyzing its inputs, processing, outputs, and logical interactions with other modules to construct a "workflow relationship" model. This is not just a physical connection diagram, but a dynamic process that includes control logic, causal timing, and constraints. Static modules are transformed into dynamic "workflow nodes." By analyzing the workflow, the system can clearly understand how an anomaly in a module will propagate to the next level through the logical chain, providing a logical basis for formulating control strategies to block the propagation of faults.

[0086] The "abnormal data" traced in S141 is mapped to specific workflow nodes, and the degree of deviation of the abnormal data from the normal workflow is quantified. This deviation is defined as a "state discrete factor". The state discrete factor reflects the "disorder" or "entropy increase" of the system's operating state. For example, if the cooling module's workflow requires "oil flow to increase linearly with temperature", but the actual data shows that "oil flow is stagnant", then a high-weight discrete factor is generated. By calculating these factors, the system clarifies "where the problem is" and "how much deviation there is".

[0087] Based on a pre-built expert knowledge base or reinforcement learning model, the system seeks the optimal mapping relationship between "state discrete factors" and "autonomous control measures." This is a multi-objective optimization process aimed at eliminating discrete factors with minimal control costs (such as the fewest switching actions and the least load loss) to bring the system back to steady state. The system not only matches single measures but also generates combined strategies. For example, it matches "forced cooling" for "thermal discrete factors" and "harmonic suppression" for "electrical discrete factors," forming a set of combined measures. The generated autonomous control measures are encapsulated into specific control instruction sets, and the sub-module responses triggered by the instructions are clearly presented. This forms a clear "fault moment response script," ensuring the interpretability and safety of the control actions.

[0088] Specifically, for the faulty modules located in S141, the system analyzes their workflows: Submersible pump module (main transformer level): its standard workflow relationship is: [detect winding temperature] → [temperature criterion logic] → [output frequency conversion speed control command] → [motor drive oil flow]; Filter control interface (high voltage incoming line level): its workflow relationship is: [detect bus harmonic content] → [over-limit judgment] → [trigger filter branch switching]; Feeder load control module (medium voltage feeder level): its workflow relationship is: [receive dispatch command / protection action] → [maintain power balance] → [switch opening and closing].

[0089] The system maps the abnormal data to the above workflow and identifies two key factors contributing to the state dispersion: First discrete factor (mechanical-fluid domain): In the submersible pump module, the detected "vibration anomaly data" and "oil flow pressure pulsation data" indicate that the mechanical efficiency of the motor drive is far lower than the preset standard value of the workflow; the system determines that there is a "surge in mechanical transmission loss discrete factor", which means that the existing cooling capacity can no longer meet the heat dissipation requirements, which is a hardware performance degradation; The second discrete factor (electrical domain): At the high-voltage incoming line level, the detected "harmonic exceedance data" is compared with the "unswitched state" of the filter interface; the workflow requires switching when harmonic exceedance occurs, but no action is taken in practice (possibly due to excessively high threshold settings or communication delays); the system determines that there is a "harmonic mitigation lag discrete factor", which results in the heat source not being cut off.

[0090] Based on the aforementioned discrete factors, the system performs mapping and matching in the decision base: For the first discrete factor (cooling capacity degradation): due to the decline in hardware performance (discrete), simply adjusting the frequency converter is ineffective; the matching strategy is to activate the "forced activation strategy of the standby cooling group" and trigger the "load transfer warning"; For the second discrete factor (harmonic mitigation lag): due to the lag in source mitigation (discrete), the matching strategy is to forcibly trigger the "forced activation command of the medium-voltage side filter branch" to cut off the heat source; the system integrates the two factors and generates the final autonomous control measures: "start the main transformer standby cooler group + forcibly activate the medium-voltage side 5th and 7th filter branches".

[0091] The autonomous control measures were issued and implemented, and the specific response content was presented: the system automatically generated control messages, bypassed the conventional threshold logic, and directly sent a closing command to the standby cooler control cabinet at the main transformer level; at the same time, it sent a forced activation command to the intelligent capacitor cabinet at the medium-voltage feeder level; the sub-working module response was as follows: the main transformer level response was that the standby oil pump started, the oil flow rate increased by 150% within 30 seconds, the rising trend of the winding hot spot temperature was curbed and began to fall; the medium-voltage feeder level response was that after the filter branch was activated, the high-voltage side monitoring module detected that the harmonic distortion rate dropped rapidly from 7.5% to below 2%, and the eddy current loss inside the transformer was significantly reduced.

[0092] Please see Figure 6 , Figure 6 This is a schematic diagram of the structure of the multi-level power equipment state perception and collaborative analysis system in this embodiment of the invention; the multi-level power equipment state perception and collaborative analysis system is applied to the above-mentioned multi-level power equipment state perception and collaborative analysis method; the multi-level power equipment state perception and collaborative analysis system includes: The multi-source heterogeneous data module 21 is used to determine multiple data regions based on the detection of multi-level power equipment, and to determine a heterogeneous data network along the multiple data regions and the power transmission direction of the multi-level power equipment. The heterogeneous data network acquires multi-source heterogeneous data of the multi-level power equipment in real time. The state perception module 22 is used to map the physical model of the multi-level power equipment to the corresponding virtual space, mark the operating state of the multi-level power equipment in the virtual space, determine multiple state components based on the operating state and multiple data combinations corresponding to the multi-source heterogeneous data, and perceive the current equipment state of the multi-level power equipment in combination with the current operating condition of the multi-level power equipment. The collaborative analysis system module 23 is used to determine the corresponding equipment operation trend based on the identification of the current equipment status of multi-level power, to deeply mine the equipment operation trend, and to output the implicit correlation features between each level. Based on the implicit correlation features, the power grid connected to the multi-level power equipment, and the corresponding sub-working modules, a collaborative analysis system is constructed. The autonomous control measures module 24 is used in this collaborative analysis system to trace the source of multiple sub-fault contents, mark the abnormal data of the corresponding sub-working modules, and determine the autonomous control measures of multi-level power equipment based on the workflow relationship of multiple abnormal data, the corresponding sub-working modules, and the mapping relationship of autonomous control measures.

[0093] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for multi-level power equipment state perception and collaborative analysis, characterized in that, include: Multiple data regions are determined based on the detection of multi-level power equipment, and a heterogeneous data network is determined along the power transmission direction of the multiple data regions and the multi-level power equipment. This heterogeneous data network acquires multi-source heterogeneous data of the multi-level power equipment in real time. The physical model of multi-level power equipment is mapped to the corresponding virtual space, and the operating status of multi-level power equipment is marked in the virtual space. Based on the operating status and multiple data combinations corresponding to the multi-source heterogeneous data, multiple state components are determined, and the current equipment status of multi-level power equipment is perceived in combination with the current operating condition of multi-level power equipment. Based on the identification of the current equipment status of multi-level power, the corresponding equipment operation trend is determined, the equipment operation trend is deeply mined, and the implicit correlation characteristics between each level are output. Based on the implicit correlation characteristics, the power grid connected to the multi-level power equipment, and the corresponding sub-working modules, a collaborative analysis system is constructed. In this collaborative analysis system, multiple sub-fault contents are traced back to their source, and the abnormal data of the corresponding sub-working modules are marked. Based on the mapping relationship between multiple abnormal data, the workflow relationship of the corresponding sub-working modules, and the autonomous control measures, the autonomous control measures of multi-level power equipment are determined.

2. The multi-level power equipment state perception and collaborative analysis method according to claim 1, characterized in that, The method involves determining multiple data regions based on the detection of multi-level power equipment, and establishing a heterogeneous data network along these data regions and the power transmission direction of the multi-level power equipment. This heterogeneous data network acquires multi-source heterogeneous data from the multi-level power equipment in real time, including: The topology of multi-level power equipment is marked, and the key operating indicators of multi-level power equipment are combined to divide the multi-level power equipment into multiple data regions. Based on the identification of multiple data regions, the corresponding sensing data is determined. The multiple data regions include "core status sensing region", "boundary coupling monitoring region" and "environment association auxiliary region".

3. The multi-level power equipment state perception and collaborative analysis method according to claim 2, characterized in that, The method of determining multiple data regions based on the detection of multi-level power equipment, and determining a heterogeneous data network along the multiple data regions and the power transmission direction of the multi-level power equipment, wherein the heterogeneous data network acquires multi-source heterogeneous data from the multi-level power equipment in real time, further includes: Based on the power detection of multi-level power equipment, the corresponding power transmission direction is determined. Sensing data of different data formats are initially aligned and iterated sequentially along the power transmission direction to construct a corresponding heterogeneous data network. At the same time, the heterogeneous data network is detected in real time and multiple sub-heterogeneous data are acquired. Based on the multiple sub-heterogeneous data, multi-source heterogeneous data of multi-level power equipment is constructed.

4. The multi-level power equipment status perception and collaborative analysis method according to claim 1, characterized in that, The process involves mapping the physical model of the multi-level power equipment to a corresponding virtual space, marking the operating status of the multi-level power equipment in the virtual space, determining multiple state components based on the operating status and multiple data combinations corresponding to the multi-source heterogeneous data, and sensing the current equipment status of the multi-level power equipment in conjunction with the current operating condition of the multi-level power equipment, including: Collect physical models of multi-level power equipment, and label the geometric, physical, and topological properties of the multi-level power equipment. Map the geometric, physical, and topological properties of the multi-level power equipment to the corresponding virtual space. In the virtual space, dynamically semantically label the operating status of the equipment based on the real-time data stream of the multi-level power equipment, and determine the operating status of the multi-level power equipment.

5. The multi-level power equipment state perception and collaborative analysis method according to claim 4, characterized in that, The process of mapping the physical model of multi-level power equipment to a corresponding virtual space, marking the operating status of the multi-level power equipment in the virtual space, determining multiple state components based on the operating status and multiple data combinations corresponding to the multi-source heterogeneous data, and sensing the current equipment status of the multi-level power equipment in conjunction with the current operating condition of the multi-level power equipment, further includes: The system matches the operating status and multiple data combinations corresponding to the multi-source heterogeneous data, and constructs corresponding association combinations. Based on the identification of each association combination, the corresponding state components are determined. The system triggers the state control of the multi-level power system according to the priority of multiple state components and the current operating conditions of the multi-level power equipment, and senses the current equipment status of the multi-level power system.

6. The multi-level power equipment state perception and collaborative analysis method according to claim 1, characterized in that, The system identifies the current equipment status based on multi-level power systems to determine the corresponding equipment operation trends, performs in-depth analysis on these trends, and outputs implicit correlation features between each level. Based on these implicit correlation features, the power grid connected to the multi-level power equipment, and the corresponding sub-working modules, a collaborative analysis system is constructed, including: The current equipment status of multi-level power is dynamically identified, and multiple status nodes are output during the identification process. These multiple status nodes are then pre-simulated in the time domain to output the corresponding equipment operation trend, which reveals the dynamic coupling relationship of multi-level power.

7. The multi-level power equipment state perception and collaborative analysis method according to claim 6, characterized in that, The method involves identifying the current equipment status based on multi-level power systems to determine the corresponding equipment operation trends, performing in-depth analysis on these trends, and outputting implicit correlation features between each level. Based on these implicit correlation features, the power grid connected to the multi-level power equipment, and the corresponding sub-working modules, a collaborative analysis system is constructed. This also includes: Based on the tracing of this dynamic coupling relationship, the content of state change is determined, and the content of state change is deeply mined to output the potential dynamic association chain of multi-level devices, and the corresponding implicit association features are matched. Based on the multi-factor fusion of each implicit association feature, the corresponding hierarchical association map is constructed. The power grids and corresponding sub-working modules connected to the multi-level power equipment are marked. A collaborative framework is determined based on the hierarchical association map and the corresponding sub-working modules of the power equipment. Multiple sub-collaborative combinations are determined in combination with the power grids connected to the multi-level power equipment. A collaborative analysis system is constructed based on the multiple sub-collaborative combinations and the current operating conditions of the multi-level power equipment.

8. The multi-level power equipment state perception and collaborative analysis method according to claim 1, characterized in that, In this collaborative analysis system, multiple sub-fault contents are traced back to their source, and the abnormal data of the corresponding sub-working modules are marked. Based on the mapping relationship between multiple abnormal data, the workflow relationship of the corresponding sub-working modules, and the autonomous control measures, the autonomous control measures of multi-level power equipment are determined, including: The collaborative analysis system is monitored in real time. As the multi-level power equipment operates, it identifies multiple fault markers. Based on the tracing of each fault marker, it determines the corresponding sub-working module and the corresponding sub-fault content. It performs multi-dimensional correlation on multiple sub-fault contents and outputs the corresponding fault combinations. By tracing the source of each fault combination, it determines the abnormal data of the sub-working module.

9. The multi-level power equipment state perception and collaborative analysis method according to claim 8, characterized in that, In this collaborative analysis system, multiple sub-fault contents are traced back to their source, and abnormal data of corresponding sub-working modules are marked. Based on the mapping relationship between multiple abnormal data, the workflow of corresponding sub-working modules, and autonomous control measures, autonomous control measures for multi-level power equipment are determined. The system also includes: In each sub-module, the corresponding workflow relationship is determined based on the detection of that sub-module. The corresponding state discrete factors are determined based on multiple abnormal data and the workflow relationship of the corresponding sub-module. The autonomous control measures of the multi-level power equipment are determined based on the matching of the mapping relationship between each state discrete factor and the autonomous control measures. These autonomous control measures present the autonomous control content of the multi-level power equipment at the time of failure and the response of the corresponding sub-module.

10. A multi-level power equipment state perception and collaborative analysis system, characterized in that, The multi-level power equipment state perception and collaborative analysis system is applied to the multi-level power equipment state perception and collaborative analysis method as described in any one of claims 1-9; The multi-level power equipment status perception and collaborative analysis system includes: The multi-source heterogeneous data module is used to determine multiple data regions based on the detection of multi-level power equipment, and to determine a heterogeneous data network along the multiple data regions and the power transmission direction of the multi-level power equipment. This heterogeneous data network acquires multi-source heterogeneous data of the multi-level power equipment in real time. The state perception module is used to map the physical model of multi-level power equipment to the corresponding virtual space, mark the operating state of multi-level power equipment in the virtual space, determine multiple state components based on the operating state and multiple data combinations corresponding to multi-source heterogeneous data, and perceive the current equipment state of multi-level power equipment in combination with the current operating condition of multi-level power equipment. The collaborative analysis system module is used to determine the corresponding equipment operation trend based on the identification of the current equipment status of multi-level power, to deeply mine the equipment operation trend, and to output the implicit correlation features between each level. Based on the implicit correlation features, the power grid connected to the multi-level power equipment, and the corresponding sub-working modules, a collaborative analysis system is constructed. The autonomous control measures module is used in this collaborative analysis system to trace the source of multiple sub-faults and mark the abnormal data of the corresponding sub-working modules. Based on the mapping relationship between multiple abnormal data, the workflow relationship of the corresponding sub-working modules and the autonomous control measures, the autonomous control measures of multi-level power equipment are determined.

Images