Method and system for monitoring state of power supply equipment driven by multi-source sensing data

Abstract

The invention discloses a multi-source sensing data driven power supply equipment state monitoring method and system, and relates to the related technical field of power supply equipment monitoring, and the method comprises the steps: recognizing a current first operation mode of power supply equipment, which comprises any one of a constant current mode, a constant voltage mode and a constant power mode; analyzing a circuit topological structure where the power supply equipment is located to obtain a first characteristic path; collecting a multi-source operation sensing data set and carrying out anomaly identification; and performing abnormal consistency detection, determining an abnormal consistency range, and outputting the abnormal consistency range as a state evaluation result of the power supply equipment. The technical problems of poor state monitoring accuracy and real-time performance, inaccurate fault element positioning and high false alarm rate of the power supply equipment caused by lack of identification and distinguishing of operation modes and incapability of adapting to multi-mode dynamic operation characteristics of the power supply equipment in the prior art are solved. The technical effects of early warning and accurate evaluation of state monitoring of the power supply equipment, reduction of false alarm rate and accurate positioning of fault elements are achieved.

Description

Technical Field

[0001] This application relates to the technical field of power equipment monitoring, specifically to a power equipment status monitoring method and system driven by multi-source sensor data. Background Technology

[0002] Power supply equipment such as switching power supplies, uninterruptible power supplies (UPS), rectifiers, and inverters are core units for energy conversion and supply. Their operational reliability directly affects the stability and safety of the entire system. Power supply equipment has increasingly complex internal structures, variable operating conditions, and dynamically changing operating modes. Once a fault occurs, it can lead to equipment downtime and production interruptions, or even huge economic losses and safety accidents. Traditional power supply equipment status monitoring and alarms are lagging, unable to provide early warnings or determine degradation trends. Furthermore, they struggle to comprehensively reflect the overall operating status of multi-component coupled power supply equipment, and cannot accurately locate potential faulty components. For example, abnormal output ripple may originate from the failure of filter capacitors, or from faults in the control loop or power switching devices; it is difficult to distinguish based solely on output voltage values. In addition, power supply equipment typically operates in different modes such as constant current, constant voltage, or constant power. The energy flow paths, stresses on various components, and their normal characteristic distribution in sensor data all differ significantly. Ignoring these differences in operating modes leads to a high misjudgment rate in power supply equipment health status assessments.

[0003] Therefore, current technologies suffer from several technical problems, including a lack of identification and differentiation of operating modes, an inability to adapt to the multi-mode dynamic operating characteristics of power supply equipment, resulting in poor accuracy and real-time performance of power supply equipment status monitoring, inaccurate location of faulty components, and a high false alarm rate. Summary of the Invention

[0004] This application provides a power equipment status monitoring method and system driven by multi-source sensor data, which solves the technical problems in the prior art that lack the ability to identify and distinguish operating modes, cannot adapt to the multi-mode dynamic operating characteristics of power equipment, resulting in poor accuracy and real-time performance of power equipment status monitoring, inaccurate location of faulty components, and high false alarm rate. It achieves the technical effects of realizing early warning and accurate assessment of power equipment status monitoring, reducing false alarm rate, and accurately locating faulty components.

[0005] This application provides a power supply device status monitoring method driven by multi-source sensor data. The method includes: identifying a first operating mode of the power supply device, the first operating mode including any one of constant current mode, constant voltage mode, and constant power mode; analyzing the circuit topology of the power supply device based on the first operating mode to obtain a first feature path corresponding to the first operating mode, wherein the first feature path includes multiple path nodes, each path node corresponding to a component in the same circuit as the power supply device; collecting multi-source operating sensor datasets corresponding to the multiple path nodes on the first feature path; performing anomaly identification on the multi-source operating sensor datasets; extracting multiple anomalous operating sensor datasets corresponding to the multiple path nodes respectively; performing anomaly consistency detection on the multiple anomalous operating sensor datasets; determining the anomaly consistency range; and outputting the anomaly consistency range as the status assessment result of the power supply device.

[0006] In a possible implementation, the power device status monitoring method driven by multi-source sensor data further performs the following processing: labeling the components corresponding to each path node in the circuit topology with prior sensitivity labels; constructing a candidate path set with the power device as the source node and other components in the circuit topology as sink nodes; obtaining a matching sensitivity label set corresponding to the first operating mode; calculating a sensitivity score for each candidate path in the candidate path set based on the matching sensitivity label set and the labeled prior sensitivity labels, outputting k candidate paths that are greater than a preset sensitivity threshold, and determining a first feature path from the k candidate paths.

[0007] In a possible implementation, the power device status monitoring method driven by multi-source sensor data further performs the following processing: constructing matching sensitivity tag sets for constant current mode, constant voltage mode, and constant power mode respectively; wherein, the matching sensitivity tag set for constant current mode includes resistance sensitivity and thermal sensitivity, the matching sensitivity tag set for constant voltage mode includes inductance sensitivity, capacitance sensitivity, and filter sensitivity, and the matching sensitivity tag set for constant power mode includes power sensitivity and coupling sensitivity.

[0008] In a possible implementation, the power device status monitoring method driven by multi-source sensor data further performs the following processing: setting a pre-switching time window and a post-switching time window for the power device to perform mode switching; for each path node of each candidate path in the k candidate paths, extracting the historical isomorphic features of the pre-switching time window and the post-switching time window; performing power coherence scoring on each candidate path based on the historical isomorphic features of each path node; and determining a first feature path from the k candidate paths based on the output k power coherence scoring results.

[0009] In a possible implementation, the power equipment status monitoring method driven by multi-source sensor data further performs the following processing: wherein the types of historical isomorphic features include steady-state features, transition features, and frequency domain features; based on the historical isomorphic features of each path node, the historical isomorphic feature response variance before and after each candidate path switching is calculated, the historical isomorphic feature response variance is converted into a power coherence score, and k power coherence score results are output.

[0010] In a possible implementation, the power device status monitoring method driven by multi-source sensor data further performs the following processing: after performing time-series alignment processing on the multi-source operation sensor dataset, an aligned multi-source operation sensor dataset is obtained; based on the aligned multi-source operation sensor dataset, real-time isomorphic features of each path node on the first feature path are obtained; a healthy operation sensor data identification template is constructed, and the real-time isomorphic features of each path node are anomaly identified using the healthy operation sensor data identification template, and multiple abnormal operation sensor datasets corresponding to the multiple path nodes are output.

[0011] In a possible implementation, the power device status monitoring method driven by multi-source sensor data further performs the following processing: performing anomaly consistency detection on the multiple abnormal operation sensor datasets, including anomaly temporal overlap, anomaly ordered consistency, directional consistency, and spectral fingerprint consistency; calculating weights for the anomaly temporal overlap, anomaly ordered consistency, directional consistency, and spectral fingerprint consistency, and outputting a sequence of continuous path nodes connected to the power device; and determining the anomaly consistency range based on the number of continuous path node sequences.

[0012] This application also provides a power supply device status monitoring system driven by multi-source sensor data. The system includes: a first operating mode identification unit, used to identify the current first operating mode of the power supply device, the first operating mode including any one of constant current mode, constant voltage mode, and constant power mode; a circuit topology analysis unit, used to analyze the circuit topology of the power supply device based on the first operating mode, and obtain a first feature path corresponding to the first operating mode, wherein the first feature path includes multiple path nodes, each path node corresponding to a component in the same circuit as the power supply device; an anomaly identification unit, used to collect multi-source operating sensor datasets corresponding to the multiple path nodes on the first feature path, perform anomaly identification on the multi-source operating sensor datasets, and extract multiple anomalous operating sensor datasets corresponding to the multiple path nodes respectively; and a consistency detection unit, used to perform anomaly consistency detection on the multiple anomalous operating sensor datasets, determine the anomaly consistency range, and output the anomaly consistency range as the status assessment result of the power supply device.

[0013] This application proposes a multi-source sensor data-driven power supply equipment condition monitoring method and system to identify the current first operating mode of the power supply equipment, including any one of constant current mode, constant voltage mode, and constant power mode; analyze the circuit topology of the power supply equipment to obtain the first characteristic path; collect multi-source operating sensor datasets and perform anomaly identification; perform anomaly consistency detection, determine the anomaly consistency range, and output the result as the condition assessment of the power supply equipment. This solves the technical problems in existing technologies, such as the lack of identification and differentiation of operating modes, inability to adapt to the multi-mode dynamic operating characteristics of power supply equipment, resulting in poor accuracy and real-time performance of power supply equipment condition monitoring, inaccurate location of faulty components, and high false alarm rate. It achieves the technical effects of enabling early warning and accurate assessment of power supply equipment condition monitoring, reducing false alarm rate, and accurately locating faulty components. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0015] Figure 1 This is a schematic flowchart of a power device status monitoring method driven by multi-source sensor data provided in an embodiment of this application.

[0016] Figure 2 This is a schematic diagram of the structure of a power equipment condition monitoring system driven by multi-source sensor data provided in an embodiment of this application.

[0017] Explanation of reference numerals in the attached figures: First operating mode identification unit 10, circuit topology analysis unit 20, anomaly identification unit 30, and consistency detection unit 40. Detailed Implementation

[0018] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application.

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0021] This application provides a power device status monitoring method driven by multi-source sensor data, such as... Figure 1 As shown, the method includes: Step S100: Identify the current first operating mode of the power supply device, the first operating mode including any one of constant current mode, constant voltage mode and constant power mode.

[0022] Preferably, the operating mode refers to the automatic control strategy adopted by the power supply to meet the load requirements, which determines the core objective of the power supply to adjust the output. Under different operating modes, the components in the circuit that bear the greatest pressure and are most prone to failure are completely different. The current first operating mode of the power supply is identified. The first operating mode includes any one of constant current mode, constant voltage mode, and constant power mode. Specifically, constant current mode means that no matter how the load changes, the output current remains stable at the set value. The power supply monitors the output current in real time. If the current shows a downward trend, the output voltage is increased to keep the current constant. If the current shows an upward trend, the output voltage is decreased to limit the current. Constant voltage mode means that no matter how the load changes, the output voltage remains stable at the set value. The power supply monitors the output voltage in real time through an internal feedback loop (such as using a PID controller). If the output voltage is detected to decrease due to increased load, the output current is increased to raise the voltage. Conversely, if the load becomes lighter and the voltage increases, the output current is decreased. Constant power mode means that no matter how the load changes, the output power (voltage × current) remains stable at the set value. If the output voltage increases, the control system correspondingly reduces the output current, and vice versa, to ensure that the power remains constant.

[0023] Step S200: Analyze the circuit topology of the power supply device based on the first operating mode to obtain the first feature path corresponding to the first operating mode. The first feature path includes multiple path nodes, and each path node corresponds to a component in the same circuit as the power supply device.

[0024] Step S200 further includes step S210, which involves labeling the components corresponding to each path node in the circuit topology with prior sensitivity tags; step S220, which involves constructing a candidate path set using the power supply device as the source node and other components in the circuit topology as sink nodes; step S230, which involves obtaining the matching sensitivity tag set corresponding to the first operating mode; and step S240, which involves calculating the sensitivity score for each candidate path in the candidate path set based on the matching sensitivity tag set and the labeled prior sensitivity tags, outputting k candidate paths that are greater than a preset sensitivity threshold, and determining the first feature path from the k candidate paths.

[0025] Preferably, the circuit topology of the power supply device is analyzed based on the first operating mode. Specifically, the components corresponding to each path node in the circuit topology are labeled with prior sensitivity tags. That is, each component in the circuit topology is labeled with a prior sensitivity tag. Circuit components may include resistors, capacitors, inductors, transistors, etc., indicating their most sensitive and easily affected physical effects. For example, high-power resistors or MOSFET switches generate a lot of heat when they are working, so they are labeled with thermal sensitivity tags; the capacitance of filter electrolytic capacitors is easy to decay and the equivalent series resistance is easy to increase, so they are labeled with capacitance sensitivity and ESR sensitivity tags; the resistance accuracy of precision sampling resistors is critical, so they are labeled with resistance sensitivity tags; power inductors or transformers are labeled with inductance sensitivity, saturation sensitivity and thermal sensitivity tags.

[0026] Preferably, using the power supply device as the source node and other components in the circuit topology as sink nodes means that current can flow from the power supply device to many different branches, eventually reaching various different components. The entire circuit topology is automatically analyzed, listing all possible current paths, thus obtaining multiple candidate paths. Each path includes the connection sequence of all components through which the current flows from the source node to the sink node, and these paths are combined to form a candidate path set. Then, the matching sensitivity label set corresponding to the first operating mode is obtained, i.e., the operating mode is mapped to the sensitivity labels. In constant current mode, the task is to maintain current stability, mainly focusing on components that impede current and generate heat; the matching labels include resistance sensitivity and thermal sensitivity. In constant voltage mode, the task is to maintain voltage stability, mainly focusing on components that store energy and filter; the matching labels include capacitance sensitivity, inductance sensitivity, and filtering sensitivity. In constant power mode, the task is to maintain power stability, mainly focusing on energy conversion efficiency and overall performance; the matching labels include power sensitivity and coupling sensitivity.

[0027] Preferably, based on the matching sensitivity label set and the labeled prior sensitivity labels, a sensitivity score is calculated for each candidate path in the candidate path set. That is, the degree of matching between the pre-labeled prior sensitivity labels of the components contained in the path and the matching labels of the current operating mode is evaluated. For example, by calculating the weight of the matching labels of all components on the path and quantifying the degree of matching, the path with a higher score indicates that it is more critical and more likely to expose fault symptoms in the current mode. Then, a preset sensitivity threshold is configured according to historical data, and k candidate paths with a value greater than the preset sensitivity threshold are output, where K is a positive integer greater than 1. Finally, based on historical data analysis, the path with the most violent and coherent reaction during mode switching is selected as the most important first feature path. The first feature path includes multiple path nodes, and each path node corresponds to a component in the same circuit as the power supply device, that is, it includes multiple physical components as data acquisition points. For example, in a buck switching power supply, the path nodes may include power switching transistors, freewheeling diodes, power inductors, output filter capacitors, current sampling resistors, etc., thereby ensuring that anomaly identification is more efficient and accurate.

[0028] Furthermore, step S230 also includes constructing matching sensitivity tag sets for constant current mode, constant voltage mode, and constant power mode respectively; wherein, the matching sensitivity tag set for constant current mode includes resistance sensitivity and thermal sensitivity, the matching sensitivity tag set for constant voltage mode includes inductance sensitivity, capacitance sensitivity, and filtering sensitivity, and the matching sensitivity tag set for constant power mode includes power sensitivity and coupling sensitivity.

[0029] Preferably, based on different operating modes of the power supply equipment, the most important component attribute tags are predefined for each mode, i.e., a set of matching sensitivity tags. Among them, the core of constant current mode is to stabilize the current. The current flowing through resistive components will generate heat. Its matching sensitivity tag set includes resistance sensitivity and thermal sensitivity to ensure that components that are prone to heat generation or resistance drift due to current fluctuations are closely monitored, such as sampling resistors and power MOSFETs. The core of constant voltage mode is to stabilize the voltage. The stability and quality of the voltage are highly dependent on energy storage and filtering components. Its tag set includes inductance sensitivity, capacitance sensitivity, and filtering sensitivity to check whether the inductor is saturated, whether the capacitor's capacitance value has decayed, or whether the ESR has increased. The core of constant power mode is efficiency and overall energy transfer. Its tag set includes power sensitivity and coupling sensitivity to assess whether the power device losses are abnormal and whether the coupling efficiency of magnetic components has decreased.

[0030] Furthermore, step S240 also includes step S241, setting a pre-switching time window and a post-switching time window for the power supply device to perform mode switching; step S242, for each path node of each candidate path in the k candidate paths, extracting the historical isomorphic features of the pre-switching time window and the post-switching time window; step S243, performing power coherence scoring on each candidate path based on the historical isomorphic features of each path node, and determining a first feature path from the k candidate paths based on the output k power coherence scoring results.

[0031] Preferably, a pre-switching time window and a post-switching time window are set for the power supply equipment to switch from one operating mode to another to capture dynamic moments. The pre-switching time window records the stable state of the system during the short period before the switch, and the post-switching time window records the transitional state and the new stable state during the short period after the switch. For example, the pre-switching time window is defined as 1-3 operating cycles before the switch, and the post-switching time window is defined as the transition to the steady state after the switch. For each path node (component) on each candidate path, historical isomorphic features of the pre-switching and post-switching time windows are extracted. These are key features characterizing the dynamic behavior of the switching mode. They typically include steady-state features such as the average voltage and current before and after the switch to measure the magnitude of change; transitional features such as rise time, overshoot, and settling time to measure the speed or smoothness of change; and frequency domain features such as changes in harmonic components at specific frequencies to analyze more subtle disturbances. Next, based on the historical isomorphic characteristics of each path node, a power coherence score is performed on each candidate path. This means determining which candidate path's overall behavior is most coherent with power mode switching. Specifically, the more drastic and synchronous the characteristic changes of multiple components on a candidate path are during operating mode switching, the higher the score of that path. A coherence score is then calculated for each candidate path, resulting in k power coherence score results. Finally, based on the k power coherence score results, the candidate path with the highest score is selected from the k candidate paths and determined as the first characteristic path.

[0032] Furthermore, step S243 also includes step S2431, wherein the types of the historical isomorphic features include steady-state features, transition features, and frequency domain features; step S2432, based on the historical isomorphic features of each path node, calculate the historical isomorphic feature response variance before and after each candidate path switching, perform power coherence score conversion on the historical isomorphic feature response variance, and output k power coherence score results.

[0033] Preferably, the types of historical isomorphic features include steady-state features, transitional features, and frequency domain features. Steady-state features may include the average voltage / current before and after the switch, reflecting the magnitude of change; transitional features may include rise time and overshoot, reflecting the rate of change and stability; frequency domain features may include specific harmonic amplitudes, reflecting hidden oscillation modes. Then, based on the historical isomorphic features of each path node, the response variance of the historical isomorphic features of each candidate path before and after the switch is calculated, i.e., the historical isomorphic feature response variance, which measures the fluctuation intensity of the overall state of the path. The larger the historical isomorphic feature response variance, the more significantly the path is affected by mode switching. Finally, the historical isomorphic feature response variance is transformed into a power coherence score using a scoring function, such as normalization or weighted summation. The higher the power coherence score, the more sensitive and coherent the path is to changes in power operating state, thus identifying it as the first characteristic path.

[0034] Step S300: Collect multi-source running sensor datasets corresponding to the multiple path nodes on the first feature path, perform anomaly identification on the multi-source running sensor datasets, and extract multiple abnormal running sensor datasets corresponding to the multiple path nodes respectively.

[0035] Preferably, multiple sensors are deployed at each path node along the first characteristic path to synchronously collect different types of operational sensor data from the path nodes, generating a multi-source operational sensor dataset, which may include voltage, current, temperature, etc. Then, anomaly identification is performed on the multi-source operational sensor dataset, that is, from the multi-dimensional data of each node, abnormal data that deviates significantly from the normal range is identified and extracted. This is usually done by comparing the real-time collected multi-source operational sensor dataset with a pre-established health template under normal operating conditions. The health template defines the fluctuation range, statistical characteristics, or waveform shape of various operational data under healthy conditions. For example, if the voltage ripple amplitude on a capacitor node suddenly exceeds the upper limit of the health template, and its temperature reading also rises abnormally, the data from the voltage and temperature sensors during this period are marked as abnormal. Finally, all the marked abnormal data are classified according to the path nodes to form multiple abnormal operational sensor datasets corresponding to multiple path nodes, thereby ensuring the efficiency and accuracy of detection and diagnosis.

[0036] Furthermore, step S300 also includes step S310, performing time-series alignment processing on the multi-source operation sensor dataset to obtain an aligned multi-source operation sensor dataset; step S320, obtaining real-time isomorphic features of each path node on the first feature path based on the aligned multi-source operation sensor dataset; step S330, constructing a healthy operation sensor data identification template, using the healthy operation sensor data identification template to perform anomaly identification on the real-time isomorphic features of each path node, and outputting multiple abnormal operation sensor datasets corresponding to the multiple path nodes respectively.

[0037] Preferably, different types of sensors have differences in acquisition start time, sampling frequency, and transmission delay. The multi-source operating sensor dataset may be misaligned on the time axis. Therefore, the multi-source operating sensor dataset is time-series aligned. That is, the data of the multi-source operating sensor dataset is synchronized to a unified timestamp through interpolation, resampling, etc., to form an aligned multi-source operating sensor dataset. This ensures that the power equipment monitoring status is analyzed at the same time and avoids misjudgment. Then, from the aligned multi-source operating sensor dataset, the real-time isomorphic features of each path node on the first feature path are obtained to characterize the component status of each path node. For example, the effective value of voltage ripple, total harmonic distortion (THD) of current waveform, instantaneous slope of temperature data, etc. are calculated to intuitively represent the health degradation trend of the power equipment. Under the health status of power equipment, the normal fluctuation range and statistical distribution of various isomorphic features are established through extensive learning and integrated into a health operation sensor data identification template. Then, the health operation sensor data identification template is used to identify anomalies in the real-time isomorphic features of each path node. That is, the extracted real-time isomorphic features are compared with the health benchmark in the health operation sensor data identification template. Features that deviate significantly from the normal mode are marked as abnormal. Finally, multiple abnormal operation sensor datasets corresponding to multiple path nodes are output, which contain multi-dimensional abnormal events on each path node.

[0038] Step S400: Perform anomaly consistency detection on the multiple abnormal operation sensor datasets, determine the anomaly consistency range, and output the anomaly consistency range as the state evaluation result of the power supply device.

[0039] Step S400 further includes step S410, performing anomaly consistency detection on the multiple abnormal operation sensor datasets, including anomaly temporal overlap, anomaly ordered consistency, directional consistency, and spectral fingerprint consistency; step S420, calculating weights for the anomaly temporal overlap, anomaly ordered consistency, directional consistency, and spectral fingerprint consistency, and outputting a sequence of continuous path nodes connected to the power supply device; step S430, determining the anomaly consistency range based on the number of continuous path node sequences.

[0040] Preferably, the true root cause of the fault is determined by cross-validating multiple evidence chains on multiple abnormal operation sensor datasets. Specifically, anomaly consistency detection is performed on multiple abnormal operation sensor datasets, including anomaly temporal overlap, anomaly order consistency, directional consistency, and spectral fingerprint consistency detection. Among these, anomaly temporal overlap detection checks whether anomalies at different path nodes occur within the same time period. If a temperature anomaly at one path node and a current anomaly at another path node completely overlap in time, the probability of a correlation is much higher than that of random occurrence. Anomaly order consistency detection checks whether there is a chronological order in which the abnormal events occur. In the current path, the impact of the fault often propagates along the path. For example, a slight breakdown of an upstream switching transistor may occur first. (Abnormality) leads to waveform distortion, followed by abnormal heating of the downstream inductor, and finally, failure of the capacitor at the end due to overvoltage; Directional consistency detection refers to checking whether the changing trends of multiple parameters conform to physical logic. For example, if the current reading of a certain node increases abnormally, its temperature reading also increases abnormally. The two changes are coordinated, which conforms to the heating principle of Ohm's law. Conversely, if the current increases while the temperature decreases, it may indicate a measurement error or an abnormality that is not related; Spectral fingerprint consistency detection refers to checking whether abnormal energy peaks appear simultaneously at the same frequency point in the spectrum diagrams of voltage and vibration signals of different nodes. For example, if the amplitude of a certain characteristic harmonic increases, spectral fingerprint consistency often directly points to a common excitation source, such as a specific switching frequency or mechanical resonance point, which is a strong feature for locating faults.

[0041] Preferably, the abnormal time overlap, abnormal ordered consistency, directional consistency, and spectral fingerprint consistency are then assigned corresponding weights. For example, spectral fingerprint consistency is the least likely to be disturbed, so its weight may be the highest. The weights of abnormal time overlap, abnormal ordered consistency, directional consistency, and spectral fingerprint consistency are 20%, 20%, 25%, and 35%, respectively. Then, a weighted calculation is performed based on the weights to output the continuous path node sequence connected to the power supply device, that is, to determine the longest continuous path node sequence that satisfies time overlap, directional consistency, and time delay monotonicity. Then, the abnormal consistency range is determined based on the number of continuous path node sequences, that is, the number of continuous path node sequences is counted and used as the state assessment result of the power supply device. For example, only 1-2 adjacent nodes (small range) indicate early, local performance degradation, while long sequence nodes (large range) may indicate a serious or spreading fault, and the scope of the abnormal fault is clarified.

[0042] In the above text, refer to Figure 1 A multi-source sensor data-driven power device status monitoring method according to embodiments of the present invention is described in detail. Next, reference will be made to... Figure 2 A power device condition monitoring system driven by multi-source sensor data is described according to an embodiment of the present invention.

[0043] The multi-source sensor data-driven power equipment status monitoring system according to embodiments of the present invention addresses the technical problems in the prior art, such as the lack of identification and differentiation of operating modes, inability to adapt to the multi-mode dynamic operating characteristics of power equipment, resulting in poor accuracy and real-time performance of power equipment status monitoring, inaccurate location of faulty components, and high false alarm rate. It achieves the technical effects of enabling early warning and accurate assessment of power equipment status monitoring, reducing false alarm rate, and accurately locating faulty components. Figure 2 As shown, the power equipment status monitoring system driven by multi-source sensor data includes: a first operating mode identification unit 10, a circuit topology analysis unit 20, an anomaly identification unit 30, and a consistency detection unit 40.

[0044] The first operating mode identification unit 10 is used to identify the current first operating mode of the power supply device, which includes any one of constant current mode, constant voltage mode, and constant power mode; the circuit topology analysis unit 20 is used to analyze the circuit topology of the power supply device based on the first operating mode, and obtain the first feature path corresponding to the first operating mode, wherein the first feature path includes multiple path nodes, and each path node corresponds to a component in the same circuit as the power supply device; the anomaly identification unit 30 is used to collect multi-source operating sensor datasets corresponding to the multiple path nodes on the first feature path, perform anomaly identification on the multi-source operating sensor datasets, and extract multiple abnormal operating sensor datasets corresponding to the multiple path nodes respectively; the consistency detection unit 40 is used to perform anomaly consistency detection on the multiple abnormal operating sensor datasets, determine the anomaly consistency range, and output the anomaly consistency range as the state evaluation result of the power supply device.

[0045] The specific configuration of the circuit topology analysis unit 20 will be described in detail below. The circuit topology analysis unit 20 further includes: labeling the components corresponding to each path node in the circuit topology with prior sensitivity tags; constructing a candidate path set using the power supply device as the source node and other components in the circuit topology as sink nodes; obtaining a matching sensitivity tag set corresponding to the first operating mode; calculating a sensitivity score for each candidate path in the candidate path set based on the matching sensitivity tag set and the labeled prior sensitivity tags, outputting k candidate paths that are greater than a preset sensitivity threshold, and determining a first feature path from the k candidate paths.

[0046] The specific configuration of the circuit topology analysis unit 20 will be described in detail below. The circuit topology analysis unit 20 further includes: constructing matching sensitivity label sets for constant current mode, constant voltage mode, and constant power mode, respectively; wherein, the matching sensitivity label set for constant current mode includes resistance sensitivity and thermal sensitivity, the matching sensitivity label set for constant voltage mode includes inductance sensitivity, capacitance sensitivity, and filter sensitivity, and the matching sensitivity label set for constant power mode includes power sensitivity and coupling sensitivity.

[0047] The specific configuration of the circuit topology analysis unit 20 will be described in detail below. The circuit topology analysis unit 20 further includes: setting a pre-switching time window and a post-switching time window for the power supply device to perform mode switching; extracting historical isomorphic features of the pre-switching time window and the post-switching time window for each path node of each of the k candidate paths; performing power coherence scoring on each candidate path based on the historical isomorphic features of each path node; and determining a first characteristic path from the k candidate paths based on the output k power coherence scoring results.

[0048] The specific configuration of the circuit topology analysis unit 20 will be described in detail below. The circuit topology analysis unit 20 further includes: wherein the types of historical isomorphic features include steady-state features, transition features, and frequency domain features; based on the historical isomorphic features of each path node, the historical isomorphic feature response variance before and after each candidate path switching is calculated; the historical isomorphic feature response variance is converted to a power coherence score, and k power coherence score results are output.

[0049] The specific configuration of the anomaly identification unit 30 will be described in detail below. The anomaly identification unit 30 further includes: performing time-series alignment processing on the multi-source operational sensor dataset to obtain an aligned multi-source operational sensor dataset; obtaining real-time isomorphic features of each path node on the first feature path based on the aligned multi-source operational sensor dataset; constructing a healthy operational sensor data identification template; using the healthy operational sensor data identification template to perform anomaly identification on the real-time isomorphic features of each path node; and outputting multiple abnormal operational sensor datasets corresponding to the multiple path nodes respectively.

[0050] The specific configuration of the consistency detection unit 40 will be described in detail below. The consistency detection unit 40 further includes: setting tiered compensation nodes according to the response cycle of the liquid cooling response scheme; verifying the control status of the data center at the tiered compensation nodes and establishing verification deviations; and performing node adaptive control compensation management at each tiered compensation node based on the corresponding verification deviation.

[0051] The power equipment status monitoring system driven by multi-source sensor data provided in this embodiment of the invention can execute the power equipment status monitoring method driven by multi-source sensor data provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0052] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0053] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A multi-source sensor data driven power supply equipment condition monitoring method, characterized in that, The method comprises: identifying a first operating mode of the power supply device, the first operating mode comprising any one of a constant current mode, a constant voltage mode and a constant power mode; analyzing a circuit topology in which the power supply device is located based on the first operating mode, to obtain a first feature path corresponding to the first operating mode, wherein the first feature path comprises a plurality of path nodes, each path node corresponding to an element in the same circuit as the power supply device; collecting a plurality of source operating sensing data sets corresponding to the plurality of path nodes on the first feature path, performing anomaly identification on the plurality of source operating sensing data sets, and extracting a plurality of abnormal operating sensing data sets corresponding to the plurality of path nodes respectively; performing anomaly consistency detection on the plurality of abnormal operating sensing data sets, determining an anomaly consistency range, and outputting the anomaly consistency range as a state evaluation result of the power supply device.

2. The multi-source sensor data driven power plant equipment condition monitoring method of claim 1, wherein, Based on the first operating mode, the circuit topology in which the power supply device is located is analyzed to obtain a first feature path corresponding to the first operating mode, and the method comprises: annotating the elements corresponding to each path node in the circuit topology with prior sensitivity labels; taking the power supply device as a source node and other elements in the circuit topology as sink nodes to construct a candidate path set; obtaining a matching sensitivity label set corresponding to the first operating mode; based on the matching sensitivity label set and the annotated prior sensitivity labels, performing sensitivity score calculation on each candidate path in the candidate path set, outputting k candidate paths greater than a preset sensitivity threshold, and determining a first feature path from the k candidate paths.

3. The multi-source sensor data driven power plant equipment condition monitoring method of claim 2, wherein, The method for obtaining a matching sensitivity label set corresponding to the first operating mode comprises: respectively constructing matching sensitivity label sets for constant current mode, constant voltage mode and constant power mode; wherein the matching sensitivity label set for the constant current mode comprises resistance sensitivity and thermal sensitivity, the matching sensitivity label set for the constant voltage mode comprises inductance sensitivity, capacitance sensitivity and filtering sensitivity, and the matching sensitivity label set for the constant power mode comprises power sensitivity and coupling sensitivity.

4. The multi-source sensor data driven power plant equipment condition monitoring method of claim 2, wherein, The method for determining a first feature path from the k candidate paths comprises: setting a pre-switching time window and a post-switching time window for mode switching of the power supply device; for each path node of each candidate path in the k candidate paths, extracting historical isomorphic features of the pre-switching time window and the post-switching time window; performing power coherence scoring on each candidate path according to the historical isomorphic features of each path node, and determining a first feature path from the k candidate paths according to the output k power coherence scoring results.

5. The multi-source sensor data driven power plant equipment condition monitoring method of claim 4, wherein, The method for performing power coherence scoring on each candidate path according to the historical isomorphic features of each path node and outputting k power coherence scoring results comprises: wherein the types of historical isomorphic features include steady-state features, transition features and frequency domain features; ​ According to the historical isomorphic characteristics of each path node, the historical isomorphic characteristic response variance before and after each candidate path switching is calculated, the historical isomorphic characteristic response variance is converted into a power coherence score, and k power coherence score results are output.

6. The multi-source sensor data driven power equipment condition monitoring method of claim 1, wherein, The method comprises: After time sequence alignment processing is performed on the multi-source operation sensing data set, an aligned multi-source operation sensing data set is obtained. According to the aligned multi-source operation sensing data set, real-time isomorphic characteristics of each path node on the first feature path are obtained. A healthy operation sensing data identification template is constructed, and the real-time isomorphic characteristics of each path node are subjected to abnormality identification using the healthy operation sensing data identification template, thereby outputting a plurality of abnormal operation sensing data sets corresponding to the plurality of path nodes respectively.

7. The multi-source sensor data driven power equipment condition monitoring method of claim 1, wherein, The abnormality consistency detection on the plurality of abnormal operation sensing data sets comprises abnormal time overlap, abnormal order consistency, direction consistency, and spectral fingerprint consistency. The abnormal time overlap, abnormal order consistency, direction consistency, and spectral fingerprint consistency are subjected to weight calculation, and a continuous path node sequence connected to the power supply equipment is output. The number of continuous path node sequences is used to determine an abnormal consistency range.

8. A multi-source sensor data driven power plant condition monitoring system, characterized by, The system is used to implement the multi-source sensing data driven power supply equipment state monitoring method according to any one of claims 1 to 7, and the system comprises: A first operation mode identification unit is configured to identify a first operation mode of the power supply equipment, wherein the first operation mode comprises any one of a constant current mode, a constant voltage mode, and a constant power mode. A circuit topology analysis unit is configured to analyze a circuit topology in which the power supply equipment is located based on the first operation mode, and obtain a first feature path corresponding to the first operation mode, wherein the first feature path comprises a plurality of path nodes, and each path node corresponds to an element in the same circuit as the power supply equipment. An abnormality identification unit is configured to collect a plurality of multi-source operation sensing data sets corresponding to the plurality of path nodes on the first feature path, identify abnormalities in the multi-source operation sensing data sets, and extract a plurality of abnormal operation sensing data sets corresponding to the plurality of path nodes respectively. A consistency detection unit is configured to detect abnormality consistency of the plurality of abnormal operation sensing data sets, determine an abnormality consistency range, and output the abnormality consistency range as a state evaluation result of the power supply equipment.

Images