A method and system for alarming the operation of a power system

Abstract

This invention discloses a method and system for power system operation alarm, belonging to the field of power system monitoring technology. It includes deploying a distributed intelligent sensing network to collect data hierarchically and extract basic physical features; performing semantic annotation and hierarchical credibility assessment on the basic physical features to generate a credible physical feature set; constructing a semantic causal graph and optimizing edge weights through perturbation simulation to extract dynamic risk factors; constructing and comparing hierarchical state vectors and dynamic health threshold vectors to generate operation alarm states; and performing attribution analysis based on the operation alarm states to generate threshold calibration signals for feedback. This invention employs a technical solution combining multi-source data fusion, hierarchical credibility assessment, causal risk modeling, and dynamic threshold adaptive calibration, enabling accurate identification and proactive early warning of potential risks in power systems, and improving the accuracy and adaptability of alarm decisions.

Description

Technical Field

[0001] This invention relates to the field of power system monitoring technology, and in particular to a power system operation alarm method and system. Background Technology

[0002] The power system is the lifeline of national energy security and a critical infrastructure upon which modern society operates. Real-time monitoring and alarming of its operational status are core elements in ensuring the safe, stable, and economical operation of the power grid. The operational alarm system aims to promptly detect potential faults, anomalies, or unstable states by continuously measuring and analyzing the operating parameters of various equipment in the power grid, providing dispatching and operation personnel with decision-making support and preventing the occurrence and escalation of faults.

[0003] Existing power system operation alarm methods typically rely on monitoring systems deployed in substations or control centers to periodically collect key electrical measurement data, such as voltage, current, frequency, and power. This collected data is transmitted to the main station backend, where real-time measurements are compared with pre-set fixed thresholds or logical rules. When a measurement exceeds a specified range, an alarm signal is triggered.

[0004] However, the aforementioned existing technologies have significant shortcomings in practical applications. First, their data sources are relatively singular, and they lack an effective mechanism for assessing the quality and reliability of the data itself, making them susceptible to interference from sensor errors or communication problems, resulting in a high false alarm rate. Second, alarm judgment is mainly based on static, pre-set thresholds, which cannot dynamically adapt to changes in power grid operation or the evolution of equipment health status, and are slow to respond to slowly developing potential faults or complex cascading risks. In addition, traditional methods often process data from individual measuring points in isolation, lacking the ability to model and analyze the mutual influence between equipment and the transmission of risks. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a power system operation alarm method and system. It employs a technical solution that combines multi-source data fusion, hierarchical reliability assessment, causal risk modeling, and dynamic threshold adaptive calibration. This approach enables accurate identification and proactive early warning of potential risks in the power system, thereby improving the accuracy and adaptability of alarm decisions.

[0006] The above objectives can be achieved through the following approach:

[0007] A power system operation alarm method includes deploying a distributed intelligent sensing network on the physical entities of the power system. The distributed intelligent sensing network switches between different trigger modes and collects electrical measurement data, environmental status data, and topology relationship data in layers. The data at each layer is preprocessed and physical features are extracted to obtain basic physical features. The basic physical features are semantically annotated, and the physical entities are associated with the topology relationship data. A hierarchical reliability assessment is performed on the basic physical features to quantify the reliability of the data source and generate a reliable physical feature set. Based on the reliable physical feature set and the topology relationship data, a system is constructed... A semantic causal graph is initialized, and multi-source disturbances are applied to simulate equipment anomalies and environmental mutations. Adaptive edge weight optimization is used to iteratively update the causal edge weights according to the node influence degree, and dynamic risk factors are extracted. The basic physical features, the set of credible physical features, and the dynamic risk factors are used to construct a hierarchical state vector. Based on the topological relationship data, a dynamic health threshold vector is constructed. The hierarchical state vector and the dynamic health threshold vector are compared dimension by dimension to generate an operational alarm state. Based on the operational alarm state, attribution analysis and parameter tracing are performed to generate a threshold calibration signal for adjusting the dynamic health threshold vector.

[0008] Optionally, obtaining the basic physical features includes: the distributed intelligent sensing network collecting low-frequency aggregated data in baseline monitoring mode, and the distributed intelligent sensing network collecting high-frequency waveform data in high-density event mode; performing time-domain statistical analysis on the low-frequency aggregated data to extract a first physical feature, and performing time-frequency domain analysis on the high-frequency waveform data to extract a second physical feature; and combining the first physical feature and the second physical feature together to form a basic physical feature.

[0009] Optionally, the distributed intelligent sensing network includes: a multimodal sensor array, an edge computing core, and a self-organizing network communication interface, wherein: the multimodal sensor array is used to perform hierarchical acquisition to obtain the electrical measurement data and the environmental status data; the edge computing core is used to execute switching logic between different trigger modes and to preprocess the data and extract physical features; the self-organizing network communication interface is used to transmit data and control commands between nodes of the distributed intelligent sensing network.

[0010] Optionally, generating a trusted physical feature set includes: associating the basic physical features with the topological relationship data to generate semantic physical features; performing a three-level trust evaluation (source layer, data layer, and network layer) on the semantic physical features to calculate a multi-dimensional trust vector; and binding the semantic physical features with the multi-dimensional trust vector to generate a trusted physical feature set.

[0011] Optionally, the calculation of the multidimensional credibility vector includes: calculating the source credibility component and the data quality credibility component based on the data source attributes and data quality indicators of the semantic physical features; performing a consistency comparison with the topological relationship data based on the semantic physical features to generate a network consistency credibility component; and combining the source credibility component, the data quality credibility component, and the network consistency credibility component together into a multidimensional credibility vector.

[0012] Optionally, the extraction of dynamic risk factors includes: constructing a semantic causal graph based on the trusted physical feature set and the topological relationship data, using the physical entities as nodes and the topological relationship data as edges; applying multi-source perturbation simulation to the semantic causal graph to generate an influence matrix; iteratively updating the semantic causal graph based on the influence matrix; and calculating and extracting dynamic risk factors based on the updated semantic causal graph.

[0013] Optionally, the method further includes: combining the trigger mode switching frequency of the distributed intelligent sensing network with the influence matrix, sorting them by contribution, and identifying key risk contributing nodes.

[0014] Optionally, generating the operational alarm status includes: combining the basic physical features, the trusted physical feature set, and the dynamic risk factor to construct a layered state vector; generating a baseline health threshold vector based on the topological relationship data, and dynamically tightening the dimensions corresponding to the key risk contribution nodes in the baseline health threshold vector to generate a dynamic health threshold vector; and comparing the layered state vector with the dynamic health threshold vector dimension by dimension to generate the operational alarm status.

[0015] Optionally, generating the threshold calibration signal for adjusting the dynamic health threshold vector includes: receiving the handling feedback result of the operation alarm status, and determining the operation alarm status based on the handling feedback result to obtain an alarm determination result; calculating and generating a threshold calibration signal based on the determination result to adjust the dynamic health threshold vector.

[0016] Based on the same inventive concept, this invention also provides a power system operation alarm system, comprising: a physical state perception module, used to deploy a distributed intelligent sensing network on the physical entities of the power system, the distributed intelligent sensing network switching between different trigger modes and collecting electrical measurement data, environmental state data, and topology relationship data in layers, preprocessing and extracting physical features from each layer of data to obtain basic physical features; a feature calibration module, used to perform feature semantic annotation on the basic physical features, associate the physical entities with the topology relationship data, and perform hierarchical reliability assessment on the basic physical features to quantify the reliability of the data source and generate a set of reliable physical features; and a causal risk modeling module, used to base the reliable physical features on the data. The feature set and the topological relationship data are used to construct and initialize a semantic causal graph. Multi-source disturbances are applied to the semantic causal graph to simulate equipment anomalies and environmental mutations. Adaptive edge weight optimization is used to iteratively update the causal edge weights according to the node influence degree to extract dynamic risk factors. The state assessment and alarm decision module is used to construct a hierarchical state vector from the basic physical features, the credible physical feature set, and the dynamic risk factors. Based on the topological relationship data, a dynamic health threshold vector is constructed. The hierarchical state vector and the dynamic health threshold vector are compared dimension by dimension to generate an operational alarm state. The adaptive optimization module is used to perform attribution analysis and parameter tracing based on the operational alarm state to generate a threshold calibration signal for adjusting the dynamic health threshold vector.

[0017] Compared with the prior art, the present invention has the following advantages:

[0018] 1. This invention constructs a complete technology chain from data perception and reliability assessment to risk modeling, and introduces a feedback calibration mechanism to achieve intelligent and accurate alarm decision-making. Through multi-mode switching and hierarchical reliability assessment mechanisms of a distributed intelligent sensing network, the comprehensiveness and reliability of input data are ensured, improving the quality of input data from the source. This effectively suppresses false alarms and missed alarms caused by factors such as sensor failure, communication delay, or data corruption, significantly enhancing the accuracy and reliability of alarm decision-making.

[0019] 2. This invention transforms the static power grid topology into a dynamic risk transmission model by constructing a semantic causal graph and extracting dynamic risk factors. This model can quantitatively assess the mutual influence between different devices and their vulnerability under specific operating conditions. This enables deep insight and forward-looking prediction of potential fault chains, overcoming the limitations of traditional alarm methods that can only make judgments based on the current state. It transforms the alarm system from passive state monitoring to proactive risk warning, providing decision support for preventative maintenance.

[0020] 3. The dynamic health threshold vector and threshold calibration signal mechanism proposed in this invention enable the alarm sensitivity to be precisely matched with the real-time risk level. By dynamically tightening the alarm thresholds of key risk nodes, alarm resources can be focused on the most vulnerable links, achieving refined risk management. Simultaneously, the feedback-driven self-optimizing closed loop provides continuous learning and evolution capabilities, ensuring its long-term adaptability to changes in power grid operating characteristics, thus achieving adaptive and refined alarm strategies.

[0021] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating a power system operation alarm method according to an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of the trusted physical feature set generation process according to an embodiment of the present invention.

[0025] Figure 3 This is a physical entity clustering analysis diagram based on the influence matrix in an embodiment of the present invention.

[0026] Figure 4 This is a key risk contribution node identification and sorting diagram in an embodiment of the present invention.

[0027] Figure 5 This is a schematic diagram of the structure of a power system operation alarm system according to an embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Reference Figure 1 One embodiment of the present invention proposes a power system operation alarm method, which adopts a technical solution combining multi-source data fusion, hierarchical credibility assessment, causal risk modeling and dynamic threshold adaptive calibration, and can achieve accurate identification and forward-looking early warning of potential risks in the power system, thereby improving the accuracy and adaptability of alarm decision-making.

[0030] The method described in this embodiment specifically includes:

[0031] A distributed intelligent sensing network is deployed on the physical entity of the power system. The distributed intelligent sensing network switches between different triggering modes and collects electrical measurement data, environmental status data and topology data in layers. The data at each layer are preprocessed and physical features are extracted to obtain basic physical features.

[0032] The basic physical features are semantically annotated, the physical entities are associated with the topological relationship data, and a hierarchical credibility assessment is performed on the basic physical features to quantify the reliability of the data source and generate a set of credible physical features.

[0033] Based on the trusted physical feature set and the topological relationship data, a semantic causal graph is constructed and initialized. Multi-source disturbances are applied to the semantic causal graph to simulate equipment anomalies and environmental mutations. Adaptive edge weight optimization is used to iteratively update the causal edge weights according to the node influence degree, and dynamic risk factors are extracted.

[0034] The basic physical features, the set of trusted physical features, and the dynamic risk factors are used to construct a hierarchical state vector. Based on the topological relationship data, a dynamic health threshold vector is constructed. The hierarchical state vector and the dynamic health threshold vector are compared dimension by dimension to generate an operational alarm status.

[0035] Based on the operational alarm status, attribution analysis and parameter tracing are performed to generate a threshold calibration signal for adjusting the dynamic health threshold vector.

[0036] By employing a technical solution that combines multi-source data fusion, hierarchical credibility assessment, causal risk modeling, and dynamic threshold adaptive calibration, we can achieve accurate identification and forward-looking early warning of potential risks in the power system, thereby improving the accuracy and adaptability of alarm decisions.

[0037] Optionally, the obtained basic physical characteristics include:

[0038] In baseline monitoring mode, the distributed intelligent sensing network collects low-frequency aggregated data; in high-density event mode, the distributed intelligent sensing network collects high-frequency waveform data.

[0039] Specifically, in this step, the distributed intelligent sensing network adaptively switches between two operating modes. During normal operation, the network operates in baseline monitoring mode, where data is collected and processed at a lower frequency to form low-frequency aggregated data, such as minute-level averages of the RMS values ​​of line voltage and current. When fluctuations in the low-frequency aggregated data exceed a preset stability range, the distributed intelligent sensing network switches to high-density event mode. In this mode, instantaneous voltage and current waveforms are captured at a high sampling frequency, thereby obtaining high-frequency waveform data.

[0040] Time-domain statistical analysis is performed on the low-frequency aggregated data to extract the first physical feature, and time-frequency domain analysis is performed on the high-frequency waveform data to extract the second physical feature;

[0041] Specifically, for low-frequency aggregated data, the processing procedure performs time-domain statistical analysis, calculating statistics such as mean and variance within a time window. These statistics constitute the first physical characteristic characterizing steady-state properties. Simultaneously, for high-frequency waveform data, the processing procedure performs time-frequency domain analysis, such as decomposing waveform components using Fourier transform to calculate the Total Harmonic Distortion (THD) to quantify the degree of waveform distortion. The calculation formula can be expressed as:

[0042] ,

[0043] in, Represents the effective value of the fundamental component; Represents the effective value of each harmonic component; The highest harmonic order is determined by a pre-set integer based on monitoring accuracy requirements. The harmonic content, transient overvoltage amplitude, and other indicators extracted through this type of analysis constitute the second physical characteristic characterizing dynamic disturbance properties.

[0044] The first physical feature and the second physical feature are combined together to form the basic physical feature.

[0045] Specifically, this step combines the first physical feature extracted in the previous step, which describes the steady-state operating characteristics, with the second physical feature, which describes the dynamic disturbance characteristics, to form a multi-dimensional set of basic physical features, providing a comprehensive and reliable basis for subsequent alarm decisions.

[0046] Optionally, the distributed intelligent sensing network includes: a multimodal sensor array, an edge computing core, and a self-organizing network communication interface, wherein:

[0047] The multimodal sensor array is used to perform hierarchical acquisition to obtain the electrical measurement data and the environmental status data;

[0048] Specifically, the distributed intelligent sensing network in this embodiment of the invention is an intelligent hardware node deployed on a physical entity, integrating sensing, computing, and communication capabilities. A multimodal sensor array serves as the sensing front-end of this node, physically integrating various types of sensors. For example, it includes a high-frequency current transformer and a voltage divider sensor for acquiring electrical measurement data, as well as temperature sensors, humidity sensors, and micro-electro-mechanical system (MEMS) vibration sensors for acquiring environmental state data. Each sensor within the array is configured to operate at different sampling rates and accuracies according to instructions from the edge computing core, performing hierarchical data acquisition.

[0049] The edge computing core is used to execute the switching logic between different trigger modes and to preprocess the data and extract physical features;

[0050] Specifically, the edge computing core is the local processing unit of the node, typically implemented by a low-power microcontroller unit (MCU) or system-on-chip (SoC). This core executes the switching logic between different trigger modes. For example, it analyzes data collected in baseline monitoring mode in real time, and immediately triggers a switch to high-density event mode once characteristic fluctuations exceed internally set dynamic thresholds. Simultaneously, this core is also responsible for preprocessing, such as digital filtering and normalization of the collected raw data, and performing physical feature extraction, such as calculating the RMS voltage, power factor, and energy characteristics of transient waveforms locally, thereby reducing the amount of data that needs to be transmitted over the network.

[0051] The self-organizing network communication interface is used to transmit data and control commands between nodes of the distributed intelligent sensing network.

[0052] Specifically, the self-organizing network communication interface is the communication unit of this node, employing a wireless communication protocol that supports mesh network topology. This interface enables nodes in the distributed intelligent sensing network to communicate not only with the central aggregation node but also to relay data among themselves. This self-organizing characteristic ensures that even if some communication links are blocked or fail, data and control commands can still be transmitted through other nodes, greatly enhancing the communication robustness and reliability of the entire sensing network.

[0053] Optionally, generating a reliable physical feature set includes:

[0054] The basic physical features are associated with the topological relationship data to generate semantic physical features;

[0055] Specifically, this step aims to further process the basic physical features obtained in the previous step to endow them with physical semantics. The process associates each data point from the basic physical features with power system topology data. This topology data details the unique identifiers of each physical entity in the power grid and their electrical connections. Through this association operation, an abstract measurement value, such as temperature data, is precisely labeled with its source, such as "temperature of phase A winding of main transformer No. 1 in a certain substation," thereby generating a semantic physical feature with clear physical meaning.

[0056] For the semantic physical features, a three-level credibility assessment is performed at the source layer, data layer, and network layer to calculate a multi-dimensional credibility vector.

[0057] Specifically, for semantically annotated semantic physical features, the processing performs a three-level credibility assessment to calculate a multi-dimensional credibility vector. The first layer is the source-end layer assessment, which primarily evaluates the reliability of the data acquisition equipment itself, such as reading attributes like the sensor's factory calibration accuracy and the date of the most recent calibration, and quantifying these as a source-end credibility component. The second layer is the data layer assessment, which focuses on evaluating the quality of the data itself, such as its real-time performance and completeness, and calculates a data quality credibility component. The third layer is the network layer assessment, a cross-validation process based on physical laws. It uses topological relationship data to compare the consistency of multiple semantically annotated physical features located on the same electrical node or associated path. For example, according to Kirchhoff's current law, the total current flowing into a node should be zero; the deviation of actual measured values ​​is calculated to generate a network consistency credibility component.

[0058] The semantic physical features are bound to the multidimensional credibility vector to generate a credibility physical feature set.

[0059] Specifically, this step involves binding each semantic physical feature to its corresponding multidimensional credibility vector. This binding operation encapsulates the feature's numerical value, semantics, and credibility into a unified data structure, thereby forming the final set of credible physical features. This provides a solid data foundation for subsequent analysis and decision-making, such as... Figure 2 As shown in the diagram, the nodes in the outer circle represent key data entities and processing modules, and the flow bands between the nodes clearly reveal the complete path of data from input, through layers of processing and evaluation, to the final fusion output.

[0060] Optionally, the calculation of the multidimensional credibility vector includes:

[0061] Based on the data source attributes and data quality indicators of the semantic physical features, the source credibility component and the data quality credibility component are calculated.

[0062] Specifically, this step aims to deeply quantify the reliability of semantic physical features. The processing calculates the source reliability component and the data quality reliability component based on the data source attributes and data quality indices embedded within each semantic physical feature. Data source attributes mainly refer to the static parameters of the sensor that acquired the feature, such as its factory accuracy level, calibration cycle, and years of use. These attributes are quantified and comprehensively calculated to obtain the source reliability component. Data quality indices refer to the dynamic attributes generated during data transmission and recording, such as data transmission latency and data packet integrity rate. The data quality reliability component is calculated by evaluating these indices.

[0063] Based on the semantic physical features, a consistency comparison is performed with the topological relationship data to generate a network consistency credibility component. The source credibility component, the data quality credibility component, and the network consistency credibility component are then combined into a multidimensional credibility vector.

[0064] Specifically, the core of generating network consistency reliability components is to utilize the inherent physical laws of the power system to cross-validate semantic physical features. This method identifies a set of physically interrelated semantic physical features based on topological relationship data. For example, for an electrical node in the power grid, according to Kirchhoff's current law, at any given time, the phasor sum of all line currents flowing into that node should theoretically be zero. The residual of the actual measured current is calculated using the following formula to assess data consistency:

[0065] ,

[0066] in, From semantic physical features Current phasors of the connecting lines; This is the total number of lines connected to this node, derived from an integer parsed from the topology data. The calculated residuals... The smaller the value, the higher the consistency between the measurement data of each sensor. This residual value is then mapped to a network consistency confidence component between 0 and 1. Finally, the processing combines the source confidence component, data quality confidence component, and network consistency confidence component calculated through the above steps into a structured multidimensional confidence vector.

[0067] Optionally, the extraction of dynamic risk factors includes:

[0068] Based on the trusted physical feature set and the topological relationship data, a semantic causal graph is constructed by initializing the physical entities as nodes and the topological relationship data as edges.

[0069] Specifically, this step aims to transform the static topology of the power system into a dynamic causal relationship model. The process utilizes the reliable physical feature set obtained in the previous steps and the topological relationship data of the power system. Physical entities in the power grid, such as generators, transformers, and lines, are used as nodes in the graph, and the electrical connections between these entities—the topological relationship data—are used as the initial edges of the graph, thus constructing a basic semantic causal graph. At this point, the graph structure only reflects physical connections; the edges have not yet been assigned quantified causal strength.

[0070] Multi-source perturbation simulation is applied to the semantic causal graph to generate an influence matrix, and the semantic causal graph is iteratively updated based on the influence matrix;

[0071] Specifically, this step aims to dynamically optimize the semantic causal graph. The process applies multi-source disturbances to the constructed semantic causal graph through computational simulation. These disturbance simulations can be performed using mature power grid simulation analysis software, such as PSASP and BPA. The types and parameters of the disturbances can be set to typical severe faults in the power grid contingency plan, such as: setting a three-phase metallic short-circuit fault lasting 100ms on a critical transmission line, simulating an N-1 tripping event of an important generator unit, or instantaneously increasing the load of a large industrial unit by 50%. For each simulated disturbance, the process calculates the propagation effect of the disturbance throughout the entire network based on the physical model of the power grid operation; that is, the degree of influence of a node's state change on the states of all other nodes. By applying multiple types of disturbances to different nodes and analyzing their global impact, an influence matrix can be generated, where each element quantifies the strength of the mutual influence between nodes. Subsequently, based on this influence matrix, the process iteratively updates the causal edge weights of the semantic causal graph, such as... Figure 3 The diagram illustrates an example of an influence matrix, where rows and columns represent physical entities within the system, and the size and color intensity of the bubbles represent the strength of influence between entities. The clustering dendrogram in the diagram automatically groups highly correlated entities based on influence patterns, thereby identifying potential risk coupling groups.

[0072] Based on the updated semantic causal graph, dynamic risk factors are calculated and extracted.

[0073] Specifically, this step aims to calculate and extract dynamic risk factors based on an iteratively updated semantic causal graph. These risk factors are key indicators for quantifying the risk status of the power grid and are typically calculated using graph theory algorithms. For example, by calculating the weighted out-degree of each node in the graph, we can obtain the node's ability to propagate its influence as a risk source, which is one dynamic risk factor; by calculating the weighted in-degree of a node, we can obtain the node's vulnerability to the influence of other nodes, which is another dynamic risk factor.

[0074] Optionally, the method further includes:

[0075] By combining the trigger mode switching frequency of the distributed intelligent sensing network with the influence matrix, and sorting them by contribution, key risk contributing nodes are identified.

[0076] Specifically, this step aims to perform a deep prioritization of risk for all physical entity nodes in the power grid. It does not rely on a single indicator but integrates two different types of information: first, the trigger mode switching frequency of the distributed intelligent sensing network, which characterizes the current operational activity and stability of the node; and second, the influence matrix, which characterizes the node's potential global influence. The process first counts the number of times the distributed intelligent sensing network associated with each physical entity switches from baseline monitoring mode to high-density event mode within a preset time window, obtaining the node's trigger mode switching frequency as its "dynamic activity" indicator. Simultaneously, the process calls the influence matrix generated in the previous step to extract the comprehensive influence of each node on all other nodes in the network, as its "static influence" indicator. The core step involves using a weighted fusion model to comprehensively calculate the normalized "dynamic activity" and "static influence" indicators to obtain a unified contribution score. Finally, the process ranks all physical entities in descending order based on this contribution score, and the top-ranked nodes are identified as key risk contribution nodes, such as... Figure 4 As shown, the results of ranking all physical entities by contribution are presented. Each entity corresponds to a "lollipop," the length of which represents the entity's final overall contribution score. By ranking from highest to lowest, the key nodes that contribute the most to the overall risk of the system can be identified.

[0077] Optionally, the generation of runtime alarm status includes:

[0078] The basic physical features, the set of credible physical features, and the dynamic risk factors are combined to construct a hierarchical state vector;

[0079] Specifically, this step aims to integrate three types of key information to construct a comprehensive hierarchical state vector. These three types of information are: basic physical characteristics that directly reflect the measured values ​​of physical quantities; a set of credible physical features processed through semanticization and credible quantification; and dynamic risk factors that proactively reveal vulnerabilities and risk transmission paths. By combining these three elements according to the dimension of physical entities, a hierarchical state vector is formed that includes the current true state, data credibility, and future risk trends. Each dimension of this vector precisely corresponds to a specific parameter of a specific device in the power grid.

[0080] A baseline health threshold vector is generated based on the topological relationship data, and the dimensions corresponding to the key risk contribution nodes in the baseline health threshold vector are dynamically tightened to generate a dynamic health threshold vector.

[0081] Specifically, this step aims to construct a benchmark for comparison, namely a dynamic health threshold vector. The process first generates an initial health threshold for each dimension of the hierarchical state vector based on power system topology data and industry operating procedures. These thresholds collectively constitute the benchmark health threshold vector. Subsequently, the process retrieves the list of key risk-contributing nodes identified in the previous steps and locates the dimensions in the benchmark health threshold vector associated with these high-risk nodes. For these specific dimensions, the process performs a dynamic tightening operation, appropriately reducing the allowable range of their health thresholds. In this way, the static benchmark health threshold vector evolves into a dynamic health threshold vector that reflects the risk distribution in real time.

[0082] The hierarchical state vector is compared with the dynamic health threshold vector dimension by dimension to generate the running alarm status.

[0083] Specifically, this step aims to perform a dimension-by-dimensional comparison operation. The process compares each element value in the hierarchical state vector with its corresponding dynamically tightened threshold in the dynamic health threshold vector. Once any element in the state vector exceeds its corresponding dynamic health threshold, the process determines that an anomaly has occurred. Based on the number of elements exceeding the threshold, the degree of exceedance, and the dynamic risk factors associated with these elements, a clear operational alarm status is ultimately generated, such as normal, warning, or critical alarm.

[0084] Optionally, generating the threshold calibration signal for adjusting the dynamic health threshold vector includes:

[0085] Receive the handling feedback result of the operation alarm status, and determine the operation alarm status based on the handling feedback result to obtain the alarm determination result;

[0086] Specifically, this step aims to build a feedback-driven, self-optimizing closed loop, ensuring that the alarm system can continuously learn and improve from historical experience. The process begins by receiving and parsing the handling feedback results for the previous operational alarm status. This handling feedback can come from manual maintenance logs, which detail the conclusions of on-site verification by maintenance personnel, or from the action records of power system automation devices. Based on this handling feedback result, the process performs a post-event judgment on the previously generated operational alarm status, thereby obtaining a clear alarm judgment result. This result is typically classified as a valid alarm (the alarm was accurate); a false alarm (an alarm was issued but no actual abnormality occurred); or a missed alarm (an event that should have triggered an alarm was not issued but was later confirmed to have occurred).

[0087] Based on the determination result, a threshold calibration signal is calculated and generated to adjust the dynamic health threshold vector.

[0088] Specifically, after obtaining the alarm determination result, the processing enters the core calculation stage to generate a threshold calibration signal with a clear adjustment direction and magnitude. This calculation process follows a set of adjustment logic: if the alarm determination result is a false alarm, it indicates that the current dynamic health threshold vector is too strict, and the processing will calculate an adjustment amount to relax the threshold; conversely, if the determination result is a missed alarm, it indicates that the threshold is too lenient, and the processing will calculate an adjustment amount to tighten the threshold; if the alarm is determined to be valid, the adjustment amount can be set to zero. The magnitude of this adjustment amount is also controlled by a learning rate to avoid excessive oscillations due to a single event. Finally, this calculation result, which contains the specific adjustment direction and value, is encapsulated as a threshold calibration signal to update the dynamic health threshold vector.

[0089] To verify the feasibility of this invention in practice, it was applied to the operation monitoring and alarm system of a municipal regional power grid. To ensure the reliability of power supply in the city's core area, this regional power grid requires real-time monitoring of multiple substations, transmission lines, and other key physical entities to achieve early warning and accurate location of potential faults. Traditional alarm systems suffer from problems such as fixed thresholds, incomplete alarm information, and difficulty in distinguishing fault severity, often leading to missed early warnings and false alarms for non-emergency disturbances. The municipal power grid dispatch center aims to use the method of this invention to construct a closed-loop alarm system encompassing intelligent sensing, data refinement, risk projection, and dynamic decision-making.

[0090] In this embodiment, the power grid dispatch center deployed the distributed intelligent sensing network of this invention at multiple key substations and important transmission lines. This network collects electrical measurement data and environmental status data, such as voltage, current, equipment temperature, and ambient humidity, through a multimodal sensor array, and performs localized feature extraction and mode switching by an edge computing core. The entire alarm system operated for six months, recording various operating conditions of the power grid during this period, and its effectiveness was compared with that of a traditional fixed-threshold alarm system operating during the same period.

[0091] In this embodiment, the present invention successfully captured and analyzed an early latent fault in a transformer. A distributed intelligent sensing network deployed on the No. 2 main transformer of the Chengnan substation operated in baseline monitoring mode, collecting low-frequency aggregated data such as effective voltage and active power at a second-level frequency. Around 10:20, the edge computing core detected a small but continuous fluctuation in the effective voltage value on the low-voltage side of the transformer, exceeding a preset stable range. It immediately executed switching logic between different trigger modes, switching the multi-modal sensor array to high-density event mode, capturing instantaneous high-frequency voltage and current waveform data at a kilohertz-level high sampling rate.

[0092] The method then processes the two types of data in parallel. For the low-frequency aggregated data before the mode switch, time-domain statistical analysis is performed, calculating that the voltage standard deviation increased by 15%. For the high-frequency waveform data acquired after the mode switch, time-frequency domain analysis is performed, and Fourier transform is used to calculate that the total harmonic distortion rate is 4.2%, significantly higher than the normal value of 2.5%. These two characteristics are combined to form the basic physical characteristics reflecting the abnormal condition of the transformer.

[0093] Next, the method performs feature calibration on the basic physical features. First, these features are associated with topological relationship data and semantically labeled as "standard deviation of voltage on the low-voltage side of main transformer No. 2 in Chengnan Substation" and "THD on the low-voltage side of main transformer No. 2 in Chengnan Substation," generating semantic physical features. Subsequently, a hierarchical reliability assessment is performed on these features: the source layer assessment confirms that the sensor calibration records are good; the data layer assessment confirms low data transmission latency and no packet loss; the network layer assessment uses topological relationships for consistency comparison and cross-validates the currents of each line flowing into the substation bus, showing minimal residuals and ruling out the possibility of sensor failures on other lines. These three components are combined into a multidimensional reliability vector and bound to the semantic physical features, confirming the reliability of the anomalous data and forming a reliable physical feature set.

[0094] Based on this reliable physical feature set, the method updates the pre-constructed semantic causal graph of the regional power grid. Through historical multi-source disturbance simulations, an influence matrix has been generated, showing that the No. 2 main transformer of the Chengnan substation has a high influence on several important downstream loads. Simultaneously, the trigger mode switching frequency of the distributed intelligent sensing network at this node has occurred three times in the past week, significantly higher than other nodes. By combining these two indicators and ranking them according to contribution, the No. 2 main transformer of the Chengnan substation is identified as a key risk contributing node. Based on the updated semantic causal graph, dynamic risk factors characterizing the risk transmission capability of this node are calculated and extracted using graph aggregation functions.

[0095] In the alarm decision-making phase, this method combines basic physical characteristics, a set of reliable physical characteristics, and dynamic risk factors to construct a hierarchical state vector. A baseline health threshold vector is generated based on topological relationship data. Since the No. 2 main transformer at the Chengnan substation is identified as a critical risk contribution node, this method dynamically tightens the dimension corresponding to this node in the baseline health threshold vector, generating a dynamic health threshold vector. For example, its baseline health threshold for total harmonic distortion (THD) is 5.0%, which becomes 3.5% after dynamic tightening. At this point, the measured THD value of 4.2% in the hierarchical state vector is compared dimension-by-dimensionally with the dynamic health threshold of 3.5%, and it is found to have exceeded the limit. Immediately at 10:23, a "warning" level operational alarm status is generated, clearly indicating that the transformer has a waveform distortion risk. Meanwhile, the traditional alarm system, whose fixed 5.0% threshold is not triggered, fails to detect this anomaly.

[0096] After receiving the alarm, the dispatch center dispatched maintenance personnel to investigate. The feedback indicated slight signs of insulation aging in the transformer's internal windings. This feedback was entered as a "valid alarm," resulting in an alarm determination. Because the alarm was accurate, a threshold calibration signal was calculated based on this determination, positively confirming the current dynamic health threshold vector adjustment logic and enabling more accurate judgments in similar future events.

[0097] Through a 6-month comparative test, the present invention demonstrated significant advantages in alarm accuracy, foresight, and adaptability, as shown in Tables 1 to 3.

[0098] Table 1. Key Risk Contribution Node Identification Data Table

[0099]

[0100] Table 2 Comparison of Dynamic Health Threshold Adjustment and Alarm Triggering

[0101]

[0102] Table 3 Comparison of Running Results

[0103]

[0104] Tables 1 to 3 above record the actual application data of the present invention in regional power grids, and demonstrate in detail the performance of the method in identifying key risks, dynamic decision-making and overall performance.

[0105] As can be seen from the data in Table 1, the present invention can effectively integrate the real-time "dynamic activity" and potential "static influence" of nodes, accurately identify the No. 2 main transformer of Chengnan Substation as the most noteworthy risk source, and provide a basis for subsequent differentiated monitoring.

[0106] Table 2 clearly illustrates the core advantages of this invention. For traditional alarm systems, a THD of 4.2% is considered normal at a fixed threshold of 5.0%. However, this invention, by identifying the transformer as a critical risk contributor, dynamically tightens the threshold to 3.5%, thus successfully identifying and issuing an early warning in the early stages of a fault. This demonstrates the decisive role of the dynamic health threshold mechanism in improving alarm sensitivity.

[0107] Table 3 provides overall performance data that macroscopically confirms the technical effectiveness of this invention. Through a closed-loop intelligent processing flow, problems can be detected earlier, and alarm quality is significantly improved. While accuracy is increased, false alarms and missed alarms are greatly reduced. This fully demonstrates that this invention can provide more reliable, intelligent, and forward-looking technical guarantees for the safe and stable operation of power systems.

[0108] Based on the same inventive concept, the present invention also provides an operation alarm system for a power system, such as... Figure 5 As shown, the system includes:

[0109] The physical state sensing module is used to deploy a distributed intelligent sensing network on the physical entities of the power system. The distributed intelligent sensing network switches between different triggering modes and collects electrical measurement data, environmental state data and topology data in layers. It preprocesses and extracts physical features from the data at each layer to obtain basic physical features.

[0110] The feature calibration module is used to perform feature semantic annotation on the basic physical features, associate the physical entities with the topological relationship data, perform hierarchical credibility assessment on the basic physical features, quantify the reliability of the data source, and generate a set of credible physical features.

[0111] The causal risk modeling module is used to construct and initialize a semantic causal graph based on the trusted physical feature set and the topological relationship data, apply multi-source perturbations to the semantic causal graph to simulate equipment anomalies and environmental mutations, use adaptive edge weight optimization to iteratively update the causal edge weights according to the node influence degree, and extract dynamic risk factors.

[0112] The status assessment and alarm decision module is used to construct a hierarchical status vector from the basic physical features, the trusted physical feature set and the dynamic risk factors, construct a dynamic health threshold vector based on the topological relationship data, and compare the hierarchical status vector with the dynamic health threshold vector dimension by dimension to generate an operation alarm status.

[0113] The adaptive optimization module is used to perform attribution analysis and parameter tracing based on the operational alarm status, and generate a threshold calibration signal for adjusting the dynamic health threshold vector.

[0114] It should be noted that the functional division and information interaction between the various modules described above are logical, but in terms of physical implementation, they can be integrated on the same software platform or deployed in a distributed manner. The connections between them represent data flow and control flow, aiming to collaboratively achieve the objectives of this invention. The above descriptions are merely exemplary embodiments of this invention and should not be construed as limiting the scope of protection of this invention.

Claims

1. A power system operation alarm method, characterized in that, The method includes: A distributed intelligent sensing network is deployed on the physical entity of the power system. The distributed intelligent sensing network switches between different triggering modes and collects electrical measurement data, environmental status data and topology data in layers. The data at each layer are preprocessed and physical features are extracted to obtain basic physical features. The basic physical features are semantically annotated, the physical entities are associated with the topological relationship data, and a hierarchical credibility assessment is performed on the basic physical features to quantify the reliability of the data source and generate a set of credible physical features. Based on the trusted physical feature set and the topological relationship data, a semantic causal graph is constructed and initialized. Multi-source disturbances are applied to the semantic causal graph to simulate equipment anomalies and environmental mutations. Adaptive edge weight optimization is used to iteratively update the causal edge weights according to the node influence degree, and dynamic risk factors are extracted. The basic physical features, the set of trusted physical features, and the dynamic risk factors are used to construct a hierarchical state vector. Based on the topological relationship data, a dynamic health threshold vector is constructed. The hierarchical state vector and the dynamic health threshold vector are compared dimension by dimension to generate an operational alarm status. Based on the operational alarm status, perform attribution analysis and parameter tracing to generate a threshold calibration signal for adjusting the dynamic health threshold vector; The obtained basic physical characteristics include: In baseline monitoring mode, the distributed intelligent sensing network collects low-frequency aggregated data; in high-density event mode, the distributed intelligent sensing network collects high-frequency waveform data. Time-domain statistical analysis is performed on the low-frequency aggregated data to extract the first physical feature, and time-frequency domain analysis is performed on the high-frequency waveform data to extract the second physical feature; The first physical feature and the second physical feature are combined together to form the basic physical feature; The generation of the reliable physical feature set includes: The basic physical features are associated with the topological relationship data to generate semantic physical features; For the semantic physical features, a three-level credibility assessment is performed at the source layer, data layer, and network layer to calculate a multi-dimensional credibility vector. The semantic physical features are bound to the multidimensional credibility vector to generate a set of credible physical features; The calculated multidimensional credibility vector includes: Based on the data source attributes and data quality indicators of the semantic physical features, the source credibility component and the data quality credibility component are calculated. Based on the semantic physical features, a consistency comparison is performed with the topological relationship data to generate a network consistency credibility component. The source credibility component, the data quality credibility component, and the network consistency credibility component are then combined into a multidimensional credibility vector.

2. The power system operation alarm method according to claim 1, characterized in that, The distributed intelligent sensing network includes: a multimodal sensor array, an edge computing core, and a self-organizing network communication interface, wherein: The multimodal sensor array is used to perform hierarchical acquisition to obtain the electrical measurement data and the environmental status data; The edge computing core is used to execute the switching logic between different trigger modes and to preprocess the data and extract physical features; The self-organizing network communication interface is used to transmit data and control commands between nodes of the distributed intelligent sensing network.

3. The power system operation alarm method according to claim 1, characterized in that, The extraction of dynamic risk factors includes: Based on the trusted physical feature set and the topological relationship data, a semantic causal graph is constructed by initializing the physical entities as nodes and the topological relationship data as edges. Multi-source perturbation simulation is applied to the semantic causal graph to generate an influence matrix, and the semantic causal graph is iteratively updated based on the influence matrix; Based on the updated semantic causal graph, dynamic risk factors are calculated and extracted.

4. The power system operation alarm method according to claim 3, characterized in that, The method further includes: By combining the trigger mode switching frequency of the distributed intelligent sensing network with the influence matrix, and sorting them by contribution, key risk contributing nodes are identified.

5. The power system operation alarm method according to claim 4, characterized in that, The generated runtime alarm status includes: The basic physical features, the set of credible physical features, and the dynamic risk factors are combined to construct a hierarchical state vector; A baseline health threshold vector is generated based on the topological relationship data, and the dimensions corresponding to the key risk contribution nodes in the baseline health threshold vector are dynamically tightened to generate a dynamic health threshold vector. The hierarchical state vector is compared with the dynamic health threshold vector dimension by dimension to generate the running alarm status.

6. The power system operation alarm method according to claim 1, characterized in that, The generation of the threshold calibration signal for adjusting the dynamic health threshold vector includes: Receive the handling feedback result of the operation alarm status, and determine the operation alarm status based on the handling feedback result to obtain the alarm determination result; Based on the determination result, a threshold calibration signal is calculated and generated to adjust the dynamic health threshold vector.

7. A power system operation alarm system, applied to a power system operation alarm method as described in any one of claims 1-6, characterized in that, The system includes: The physical state sensing module is used to deploy a distributed intelligent sensing network on the physical entities of the power system. The distributed intelligent sensing network switches between different triggering modes and collects electrical measurement data, environmental state data and topology data in layers. It preprocesses and extracts physical features from the data at each layer to obtain basic physical features. The feature calibration module is used to perform feature semantic annotation on the basic physical features, associate the physical entities with the topological relationship data, perform hierarchical credibility assessment on the basic physical features, quantify the reliability of the data source, and generate a set of credible physical features. The causal risk modeling module is used to construct and initialize a semantic causal graph based on the trusted physical feature set and the topological relationship data, apply multi-source perturbations to the semantic causal graph to simulate equipment anomalies and environmental mutations, use adaptive edge weight optimization to iteratively update the causal edge weights according to the node influence degree, and extract dynamic risk factors. The status assessment and alarm decision module is used to construct a hierarchical status vector from the basic physical features, the trusted physical feature set and the dynamic risk factors, construct a dynamic health threshold vector based on the topological relationship data, and compare the hierarchical status vector with the dynamic health threshold vector dimension by dimension to generate an operation alarm status. The adaptive optimization module is used to perform attribution analysis and parameter tracing based on the operational alarm status, and generate a threshold calibration signal for adjusting the dynamic health threshold vector.

Images