Power system source network load storage panoramic scheduling operation monitoring and collaborative alarm method

Abstract

The invention discloses a source-grid-load-storage panoramic scheduling operation monitoring and collaborative warning method for an electric power system. The method comprises the following steps: collecting real-time operation data of a power supply side, a power grid side, a load side and an energy storage side in the electric power system; performing dynamic weight adaptive clustering processing on the plurality of pieces of real-time operation data to generate a panoramic operation situation quantification result, the dynamic weight adaptive clustering processing being used for adaptively adjusting the weight of the plurality of pieces of real-time operation data according to a system operation stage and generating a panoramic operation situation label and a panoramic operation situation index through clustering analysis; and based on the panoramic operation situation quantification result, triggering collaborative alarm for the power supply side, the power grid side, the load side and the energy storage side. Through the dynamic weight adaptive clustering processing, the defects of lack of overall situation assessment caused by information enumeration in traditional scheduling monitoring and cross-link association missing caused by alarm isolation are overcome.

Description

Technical Field

[0001] This invention relates to the field of power system dispatch automation technology, and in particular to a method for panoramic dispatch operation monitoring and collaborative alarm of power system source-grid-load-storage. Background Technology

[0002] With the deepening of the construction of new power systems, the volatility of high-proportion renewable energy sources on the power generation side, the growth of diversified interactive demands on the load side, and the large-scale integration of energy storage have made the operation of power systems increasingly complex, and the coupling and interaction of various links in the "source-grid-load-storage" system are becoming increasingly close. Against this backdrop, the dispatch center needs to have a panoramic perception and control of the entire system's operating status, and traditional dispatch and operation monitoring methods are no longer sufficient to meet this requirement.

[0003] Traditional dispatch and monitoring systems typically collect and display information separately for each link: "source," "network," "load," and "storage." Their monitoring screens are often simply a juxtaposition and accumulation of various real-time data, curves, and alarm information. This approach is essentially an "information listing" model, lacking the ability to deeply integrate and intelligently analyze multi-dimensional, multi-temporal scale operational data. It fails to quantitatively assess the overall operational status of the system. Dispatchers struggle to quickly and intuitively determine whether the system is in a relaxed, balanced, tense, or critical state, resulting in a lag and inaccuracy in grasping the macro-level operational situation. This hinders the provision of accurate and timely situational awareness input for subsequent coordinated control decisions.

[0004] Furthermore, existing alarm systems primarily rely on independent alarms from various disciplines. For example, alarms related to grid protection actions, renewable energy disconnection, and load control activation are often generated and displayed in isolation. When complex cascading events occur across multiple systems, a massive influx of discrete alarm messages emerges instantly. This not only exacerbates information overload but also, due to the lack of causal relationships between alarms, dispatchers must expend considerable time and effort on manual sorting and analysis to even begin to deduce the root cause of the fault and the scope of its impact. This "phenomenal alarm" model severs the intrinsic connection between "source-grid-load-storage," making it difficult to quickly formulate a coordinated response strategy in emergencies, thus reducing the efficiency of emergency response and the resilience of the system.

[0005] Therefore, how to break through the limitations of traditional separate monitoring and isolated alarms, design a method that can integrate real-time, adaptive panoramic operation data of "source-grid-load-storage", and on this basis achieve accurate situation quantification and collaborative intelligent alarms, has become a key technical problem that needs to be solved to improve the intelligent level of dispatching and operation of new power systems and ensure the safe and stable operation of the power grid. Summary of the Invention

[0006] The purpose of this invention is to provide a method for panoramic dispatching and operation monitoring and collaborative alarming of power system sources, grids, loads and storage. Through the dynamic weight adaptive clustering process, it overcomes the shortcomings of traditional dispatching and monitoring, such as the lack of overall situation assessment due to information listing and the lack of cross-link correlation due to isolated alarms. It realizes quantitative assessment and collaborative alarming of the panoramic operation status of power system sources, grids, loads and storage.

[0007] To address the aforementioned technical problems, a first aspect of this invention provides a method for panoramic dispatching and operation monitoring and coordinated alarming of power system sources, grid, load, and storage, comprising the following steps: Collect real-time operational data from the power source side, grid side, load side, and energy storage side of the power system; Dynamic weight adaptive clustering is performed on several real-time running data to generate a panoramic running status quantification result. The dynamic weight adaptive clustering is used to adaptively adjust the weights of several real-time running data according to the system running stage and generate panoramic running status labels and panoramic running status indices through cluster analysis. Based on the quantitative results of the panoramic operation status, coordinated alarms are triggered for the power supply side, grid side, load side, and energy storage side.

[0008] Furthermore, the step of performing dynamic weight adaptive clustering processing on several of the real-time operational data to generate a panoramic operational status quantification result includes: The corresponding operational characteristic quantities are extracted from the real-time operational data of the power supply side, grid side, load side and energy storage side respectively; According to the system operation stage, weights are dynamically assigned to several operational feature quantities, and a current weighted feature vector is generated based on the real-time operational data. Clustering is performed on multiple historical weighted feature vectors formed from historical data to establish a tag library containing several panoramic operation status labels; The current weighted feature vector generated based on the real-time operation data is matched with the tag library to output the corresponding panoramic operation status tag, and the quantized distance between the current weighted feature vector and the cluster center of the matched panoramic operation status tag is calculated as the panoramic operation status index.

[0009] Furthermore, the step of extracting corresponding operational characteristic quantities from the real-time operational data of the power supply side, grid side, load side, and energy storage side includes: Based on the real-time operating data of the power supply side, power supply side characteristic quantities are calculated, including power output fluctuation rate and power output prediction deviation. Based on the real-time operating data of the power grid side, the characteristic quantities of the power grid side are calculated, including the power flow margin and voltage stability margin of the key section. Based on the real-time operating data of the load side, load side characteristic quantities are calculated, including demand response potential and probability of uncontrollable load mutations. Based on the real-time operating data of the energy storage side, the characteristic quantities of the energy storage side are calculated, including the state of charge and adjustable power. The operating characteristic is obtained by combining the power supply side characteristic, the grid side characteristic, the load side characteristic, and the energy storage side characteristic.

[0010] Furthermore, the step of dynamically assigning weights to several operational features based on the system's operational phase, and generating a current weighted feature vector based on the real-time operational data, includes: Based on the total system load and total renewable energy output in the real-time operating data, the current system operation stage is identified, which includes the peak load stage, the renewable energy generation stage, and the normal balance stage. Based on the system operation phase, a preset benchmark weight configuration is obtained, which includes the initial weight values ​​of the power supply side characteristic, the grid side characteristic, the load side characteristic, and the energy storage side characteristic. Based on the real-time values ​​of the power supply side characteristics, the grid side characteristics, the load side characteristics, and the energy storage side characteristics, the initial weight values ​​in the benchmark weight configuration are dynamically and collaboratively corrected to generate the final weight of each characteristic. The final weights are used to weight the values ​​of the power source side characteristic, the grid side characteristic, the load side characteristic, and the energy storage side characteristic, respectively, and all the weighted values ​​are combined into a vector in a predetermined order to generate the current weighted feature vector.

[0011] Furthermore, the dynamic and coordinated correction of the initial weight values ​​in the benchmark weight configuration based on the real-time values ​​of the power supply-side characteristics, the grid-side characteristics, the load-side characteristics, and the energy storage-side characteristics to generate the final weight for each characteristic includes: The deviation of the real-time value of the output fluctuation rate in the power supply side characteristic quantity from its preset standard value is calculated and used as the power supply side fluctuation deviation. The negative deviation of the real-time value of the power flow margin of the key section in the power grid side characteristic quantity relative to its preset safety threshold is calculated as the power grid congestion deviation. The growth deviation of the real-time value of the uncontrollable load mutation probability among the load-side characteristic quantities relative to its historical baseline value is calculated as the load mutation deviation. The deviation of the real-time value of the state of charge in the energy storage side characteristic quantity from the boundary of its preset optimal range is calculated as the energy storage state deviation. Based on the power supply side fluctuation deviation, the grid congestion deviation, the load change deviation, and the energy storage state deviation, the weight correction coefficients corresponding to the power supply side characteristic, the grid side characteristic, the load side characteristic, and the energy storage side characteristic are calculated respectively through a preset collaborative correction function. The initial weight value of each feature is multiplied by the corresponding weight correction coefficient to obtain the final weight of the corresponding feature.

[0012] Furthermore, the multiple historical weighted feature vectors formed based on historical data are clustered to establish a tag library containing several panoramic operational status tags, including: Obtain multiple historical weighted feature vectors classified and stored according to the system operation stage within a historical time period, with each historical weighted feature vector corresponding to a specific historical moment; Clustering algorithms are used to perform cluster analysis on the set of historical weighted feature vectors belonging to the same stage of system operation, dividing the feature space into several clusters and determining the center vector of each cluster. For each cluster, based on the magnitude of the characteristic components in its central vector that represent the power supply side fluctuation level, grid side congestion level, load side sudden change risk and energy storage side support capacity, the cluster is assigned the panoramic operation status label according to the preset label generation rules. Establish a mapping relationship between the center vector of the cluster and the corresponding panoramic operation status label, and the label library is composed of the center vector of all clusters and their mapped panoramic operation status labels.

[0013] Further, the step of calculating the quantized distance between the current weighted feature vector and the cluster centers of the matched panoramic operational status labels as the panoramic operational status index includes: Obtain the cluster center vector corresponding to the panoramic operation status label; Based on all historical weighted feature vectors belonging to the cluster center vector, calculate the statistical distribution of the distance between them and the cluster center vector, and determine the distance benchmark for measuring the degree of tension based on the statistical distribution. Calculate the weighted Euclidean distance between the current weighted feature vector and the cluster center vector; The calculated weighted Euclidean distance is compared with the distance benchmark. By using a preset normalization mapping function, the weighted Euclidean distance is converted into a continuous scalar within a preset numerical range, thus obtaining the panoramic operating status index that characterizes the degree to which the overall operating state of the system deviates from its typical pattern.

[0014] Further, the step of calculating the statistical distribution of the distance between the cluster center vector and the cluster center vector, and determining a distance benchmark for measuring the degree of tension based on the statistical distribution, includes: Calculate the set of historical distances between all historical weighted feature vectors belonging to the cluster center vector and the cluster center vector, and calculate its statistical mean and standard deviation based on the set of historical distances; Based on the numerical distribution characteristics of the historical distance set, historical distance values ​​that represent significant anomalies in the operating status within the cluster are selected from the historical distance set and used as distance references to represent typical abnormal fluctuations within the cluster. Based on the preset importance level of the system operation mode represented by the panoramic operation status label corresponding to the cluster, the preliminary benchmark value determined based on the statistical mean and the distance reference is adjusted. Based on the different requirements for operational stability at the current system operation stage, the distance benchmark value after importance adjustment is finally calibrated to generate the distance benchmark used for current situation assessment.

[0015] Further, the step of comparing the calculated weighted Euclidean distance with the distance benchmark, and converting the weighted Euclidean distance into a continuous scalar within a preset numerical range using a preset normalization mapping function, to obtain the panoramic operational status index characterizing the degree to which the overall system operating state deviates from its typical pattern, includes: The weighted Euclidean distance is compared with the distance reference to determine the degree of relative deviation of the weighted Euclidean distance from the distance reference. Based on the different numerical intervals in which the relative deviation is located, the weighted Euclidean distance is mapped to a first mapping value located in the first preset numerical interval through a piecewise linear mapping function. The first mapping value is correlated and adjusted with the safety margin level of the typical operating mode represented by the panoramic operating status label to obtain the second mapping value; The second mapping value is restricted to a second preset numerical closed interval, and the restricted value is output as the panoramic operation status index.

[0016] Furthermore, based on the quantitative results of the panoramic operation status, the triggering of coordinated alarms for the power supply side, grid side, load side, and energy storage side includes: The panoramic operation status index is compared with the preset multi-level alarm threshold, and the current alarm level of the power system is determined based on the comparison result; The severity of anomalies in the power supply-side characteristic quantities, grid-side characteristic quantities, load-side characteristic quantities, and energy storage-side characteristic quantities that constitute the panoramic operation status label is analyzed to determine the key anomaly links. Based on the alarm level and the key abnormal link, the corresponding target alarm strategy is matched from the preset collaborative alarm strategy library; According to the target alarm strategy, alarm commands with logical correlation are generated and sent simultaneously to the corresponding links of the target alarm strategy in the power supply side, grid side, load side and energy storage side to trigger the coordinated alarm.

[0017] Accordingly, a second aspect of the present invention provides an electronic device, including: at least one processor; and a memory connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to cause the at least one processor to perform the above-described power system source-grid-load-storage panoramic dispatching operation monitoring and collaborative alarm method.

[0018] Accordingly, a third aspect of the present invention provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the above-described method for panoramic dispatching, operation monitoring, and coordinated alarming of power system source-grid-load-storage systems.

[0019] The above-described technical solutions of the embodiments of the present invention have the following beneficial technical effects: 1. By implementing dynamic weighted adaptive clustering, the problem of difficulty in quantifying and assessing the overall system operation status caused by the simple parallel arrangement of "source, network, load, and storage" operation data in traditional scheduling and monitoring is effectively overcome. It can adapt to different operation stages, deeply integrate multi-dimensional real-time data, and output intuitive panoramic operation status labels and continuous status indices. This achieves a fundamental transformation from listing massive amounts of information to one-click, quantitative perception of the overall system health, greatly improving the speed and accuracy of dispatchers' grasp of the macro-operation status. 2. By establishing a collaborative alarm triggering mechanism based on the quantitative results of panoramic operational status, the "information silo" phenomenon caused by independent alarms in each link in the traditional way has been fundamentally changed. It can intelligently identify key abnormal links and match collaborative strategies based on panoramic status index and tags, and send logically related alarm instructions to multiple related links at the same time. This upgrades alarm information from describing isolated phenomena to revealing cross-link causal relationships and the scope of impact, which significantly improves the information value and decision support capabilities of alarms, and helps dispatchers quickly locate the root cause and initiate collaborative handling. 3. By organically linking panoramic situational awareness, quantitative assessment and intelligent alarms, a complete "perception-cognition-decision" dispatch support closed loop is constructed; it not only provides advanced situational warnings, but also points out the direction of handling through collaborative alarms, thereby transforming the dispatcher's role from passively responding to a large number of isolated alarms to actively managing the panoramic operation of the system; it greatly improves the intelligence level and overall safety resilience of dispatch operations in dealing with the complexity and uncertainty of new power systems. Attached Figure Description

[0020] Figure 1 This is a flowchart of the power system source-grid-load-storage panoramic dispatching operation monitoring and collaborative alarm method provided in the embodiments of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and the accompanying drawings. It should be understood that these descriptions are merely exemplary and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0022] The first aspect of this invention provides a method for panoramic dispatching and operation monitoring and collaborative alarming of power system sources, grid, load, and storage. Its application scenario is a new type of power system dispatching and operation control center with a high proportion of renewable energy access. In this scenario, the volatility, randomness, and multi-stage coupling of system operation are significantly enhanced. Dispatch operations need to be expanded from traditional grid power flow monitoring to panoramic dynamic perception and collaborative decision support of the operational status and interactive relationships of all aspects of "sources, grid, load, and storage."

[0023] Please refer to Figure 1 The above-mentioned method for panoramic dispatching and operation monitoring and coordinated alarming of the power system's power sources, grids, loads, and storage specifically includes the following steps: Step S100: Collect real-time operating data from the power source side, grid side, load side, and energy storage side of the power system.

[0024] In practice, the dispatch automation system synchronously acquires time-synchronized cross-sectional data covering the entire network through the Supervisory Control and Data Acquisition (SCADA) system, Wide Area Measurement System (WAMS), and various management information systems. Power source data mainly includes real-time active / reactive power output, planned output values, short-term power forecasts, and station operating status of each grid-connected power plant (especially wind farms and photovoltaic power plants). Grid-side data mainly includes power flow values ​​and stability limits of key transmission sections, voltage amplitude and phase angle of important buses, and position signals of major switches and disconnectors. Load-side data mainly includes the current status and capacity of total load in each zone, interruptible / adjustable loads, and meteorological information related to load characteristics. Energy storage-side data mainly includes real-time charging and discharging power, state of capacity (SOC), and upper limits of charging and discharging capacity of various energy storage power stations (such as electrochemical energy storage and pumped storage). These real-time operational data collectively constitute the raw data foundation for panoramic operational status analysis.

[0025] Step S200: Perform dynamic weight adaptive clustering on several real-time running data to generate a panoramic running status quantification result. The dynamic weight adaptive clustering is used to adaptively adjust the weights of several real-time running data according to the system running stage and generate panoramic running status labels and panoramic running status indices through cluster analysis.

[0026] First, the system calculates characteristic quantities from the raw data for each side, such as the power output fluctuation rate and power output prediction deviation on the power source side, the power flow margin and voltage stability margin of key sections on the grid side, the demand response potential and probability of uncontrollable load mutations on the load side, and the state of charge and adjustable power on the energy storage side. Next, the system identifies operating phases such as "peak load," "large-scale renewable energy generation," or "normal balance" based on the current total system load and total renewable energy output, and calls upon the corresponding baseline weight configuration for each phase. More importantly, the system dynamically adjusts the weights of each characteristic quantity based on its real-time values. For example, it increases the weight when the power flow margin of key sections decreases sharply, and it increases the weight when the power output fluctuation rate on the power source side increases significantly. This generates a weighted feature vector that sensitively reflects the most pressing constraints and weak points of the current system—the current weighted feature vector. Meanwhile, the system utilizes a large amount of historical operational data to generate a historical weighted feature vector sample library according to the same feature extraction and dynamic weighting rules. This sample library is analyzed using a clustering algorithm to form multiple clusters representing different typical operational modes and their central vectors. Each cluster is assigned a panoramic operational status label with clear physical meaning, such as "stable source and flat load," "fluctuating source and tense load," or "network congestion and saturated storage," thus constructing a status label library. Finally, by matching the current weighted feature vector with the status label library, the most matching status label is output, and the quantized distance between the current weighted feature vector and the corresponding cluster central vector is calculated. Through a preset normalization mapping process, this distance is converted into a continuous scalar value within a preset range (e.g., 0 to 100), namely the panoramic operational status index. This index intuitively represents the degree to which the overall system operational state deviates from its historical typical mode.

[0027] Step S300: Based on the quantitative results of the panoramic operation status, trigger coordinated alarms for the power supply side, grid side, load side and energy storage side.

[0028] In practice, the system first compares the panoramic operational status index with multiple preset thresholds to determine different alarm levels, such as "attention," "early warning," and "emergency." Simultaneously, it analyzes the output panoramic operational status labels, identifying the key anomalies causing the current suboptimal situation by analyzing the severity of anomalies in the characteristic quantities constituting each label. For example, it determines whether grid congestion or power fluctuations are the primary causes. Then, based on the determined "alarm level" and "key anomaly," the system retrieves and matches the corresponding target alarm strategy from a pre-defined collaborative alarm strategy knowledge base. This strategy clearly defines which specific links among the power source, grid, load, and storage should be sent alarm commands under the current specific panoramic operational status, and with what logical relationship. For example, for a situation where "large-scale renewable energy generation leads to section overload," the matched strategy might include: sending a "regulate output" request command to relevant renewable energy power plants, sending a "adjust charging plan or switch to discharge mode" command to energy storage power plants, and issuing a "start demand response preparation" notification to load aggregators. Based on the target alarm strategy, the scheduling system generates and sends a series of alarms and instructions that are closely related in terms of time, target and logic to the monitoring or control systems of the above-mentioned multiple links in a near-synchronous manner, thereby realizing a closed loop from global situation assessment to triggering coordinated actions of multiple links.

[0029] Through the above steps, the isolated data and alarms representing local states in traditional dispatch monitoring are transformed into a quantitative assessment of the overall system operation status based on deep fusion of multi-stage data and adaptive weight analysis, further driving the generation of logically related alarms across stages. This realizes a paradigm shift in dispatch monitoring operations from "parallel information display" to "overall situational awareness," and from "independent reporting of phenomena" to "causal collaborative push" of alarms. It provides dispatchers with intelligent tools covering the entire process of perception, cognition, and decision support, significantly enhancing their dispatch capabilities to cope with the complex operating states of new power systems.

[0030] Furthermore, step S200 involves performing dynamic weight adaptive clustering on several real-time operational data points to generate a panoramic operational status quantification result, including: Step S210: Extract the corresponding operating characteristic quantities from the real-time operating data of the power supply side, grid side, load side and energy storage side respectively.

[0031] The massive and heterogeneous raw operational data is transformed into a standardized set of indicators that can uniformly characterize the core operational status and flexibility of each link. Specifically, for the power supply side, based on the real-time active power output sequence of new energy power sources such as wind farms and photovoltaic power plants, as well as conventional units, the output volatility is calculated. This indicator reflects the short-term instability of power output. At the same time, the deviation between its real-time output and the short-term (e.g., the power prediction deviation) power forecast is calculated to quantify the predictability level of the power supply side. For the grid side, key transmission sections that affect system security are selected, and the difference between their current actual power flow and static stability, thermal stability, and other limits is calculated to obtain the power flow margin of the key sections. The smaller this value, the more strained the transmission channel is. For important buses with weak voltage stability, the minimum difference between their current voltage amplitude and the upper and lower limits of safe operating voltage is calculated as the voltage stability margin. For the load side, the system first aggregates and statistically analyzes the total contracted capacity of various interruptible and adjustable loads with the currently called capacity. The difference between the two represents the current demand response potential, reflecting the scale of resources that the load side can proactively adjust. Furthermore, by combining historical load curves from the same period, real-time meteorological data (such as temperature and humidity), and information on holidays, a statistical learning model is used to assess the probability of unexpected and drastic load fluctuations in the near future, i.e., the probability of uncontrollable load mutations. For the energy storage side, the system directly obtains the current State of Charge (SOC) of the energy storage facilities and calculates the difference between the maximum allowable charging or discharging power under the current operating mode and the current actual power. This difference is used as the adjustable power to characterize the real-time adjustment capability of the energy storage. These features together constitute a standardized feature vector describing the overall operating status of the system.

[0032] Step S220: Based on the system operation stage, dynamically assign weights to several operating feature quantities, and generate the current weighted feature vector based on real-time operating data.

[0033] The identification of system operation phases relies on macroscopic indicators such as total system load and total renewable energy output from real-time operational data. For example, when the total system load remains high while the proportion of renewable energy output is relatively low, it is identified as a "peak load phase"; when the proportion of renewable energy output exceeds a specific threshold and fluctuates significantly, it is identified as a "high-energy-output phase"; other situations are classified as "normal balance phases". Each operation phase has a pre-set set of benchmark weight configurations, which define the initial importance of each characteristic quantity in assessing the overall system status under such typical scenarios. However, the core innovation of this step lies in the dynamic adjustment mechanism of the weights. The system monitors the values ​​of each characteristic quantity in real time. If a certain characteristic quantity shows that the system state is rapidly evolving in an unfavorable direction (for example, the power flow margin at critical sections is rapidly decreasing, or the power output volatility is rising sharply), the system will automatically increase the proportion of that characteristic quantity in the weight allocation, that is, give it a greater weight. This increase is not a fixed value, but is dynamically calculated based on the degree to which the characteristic quantity deviates from its normal or safe benchmark. After dynamically adjusting the weights of all feature quantities, the system uses these finalized weights to perform weighted calculations on the real-time values ​​of eight feature quantities: power supply output volatility, output prediction deviation, grid power flow margin, voltage stability margin, load demand response potential, probability of uncontrollable load mutations, and energy storage state of charge and adjustable power. The eight weighted values ​​are then arranged in a preset order to form an eight-dimensional current weighted feature vector. This vector not only contains the state information of each component but also, through the weights, implies the urgency ranking of the impact of each state on the overall system security at the current moment.

[0034] Step S230: Cluster multiple historical weighted feature vectors formed based on historical data to establish a tag library containing several panoramic operation status labels.

[0035] First, collect historical operational data over a long period (e.g., more than one year). Following the same process as steps S210 and S220, generate a corresponding historical weighted feature vector for each sampling moment in history and record its corresponding system operation stage. Then, for all historical weighted feature vector sets under the same operational stage (e.g., the "new energy development stage"), analyze them using unsupervised clustering algorithms (e.g., K-means, DBSCAN). The clustering algorithm automatically divides a large number of historical samples into several clusters based on the distance between vectors in the feature space. Samples within each cluster have similar operational feature patterns, and a vector representing the center point of that cluster, i.e., the cluster center vector, is calculated. Next, domain experts or automatic analysis based on rules analyze the combination patterns of various feature quantities reflected by each cluster center vector, assigning it an intuitive and easily understood text label with clear physical meaning, i.e., a panoramic operational status label. For example, a cluster center showing low power supply volatility, sufficient grid margin, and low load-side abrupt change probability might be labeled as "stable power supply and smooth grid operation." Conversely, a cluster center showing high power supply volatility, small grid critical section margin, and low adjustable power of energy storage might be labeled as "power supply volatility causing grid strain and insufficient energy storage support." The center vectors of all clusters and their corresponding panoramic operational status labels together constitute a situational knowledge base, or label library, which maps the high-dimensional, abstract feature vector space to a comprehensible, finite set of typical operational scenarios.

[0036] Step S240: Match the current weighted feature vector generated based on real-time operation data with the label library, output the corresponding panoramic operation status label, and calculate the quantized distance between the cluster centers of the current weighted feature vector and the matched panoramic operation status label as the panoramic operation status index.

[0037] In real-time applications, the system calculates the similarity (e.g., Euclidean distance or cosine similarity) between the current weighted feature vector generated in step S220 and all cluster center vectors in the tag library established in step S230 for the same operational phase. By finding the cluster center closest to or with the highest similarity to the current vector, the system can determine which historical typical pattern the current panoramic operational state most belongs to and output the corresponding panoramic operational status label, such as "source wave load tension." This provides the scheduler with a qualitative and intuitive understanding of the current complex situation. Furthermore, the system precisely calculates the specific distance value between the current weighted feature vector and its matched cluster center vector. This distance value quantifies the degree of deviation between the current real-time state and the most similar "typical" historical state. To obtain a standardized and easily understandable metric, this raw distance value is processed by a preset normalization function. For example, based on the statistical distribution (such as mean, standard deviation, or percentile) of the distances from all historical samples within a cluster to its center, the current distance can be mapped to a scalar value within a fixed range (e.g., 0 to 100). This final value is the panoramic operational status index. The closer the index value is to 0, the closer the current state is to the typical historical safe and stable pattern; the higher the index value, the greater the deviation, and the more abnormal or stressful the system operation. This index provides a continuous and quantitative measure of the overall safety margin of the system.

[0038] Furthermore, in step S210, corresponding operational characteristic quantities are extracted from the real-time operational data of the power supply side, grid side, load side, and energy storage side, including: Step S211: Based on the real-time operating data of the power supply side, calculate the power supply side characteristic quantities, including the power output fluctuation rate and the power output prediction deviation.

[0039] Output volatility aims to quantify the instability of total power output over a short timescale. It is typically calculated based on an active power sequence within a rolling time window (e.g., 15 minutes), obtained by calculating the ratio of the sequence's standard deviation to its mean. A larger ratio indicates greater overall power output instability, placing greater pressure on grid frequency regulation and reserve capacity requirements. Output prediction deviation, on the other hand, assesses the predictability of power output. It is calculated as the absolute or relative difference between the current actual total active power output and a prior (e.g., a previously released short- or ultra-short-term power forecast). This deviation directly reflects the prediction accuracy; a larger deviation indicates a greater deviation between actual operation and the expected plan, resulting in greater adjustment pressure on the dispatch plan. In systems with a high proportion of wind and solar power, these two characteristics are core indicators for characterizing the impact of power-side uncertainty on system operation.

[0040] Step S212: Based on the real-time operating data of the power grid side, calculate the characteristic quantities of the power grid side, including the power flow margin and voltage stability margin of the key sections.

[0041] The calculation of power flow margin at critical sections first requires identifying the critical transmission sections that affect system security and power transmission. These sections are usually predefined by the dispatching agency based on grid structure, stability analysis, and operational experience. The margin is calculated as the difference between the current actual transmission power (power flow) at the section and the minimum of its various stability limits (such as static safety limits, transient stability limits, or thermal stability limits). This difference reflects the available transmission capacity of the transmission channel; a smaller or negative value indicates a higher risk of exceeding the limit, directly constraining the safe operation of the grid. Voltage stability margin, on the other hand, is calculated for voltage-weak nodes or areas as the shortest distance between the real-time voltage amplitude of the critical bus (or node) and the upper and lower limits of safe voltage operation determined according to safety guidelines. Specifically, it is the smaller of the voltage amplitude deviation from the upper and lower limits. This margin characterizes the safe distance of the current voltage level from the limit boundary and is an important indicator for assessing voltage stability.

[0042] Step S213: Based on the real-time operating data of the load side, calculate the load side characteristic quantities, including demand response potential and uncontrollable load mutation probability.

[0043] Demand response potential refers to the total capacity of the load side that can be called upon or incentivized by the dispatch center to participate in grid regulation at the current moment. Its calculation is typically based on information such as contracted capacity, availability status, and current response volume reported by various demand response resources (e.g., interruptible loads, adjustable loads, electric vehicle aggregators, etc.). It is obtained by aggregating interruptible capacity and subtracting used capacity from the current maximum adjustable capacity. This potential value reflects the scale of the load side's ability to participate in balance regulation as a flexible resource. Uncontrollable load mutation probability, on the other hand, is used to characterize the extreme risks of natural load fluctuations. By analyzing the load curve fluctuation characteristics of the same historical period (similar date types, similar weather conditions), combined with real-time weather forecast data (e.g., forecasts of drastic temperature changes), and information on special events, statistical models or probability distribution fitting are used to estimate the probability of unexpected and significant load fluctuations in the near future (next dispatch period). The higher the probability value, the greater the balance uncertainty brought to the system by the load side.

[0044] Step S214: Based on the real-time operating data of the energy storage side, calculate the characteristic quantities of the energy storage side, including the state of charge and adjustable power.

[0045] State of Charge (SOC) is a direct indicator of the current energy reserve level of an energy storage device, calculated as the percentage of the electrical energy currently stored to its rated total capacity. This state not only determines the continuous charging and discharging capability of the energy storage but is also closely related to its operating strategy and lifespan. For example, excessively high or low SOC can limit its regulation capabilities or accelerate device aging. Adjustable power characterizes the immediate ability of an energy storage device to further increase or decrease its power at the current operating point (current charging / discharging power and SOC state). It is calculated as the difference between the maximum allowable power limit of the energy storage device under the current permissible operating mode (charging, discharging, or standby) and the current actual operating power. This difference reflects the upward or downward adjustment power range that the energy storage device can provide in the short term and is a key parameter for assessing its ability to quickly respond to system fluctuations.

[0046] Step S215: Based on the power supply side characteristic quantity, grid side characteristic quantity, load side characteristic quantity and energy storage side characteristic quantity, the operation characteristic quantity is obtained by combination.

[0047] The eight specific characteristic indicators (power generation side: output volatility, output prediction deviation; grid side: power flow margin at key sections, voltage stability margin; load side: demand response potential, probability of uncontrollable load mutations; energy storage side: state of charge, adjustable power) calculated independently from the four stages of power generation, grid, load, and energy storage are arranged and combined in a predetermined and unified order (e.g., power generation side first, then grid side, then load side, and finally energy storage side) to form a fixed-dimensional, structured numerical vector, namely the "operational characteristic quantity" vector. This vector constitutes a standardized and digital profile description of the panoramic operation status of the power system from "power generation-grid-load-storage". The diverse and heterogeneous data, originally scattered across different professional systems with different physical units and dimensions, are normalized into a comprehensive data object that can be used for subsequent unified mathematical processing (such as weight allocation, vector comparison, and cluster analysis), laying the data structure foundation for quantitatively assessing the system's operational status from a holistic perspective.

[0048] Further, in step S220, weights are dynamically assigned to several operational features based on the system's operational phase, and a current weighted feature vector is generated based on real-time operational data, including: Step S221: Based on the total system load and total renewable energy output in the real-time operating data, identify the current system operation stage. The system operation stages include the peak load stage, the renewable energy generation stage, and the normal balance stage.

[0049] The division of system operation phases is based on the dominant operating characteristics and main contradictions exhibited by the power system at different time scales. The specific identification logic is as follows: continuously monitor and calculate the real-time values ​​and proportional relationships between the total active power load and the total output of new energy sources (mainly wind and solar power). When the total system load is consistently at or above a specific percentage threshold (e.g., 85%) of the daily maximum load forecast curve, and the proportion of total new energy output is relatively low, the system is determined to be in the "peak load phase." The main contradiction in this phase lies in balancing the peak-shaving capacity of conventional power sources with load demand. When the proportion of total new energy output to the total system load exceeds a preset high threshold (e.g., 50%), and its power fluctuates significantly in a short period, the system is determined to be in the "large-scale new energy generation phase." The core challenge in this phase is the grid's ability to absorb fluctuating power sources and the resulting potential frequency and voltage stability issues. When the system operation status does not meet the criteria for either the peak load phase or the large-scale new energy generation phase, it is classified as the "normal balance phase." In this phase, the system operates relatively smoothly, and the adjustment margins of each component are usually ample. This stage identification provides context for subsequent differentiated weight configuration.

[0050] Step S222: Based on the system operation stage, obtain the preset benchmark weight configuration. The benchmark weight configuration includes the initial weight values ​​of the power source side characteristic quantity, grid side characteristic quantity, load side characteristic quantity and energy storage side characteristic quantity respectively.

[0051] The baseline weighting configuration is based on a predefined set of data tables or rules derived from historical analysis and expert experience of key influencing factors for system safety operation at each operational stage. For the "peak load phase," the baseline configuration typically assigns relatively high initial weights to grid-side characteristics (especially power flow margin at key sections) and power source characteristics (reflecting peak-shaving capacity), as grid transmission capacity constraints and power source output limits are the primary concerns at this stage. For the "peak renewable energy generation phase," the baseline configuration significantly increases the initial weights of power source output volatility, output prediction deviation, and grid-side voltage stability margin to emphasize the importance of renewable energy volatility and grid connection point voltage control. For the "normal balance phase," the weighting in the baseline configuration is relatively balanced, or slightly biased towards characteristics reflecting the overall system's regulatory flexibility, such as adjustable power on the energy storage side and demand response potential on the load side. These preset initial weight values ​​constitute the baseline starting point for dynamic weight adjustment, ensuring that the initial bias of the assessment under different operating scenarios aligns with operational experience.

[0052] Step S223: Based on the real-time values ​​of the power source side characteristic quantity, grid side characteristic quantity, load side characteristic quantity and energy storage side characteristic quantity, dynamically and collaboratively correct the initial weight value in the benchmark weight configuration to generate the final weight of each characteristic quantity.

[0053] The system monitors the specific values ​​of the eight characteristic quantities calculated in steps S211 to S214 in real time. The core of the correction logic is that when the real-time value of a certain characteristic quantity indicates that the system link it represents is deteriorating or approaching a safety boundary, the system will automatically increase the weight of that characteristic quantity, increasing its influence in the overall assessment. For example, if the real-time calculated critical section power flow margin value drops rapidly, approaching zero or even the alarm threshold, the system will dynamically increase the weight allocated to the "critical section power flow margin" characteristic quantity based on the baseline weight, regardless of the current operating stage. Similarly, if the power supply side output fluctuation rate rises sharply, or the probability of uncontrollable load mutations on the load side increases significantly, their corresponding weights will also be increased. This correction is "coordinated," meaning the system may simultaneously adjust multiple weights in conjunction with the abnormal conditions of multiple characteristic quantities. The magnitude of the weight increase is usually positively correlated with the degree to which the characteristic quantity deviates from its normal reference value, and is calculated using a predefined correction function (such as a linear function, exponential function, etc.). Ultimately, each feature quantity obtains a "final weight" that has been calibrated in real time. This set of weights dynamically reflects the urgency ranking of the impact of the current operational status on the overall security posture of the system.

[0054] Step S224: Use the final weights to weight the values ​​of the power source side characteristic quantity, grid side characteristic quantity, load side characteristic quantity and energy storage side characteristic quantity respectively, and combine all the weighted values ​​into a vector in a predetermined order to generate the current weighted feature vector.

[0055] In this step, the system multiplies the "final weight" of each feature calculated in step S223 with the corresponding feature value in the original "operating feature quantity" vector formed in step S215. Specifically, the power supply output fluctuation rate is multiplied by its final weight, the output prediction deviation is multiplied by its final weight, and so on, until all eight features have been weighted. The weighted values ​​have enhanced physical meaning: the magnitude of the value not only reflects the absolute level of the feature itself, but also, through the amplification or reduction of the weight, reflects the relative importance and urgency of the state in the current global context. Subsequently, these eight weighted values ​​are arranged and concatenated in the same predetermined order as the original operating feature quantity vector (e.g., source first, then grid, reload, and finally storage) to form a new numerical vector, namely the "current weighted feature vector". This vector is the product of the original operating status data after dual calibration with context (operation phase) and real-time status. It encapsulates the core information of the current power system's panoramic operating status, and the magnitude relationship of the data in each dimension has already included the system's quantitative judgment on the current main contradictions and risk points, providing accurate input for subsequent clustering matching and situation index calculation.

[0056] Furthermore, in step S223, based on the real-time values ​​of power source-side characteristics, grid-side characteristics, load-side characteristics, and energy storage-side characteristics, the initial weight values ​​in the benchmark weight configuration are dynamically and collaboratively corrected to generate the final weight for each characteristic, including: Step S2231: Calculate the deviation of the real-time value of the power output fluctuation rate in the power supply side characteristic quantity from its preset standard value, and use it as the power supply side fluctuation deviation.

[0057] The deviation of the real-time output volatility value from its preset standard value is calculated as the power supply side volatility deviation. The preset standard value is an acceptable benchmark level of power output volatility based on historical statistics under specific operating phases (such as periods of high renewable energy generation). This benchmark can be determined, for example, by calculating the average or a specific quantile of the output volatility under similar weather conditions during the same historical period. The calculation of the power supply side volatility deviation aims to quantify the extent to which the current actual volatility level exceeds this reasonable benchmark. Specifically, this deviation is usually calculated as the difference between the current real-time output volatility value and the preset standard value, divided by the preset standard value to obtain the relative ratio; or it can be expressed as a piecewise function, where the deviation is zero when the volatility does not exceed the standard value, and increases proportionally or non-linearly after exceeding it. In operating scenarios with high renewable energy penetration, if wind and solar resources experience drastic changes, the real-time output volatility may increase sharply, leading to a significant increase in the calculated power supply side volatility deviation. This directly indicates that the power supply side is exerting extraordinary volatility pressure on the system.

[0058] Step S2232: Calculate the negative deviation of the real-time value of the power flow margin of the key section in the power grid side characteristic quantity relative to its preset safety threshold, and use it as the power grid congestion deviation.

[0059] The grid congestion deviation is calculated as the negative deviation of the real-time power flow margin of key sections in the grid-side characteristic quantities relative to its preset safety threshold. The preset safety threshold is a margin threshold greater than zero, marking the boundary between a section's operation and a state requiring vigilance. Its setting typically considers operating procedures, reserved adjustment space, and transient stability requirements. The grid congestion deviation specifically measures the severity of the power flow margin falling below this safety threshold, focusing on negative deviation. A common calculation method is to define the grid congestion deviation as zero when the real-time power flow margin is greater than or equal to the safety threshold; when the real-time margin is less than the safety threshold, the difference between the safety threshold and the real-time margin is calculated and divided by the safety threshold or a reference value to obtain a positive index characterizing the severity of congestion. When a transmission section approaches its stability limit due to load growth, concentrated power output, or network topology changes, the real-time power flow margin decreases, and the grid congestion deviation increases from zero. This index keenly captures the risk level of grid-side transmission capacity being limited and constituting a major bottleneck in system operation.

[0060] Step S2233: Calculate the growth deviation of the real-time value of the uncontrollable load mutation probability in the load-side characteristic quantity relative to its historical baseline value, and use it as the load mutation deviation.

[0061] The load mutation deviation is calculated as the deviation of the real-time value of the probability of uncontrollable load mutations relative to its historical baseline value. The historical baseline value represents the average statistical level of the probability of unexpected and drastic load mutations under similar date types, similar time periods, and similar meteorological conditions, and can be obtained by analyzing long-term historical data. The load mutation deviation aims to measure the magnitude of the abnormal increase in the currently assessed mutation probability. It is usually calculated by dividing the difference between the currently calculated real-time probability of uncontrollable load mutations and the corresponding historical baseline value by the historical baseline value to obtain a relative growth rate. For example, when encountering extreme weather warnings or sudden social events, the load mutation probability assessed based on real-time meteorological models and event information may far exceed the normal level for the same period in history, resulting in a significant increase in the calculated load mutation deviation. This indicator quantifies the abnormal increase in load-side uncertainty and reflects the potential risk of the system facing unplanned and drastic load fluctuations.

[0062] Step S2234: Calculate the deviation of the real-time value of the state of charge in the energy storage side characteristic quantity from the boundary of its preset optimal range, and use it as the energy storage state deviation.

[0063] The deviation of the real-time value of the state of charge (SOC) among the energy storage-side characteristic quantities from the boundary of its preset optimal range is calculated as the energy storage state deviation. The preset optimal range is a recommended operating range of SOC set to ensure the efficient and safe operation of energy storage facilities and extend their lifespan, for example, set between 20% and 80% of the rated capacity. The energy storage state deviation measures the degree to which the current SOC deviates from this optimal range. During calculation, it is first determined whether the current SOC value is within the optimal range. If it is within the range, the deviation can be set to zero or a very small value; if it is below the lower limit of the range, the difference between the lower limit of the range and the current SOC is calculated, and the deviation is expressed as the ratio of this difference to the range width or other normalization methods; if it is above the upper limit of the range, the difference between the current SOC and the upper limit of the range is calculated and similar processing is performed. The larger the deviation, the further the energy state of the energy storage device deviates from the ideal operating point, and its usable depth of charge and discharge may be limited, or the risk of overcharging and over-discharging increases, thereby weakening its real-time adjustment capability and reliability as a system flexibility resource.

[0064] Step S2235: Based on the power supply side fluctuation deviation, grid congestion deviation, load change deviation, and energy storage state deviation, the weight correction coefficients corresponding to the power supply side characteristic quantity, grid side characteristic quantity, load side characteristic quantity, and energy storage side characteristic quantity are calculated respectively through a preset collaborative correction function.

[0065] Based on the deviations of power supply fluctuations, grid congestion, load abrupt changes, and energy storage status, a pre-defined collaborative correction function is used to calculate the corresponding weight correction coefficients for each of the power supply-side, grid-side, load-side, and energy storage-side characteristic quantities. The pre-defined collaborative correction function is a multi-input, multi-output mapping rule or computational model. Its core design principle is that an increase in the deviation of a certain link should not only lead to an increase in the weight correction coefficients of its own related characteristic quantities but may also affect the weight adjustments of characteristic quantities in other related links, reflecting the coupling influence between the states of various links in the system. For example, this function can be designed to calculate the correction coefficient for each characteristic quantity based on a weighted sum of each deviation or a more complex nonlinear combination (such as considering interaction terms in the product of deviations). Specifically, for power source-side characteristic quantities (output fluctuation rate, output prediction deviation), their weight correction coefficients are mainly positively affected by the power source-side fluctuation deviation. For grid-side characteristic quantities (critical section power flow margin, voltage stability margin), their weight correction coefficients are mainly positively affected by the grid congestion deviation, and may also be slightly affected by the power source-side fluctuation deviation (because power fluctuations may trigger power flow changes). For load-side and energy storage-side characteristic quantities, their correction coefficients are mainly driven by load mutation deviation and energy storage state deviation, respectively. This function ensures that the final weight adjustment is systematic, reflecting the comprehensive urgency of abnormal states in each link from a panoramic perspective.

[0066] Step S2236: Multiply the initial weight value of each feature by the corresponding weight correction coefficient to obtain the final weight of the corresponding feature.

[0067] The initial weight value of each feature is multiplied by the corresponding weight correction coefficient to obtain the final weight of the corresponding feature. This step performs specific weight value adjustments. The system reads the initial weight value of each feature (a total of eight) from the baseline weight configuration obtained in step S222. Then, a scalar multiplication operation is performed between the corresponding weight correction coefficient calculated in step S2235 (this coefficient is usually a value greater than or equal to 1; when the corresponding deviation is zero or very small, the coefficient may be 1 or slightly higher than 1, indicating no correction or slight correction) and the corresponding initial weight value. For example, if the initial weight corresponding to the power flow margin of a certain critical section is 0.15, and the calculated weight correction coefficient is 1.5 (due to the large deviation of the power grid congestion), then its final weight is updated to 0.225. Through this multiplication operation, the weight of each feature is amplified according to the real-time abnormal deviation of its associated link; the greater the deviation of the link, the higher the proportion of amplification of the weight of its related feature. Ultimately, all eight features received updated, dynamically adjusted final weight values. This set of weight values ​​accurately reflects the relative importance of each operating feature in assessing the overall system status at the current moment.

[0068] Furthermore, in step S230, multiple historical weighted feature vectors formed based on historical data are clustered to establish a tag library containing several panoramic operational status labels, including: Step S231: Obtain multiple historical weighted feature vectors stored according to the system operation stage within the historical time period, with each historical weighted feature vector corresponding to a specific historical moment.

[0069] This step is the data preparation stage for building the tag library. It extracts a complete dataset from a long-running historical database, containing all recorded time points (e.g., every 5 or 15 minutes) from a past period (e.g., one year or several years). For each historical time point, the system retrospectively calculates the corresponding "historical weighted feature vector" following the same logical flow as real-time processing (i.e., steps S210 to S220). This means that for any historical time point, an eight-dimensional vector is generated, where each dimension represents the value of eight characteristic quantities, such as power supply output fluctuation rate and output prediction deviation, corresponding to that time point. These values ​​have undergone dynamic weighting and processing based on the identified operating stage and real-time status at that historical time point. Subsequently, the system categorizes and stores all historical weighted feature vectors into three different subsets according to the identified operating stage (peak load, large-scale renewable energy generation, or normal balance stage) for each historical time point. This categorized storage by operational phase is crucial because it ensures that subsequent cluster analysis is conducted within datasets with similar dominant operational characteristics, avoiding interference between data from different operational modes, and thus enabling a clearer identification of different typical sub-patterns within the same phase.

[0070] Step S232: Use a clustering algorithm to perform cluster analysis on the set of historical weighted feature vectors belonging to the same system operation stage, divide the feature space into several clusters, and determine the center vector of each cluster.

[0071] This step is the core application of unsupervised machine learning, independently applying a clustering algorithm to each subset of operational stages obtained in step S231 (e.g., all historical vectors in the "New Energy Development Stage"). Taking the K-means algorithm as an example, the algorithm automatically finds and iteratively optimizes a set of center points (i.e., cluster center vectors) in the eight-dimensional feature space, minimizing the sum of the distances from all historical vectors to their nearest center point. After the algorithm finishes running, hundreds or thousands of historical vectors belonging to the same operational stage are divided into a finite number (e.g., K) of clusters. All historical vectors within the same cluster are close to each other in the feature space, meaning that the overall operational state of the system they represent is highly similar. The center vector of each cluster is the mean vector of all vectors within that cluster, mathematically best representing the common state pattern of that cluster. For example, in the subset of the "peak of renewable energy generation," clustering algorithms might divide the data into several typical clusters: one cluster's center vector shows low power supply volatility and sufficient grid margin, representing "stable and abundant renewable energy generation"; another cluster's center vector shows high power supply volatility and tight grid margins in some sections, representing "renewable energy fluctuations causing localized congestion"; and yet another cluster's center vector might show extremely low adjustable energy storage power, representing "abundant renewable energy generation but depleted energy storage regulation capacity." In this way, the complex continuous historical state space is discretized into several representative typical pattern points (center vectors).

[0072] Step S233: For each cluster, based on the magnitude of the characteristic components in its central vector that represent the power supply side fluctuation level, grid side congestion level, load side sudden change risk and energy storage side support capacity, assign a panoramic operation status label to the cluster according to the preset label generation rules.

[0073] The analysis focuses on the central vector of each cluster, specifically examining the magnitudes of its eight components. For example, it determines whether the power supply volatility component exceeds the "high volatility" threshold, whether the grid-side critical section power flow margin component is below the "alert" threshold, whether the load-side uncontrollable load mutation probability component is above the "high risk" threshold, and whether the energy storage-side state of charge component is far from the midpoint of the optimal range. Based on the combination of these components, the rule base matches the most appropriate text description. For instance, a rule might specify: if power supply volatility is "high" and grid power flow margin is "tight," the label is "source volatility causing grid tension"; if power supply volatility is "medium," load mutation probability is "high," and energy storage adjustable power is "low," the label is "source and load dual disturbances with insufficient energy storage support"; if all components are within the "relaxed" or "normal" range, the label is "stable source, smooth grid, and balanced load." These labels, such as "source volatility and load tension" and "grid congestion and energy storage saturation," concisely summarize the core characteristics of the panoramic operation scenario represented by the cluster, enabling dispatchers to understand them intuitively.

[0074] Step S234: Establish the mapping relationship between the center vector of the cluster and the corresponding panoramic operation status label. The label library is composed of the center vector of all clusters and their mapped panoramic operation status labels.

[0075] Each cluster center vector (an eight-dimensional numerical array) calculated in step S232 is used as the "key," and the "panoramic operation status label" (a string) assigned to it in step S233 is used as the "value," forming a one-to-one key-value pair. These "center vector-label" pairs under all operation stages are systematically organized together to form a database or knowledge base that can be efficiently retrieved, namely, a "label library." For example, the label library may contain dozens of such entries, covering all the typical patterns identified under each stage, such as "peak load," "new energy power generation," and "routine balance." The physical essence of this label library is a lookup table or classifier that maps typical state points (center vectors) in a high-dimensional continuous feature space to discrete, semantic operation scenario categories (labels).

[0076] Further, the calculation of the quantized distance between the current weighted feature vector and the cluster centers of the matched panoramic operational situation labels in step S240, as the panoramic operational situation index, includes: Step S241: Obtain the cluster center vector corresponding to the panoramic operation status label.

[0077] In real-time operation, once the matching process in step S240 determines that the current state best belongs to a specific panoramic operational status label (e.g., "source fluctuations causing grid tension"), the system immediately retrieves the cluster center vector uniquely mapped to that label from a pre-built label library. This cluster center vector is an eight-dimensional numerical vector representing the average or core feature pattern of all historical operational states categorized under that label in historical data. For example, for the label "source fluctuations causing grid tension," each component in its corresponding cluster center vector characterizes the typical level of power source volatility, the typical tension level of power flow margin at key grid sections, and typical values ​​of other characteristics under similar historical scenarios. Obtaining this vector provides a benchmark for subsequent quantitative comparisons and serves as a reference origin for measuring the degree to which the current real-time state deviates from this typical pattern.

[0078] Step S242: Based on all historical weighted feature vectors belonging to the cluster center vector, calculate the statistical distribution of the distance between them and the cluster center vector, and determine the distance benchmark used to measure the degree of tension based on the statistical distribution.

[0079] To scientifically assess the degree of deviation, a reasonable scale is needed. From historical data, all historical weighted feature vector samples that were assigned to the desired cluster during clustering are extracted. The Euclidean distance from each historical vector to the cluster center vector is calculated, resulting in a set of historical distances. The statistical properties of this set are analyzed, such as calculating its mean (μ) and standard deviation (σ). The distance benchmark can be determined based on this statistical distribution. One embodiment is to set the distance benchmark as "μ + nσ" (where n is a coefficient set according to the sensitivity requirement for anomalies, such as 2 or 3). Physically, this means that in historical similar patterns, the distance from most (e.g., over 95%) sample points to the center point is less than this benchmark value. This benchmark value thus defines a boundary of a historically empirical "normal fluctuation range." Another embodiment is to select a high percentile (e.g., the 95th percentile) of this historical distance set as the benchmark. Establishing this distance benchmark makes subsequent assessments of real-time deviations no longer a simple judgment of absolute distance, but a relative assessment relative to the historical normal fluctuation range of the pattern itself.

[0080] Step S243: Calculate the weighted Euclidean distance between the current weighted feature vector and the cluster center vector.

[0081] To more accurately reflect the differences in the impact of different feature dimensions on the current situation, this step uses weighted Euclidean distance instead of ordinary Euclidean distance. The formula for calculating weighted Euclidean distance is, in principle, to square the difference between the current vector and the center vector in each dimension, but before squaring, the difference in each dimension must be multiplied by a weight coefficient corresponding to that dimension. This weight coefficient is related to the "final weight" used in step S223 when generating the current weighted feature vector. A direct implementation is to use the same weight vector. This means that when calculating the distance, the differences in feature dimensions that were deemed more important (higher weight) when generating the current vector will contribute more to the total distance. For example, if the current grid congestion deviation is large, resulting in a high weight for the "critical section power flow margin" feature, then the difference between the current vector and the center vector in the "power flow margin" dimension will be significantly amplified in the distance calculation. The distance calculated in this way not only measures the difference in the overall state but also reinforces the differences in the core security dimensions that are of most concern in the current specific situation, making the distance measurement more relevant to the tension of the situation.

[0082] Step S244: The calculated weighted Euclidean distance is compared with the distance benchmark. Through a preset normalization mapping function, the weighted Euclidean distance is converted into a continuous scalar within a preset numerical range to obtain a panoramic operating status index that characterizes the degree to which the overall operating status of the system deviates from its typical mode.

[0083] First, the system compares the weighted Euclidean distance (denoted as d) calculated in step S243 with the distance reference (denoted as D) determined in step S242. ref This is compared to obtain a relative ratio or relationship. A pre-defined normalized mapping function maps the distance d to a fixed, easily understood interval, such as 0 to 100, based on this relationship. This function is typically designed to be non-linear to reflect the sensitivity differences between different levels of stress. A typical mapping method is: when d is much smaller than D... ref When d approaches D, the mapping function outputs a low value close to 0, indicating a normal state; when d approaches D... ref When d equals or exceeds D, the output value begins to increase rapidly and non-linearly, entering the warning range (e.g., 40-70); ref When the output value reaches a high value, it enters the alarm range (e.g., 70-100). In this way, an abstract, multidimensional distance value is transformed into a unified, monotonically increasing scalar exponent. The larger the exponent value, the further the current real-time operating state deviates from its best-matching typical historical safety mode, and the more tense and abnormal the overall system operating situation.

[0084] Further, step S242, which involves calculating the statistical distribution of the distance between the vector and the cluster center vector, and determining a distance benchmark for measuring the degree of tension based on the statistical distribution, includes: Step S2421: Calculate the set of historical distances between all historical weighted feature vectors belonging to the cluster center vector and the cluster center vector, and calculate its statistical mean and standard deviation based on the set of historical distances.

[0085] First, all historical weighted feature vector samples that were assigned to the specified cluster during clustering are extracted from the historical database. For each sample vector, its Euclidean distance to the cluster center vector is calculated, resulting in a set of distances for all sample points—the historical distance set. Then, basic statistical moment calculations are performed on this set to obtain its arithmetic mean and standard deviation. The mean reflects the typical average distance of each sample point from its central pattern under similar historical conditions, representing the "normal fluctuation level" of the pattern. The standard deviation measures the dispersion of these distance values; a larger standard deviation indicates greater historical variability in similar operating conditions and a wider fluctuation range. These two statistical parameters together characterize the naturally occurring dispersion of this specific operating pattern (such as "local congestion caused by new energy fluctuations") in history, providing a data-driven objective basis for subsequently defining what constitutes "abnormal deviation."

[0086] Step S2422: Based on the numerical distribution characteristics of the historical distance set, select historical distance values ​​from the historical distance set that represent significant anomalies in the operating status within the cluster, and use them as distance references to represent typical abnormal fluctuations within the cluster.

[0087] Relying solely on the mean and standard deviation may be insufficient to capture "stressful" states that, while belonging to the same historical pattern, are approaching the safety margin. Therefore, a thorough analysis of the complete distribution of historical distance sets is necessary to identify and extract thresholds representing significant abnormal fluctuations. Specifically, the cumulative distribution or percentile information of this distance set is systematically analyzed. One implementation involves calculating a high percentile of the set, such as the 95th or 99th percentile. This means that among all historical states belonging to this pattern, 95% or 99% of the states are less than this percentile value in distance from the center point. This high percentile value can serve as a distance reference value for "typical abnormal fluctuations." These correspond to situations that are relatively rare but have occurred historically in the operation of this type of pattern, where states deviate significantly from the center. These situations are often associated with higher operational risks or more intense regulatory pressures. Using this value as a reference aligns the sensitivity of the assessment with real historical risk cases, making the benchmark setting more meaningful for practical early warning.

[0088] Step S2423: Adjust the preliminary benchmark value determined based on the statistical mean and distance reference by combining the preset importance level of the system operation mode represented by the panoramic operation status label corresponding to the cluster.

[0089] Different panoramic operational status labels imply different levels of system risk. For example, the "stable source and smooth network" label represents a low-risk, stable mode, while the "network congestion and storage saturation" label represents a high-risk, tense mode, and a "priority level" coefficient is preset for each type of label. First, based on the mean of step S2421 and the distance reference (such as the high percentile value) of step S2422, a preliminary distance benchmark value is formed through weighted averaging or other composite methods. Then, this preliminary benchmark value is scaled and adjusted according to the importance level of the currently matched label. For high-risk level labels (such as "network congestion and storage saturation"), the adjustment coefficient is usually less than 1, thus lowering the distance benchmark value. This means that the tolerance for state fluctuations under this type of tense mode is lower, and even small state deviations may trigger a higher situation index to achieve earlier and more stringent warnings. Conversely, for low-risk level labels, the adjustment coefficient may be greater than or equal to 1, appropriately relaxing the benchmark. This adjustment reflects a differentiated security strategy, ensuring more sensitive monitoring of key vulnerability modes.

[0090] Step S2424: Based on the different requirements for operational stability at the current system operation stage, the distance reference value after importance adjustment is finally calibrated to generate a distance reference for current situation assessment.

[0091] The system's operational phases inherently present different stability and control requirements. For example, during peak load periods, grid equipment operates close to its limits, exhibiting weaker tolerance to state fluctuations and demanding higher monitoring sensitivity; the "normal balance phase," on the other hand, is relatively more lenient. Therefore, after obtaining the baseline value adjusted for tag importance, the system needs to perform a final calibration based on the currently identified operational phase (peak load, renewable energy generation, or normal balance). The system presets a calibration factor for each operational phase. During peak load periods or renewable energy generation phases, the calibration factor may be set to a value less than 1, further tightening the distance baseline and making the evaluation criteria more stringent to match the higher safety sensitivity of this phase. During the "normal balance phase," the calibration factor may be 1 or slightly greater than 1, maintaining or slightly relaxing the criteria. After this round of calibration based on operational phases, a dynamic distance baseline is ultimately generated that fully adapts to the current operational context (considering historical pattern characteristics, pattern risk levels, and current phase requirements). This baseline value is no longer static but intelligently adjusted according to the matched situation tags and the current operational phase, enabling subsequent situation index calculations to achieve true context-aware assessment.

[0092] Further, in step S244, the calculated weighted Euclidean distance is compared with a distance benchmark. Using a preset normalization mapping function, the weighted Euclidean distance is converted into a continuous scalar within a preset numerical range, resulting in a panoramic operational status index characterizing the degree to which the overall system operating state deviates from its typical pattern. This index includes: Step S2441: Compare the weighted Euclidean distance with the distance reference to determine the relative deviation of the weighted Euclidean distance from the distance reference.

[0093] By establishing a dimensionless ratio, the system measures the degree of deviation from the real-time state relative to historical experience benchmarks, either as a multiple or a proportion. In practice, the system calculates the current weighted Euclidean distance (denoted as d). current ) and the dynamic distance reference (denoted as D) generated in step S2424 base The ratio of r to d, i.e., r = d current / D base The relative deviation *r* is a key mediating variable. When *r* < 1, it indicates that the distance from the current state to the center of the typical pattern is still less than the baseline set by historical experience, and is within the normal fluctuation range of the pattern; when *r* = 1, it indicates that the current distance has just reached the baseline; when *r* > 1, it indicates that the current deviation has exceeded the normal fluctuation range of history and entered an abnormal or tense region. This ratio unifies distances with different physical dimensions and absolute values ​​to the baseline on a comparable relative scale, laying the foundation for subsequent standardized mapping.

[0094] Step S2442: According to different numerical ranges of the relative deviation degree, map the weighted Euclidean distance to a first mapped value within a first preset numerical range through a piecewise linear mapping function.

[0095] Convert the relative deviation degree r into a preliminary and fixed-range score. The preset piecewise linear mapping function defines the linear conversion relationship between different r value ranges and the first preset numerical range (such as 0 to 100). Generally, this function is designed to be monotonically increasing. For example, it can be defined that when r ≤ r1 (such as r1 = 0.8), the mapping function outputs a relatively low and slowly increasing value, indicating normal status; when r1 < r ≤ r2 (such as r2 = 1.0), the slope of the mapping function increases, and the output value increases rapidly to the median value, indicating that the status requires attention; when r > r2, the mapping function adopts a larger slope, making the output value quickly approach the upper limit, indicating a tense or abnormal status. Through this piecewise linear mapping, not only is the monotonic relationship between r and the output value (the first mapped value) ensured, but different sensitivities can also be assigned to different regions of different tension levels, enabling the index to produce more significant changes when approaching and exceeding the benchmark value, thereby providing a clearer warning gradient.

[0096] Step S2443: Correlate and adjust the first mapped value with the safety margin level of the typical operation mode characterized by the panoramic operation situation label to obtain a second mapped value.

[0097] Different panoramic operation situation labels essentially represent operation modes of the system at different safety levels. For example, the mode corresponding to the "stable source and smooth network" label has a high safety margin, while the mode corresponding to the "network congestion and storage saturation" label has a naturally tight margin. Therefore, it is necessary to correct the initially obtained first mapped value based on the inherent risk of the mode. The system presets a "safety margin level" coefficient kk (usually 0 < k ≤ 1) for each panoramic operation situation label. For labels with high safety risks and low inherent margins (such as "network congestion and storage saturation"), the k value is set smaller (such as 0.8); for labels with low risks and high margins, the k value is set larger (such as 1.0 or 1.1). The adjustment operation is usually to multiply the first mapped value by this coefficient k to obtain the second mapped value. For high-risk labels, this multiplication operation will reduce the second mapped value (when k < 1), which does not weaken the warning, but requires stricter requirements when comparing with the unified threshold in subsequent steps. The logic is that for a mode that is already very tense, even if the status only deteriorates slightly, its relative risk is higher, so it is necessary to adjust it to make its index more likely to enter the high-value warning range. On the contrary, for safe modes, it is appropriately relaxed. This enables the index to reflect the superimposed risk concept of "on what basis has one advanced one step".

[0098] Step S2444: The second mapping value is restricted to a second preset numerical closed interval, and the restricted value is output as the panoramic operation status index.

[0099] To ensure the standardization and consistency of the output indicators, regardless of the second mapping value obtained through the aforementioned calculations, it needs to be constrained within a unified, predefined closed interval, for example, forcibly limited to [0, 100]. This operation is achieved through a limiting function: if the second mapping value is less than the lower limit of the interval (e.g., 0), output 0; if it is greater than the upper limit of the interval (e.g., 100), output 100; if it is within the interval, output the original value. The final value after this step is the "Panoramic Operational Status Index." This index is a continuous scalar between a fixed minimum and a maximum value. The closer the value is to the lower limit, the closer the overall system operation status is to its matched typical safety mode, and the more stable the situation; the closer the value is to the upper limit, the further the system status deviates from the typical mode, and the more tense and risky the situation is assessed after considering the risk characteristics of the mode itself. This standardized index provides dispatchers with an intuitive, unified, and comparable risk measurement, directly supporting the setting of alarm thresholds for different levels and the triggering of collaborative alarms.

[0100] Furthermore, the quantitative results based on the panoramic operational status in step S300 trigger coordinated alarms for the power supply side, grid side, load side, and energy storage side, including: Step S310: Compare the panoramic operation status index with the preset multi-level alarm threshold, and determine the current alarm level of the power system based on the comparison result.

[0101] The preset multi-level alarm thresholds are critical values ​​that divide the continuous panoramic operational status index into several discrete alarm levels based on scheduling operation procedures, system security risk tolerance, and historical event analysis. For example, three levels of thresholds can be set: when the index is below the first threshold (e.g., 40), the system is in a "normal" state with no alarms; when the index is between the first and second thresholds (e.g., 70), a "warning" level alarm is triggered, prompting the dispatcher that the system's operational status is becoming tense, requiring enhanced monitoring and contingency planning; when the index exceeds the second threshold, an "emergency" level alarm is triggered, indicating that the system is in a high-risk state and immediate intervention is required. The system compares the calculated panoramic operational status index with these thresholds in real time to quickly and objectively determine the current alarm level for the entire system. This step realizes the transformation from continuous quantitative situational awareness to discrete, actionable alarm decision-making.

[0102] Step S320: Analyze the severity of anomalies in the power supply-side characteristic quantities, grid-side characteristic quantities, load-side characteristic quantities, and energy storage-side characteristic quantities that constitute the panoramic operation status label, and determine the key anomaly links.

[0103] The panoramic operational status label itself is a qualitative summary of the combination patterns of characteristic quantities in each link. To initiate targeted coordinated actions, further quantitative analysis is needed to determine which link(s)' characteristic quantities are abnormally dominant in the current suboptimal situation. The system backtracks the real-time calculated values ​​of the eight characteristic quantities on which the label was generated. By comparing the real-time value of each characteristic quantity with its preset normal reference value or safety threshold, the deviation or severity of each deviation is calculated. For example, the deviations of the power supply side output volatility from the standard value, the grid side power flow margin from the safety threshold, the load side sudden change probability from the baseline value, and the energy storage side SOC from the optimal range are compared. By comparing the relative magnitudes of these deviations, the one or two links with the most significant deviations and the greatest contribution to the deterioration of the current situation are identified and designated as "critical abnormal links." For example, if the analysis finds that the negative deviation of the grid side power flow margin is much higher than the deviations of all other characteristic quantities, then the "grid side" is determined to be the current critical abnormal link; if the deviations of the power supply side volatility and the load side sudden change probability are both high and comparable, then the "power supply side and load side" may be determined to be common critical abnormal links. This positioning process provides a target for subsequent precise collaboration.

[0104] Step S330: Based on the alarm level and key abnormal links, match the corresponding target alarm strategy from the preset collaborative alarm strategy library.

[0105] The collaborative alarm strategy library is a predefined knowledge base or rule base. Each strategy is associated with a specific combination of "alarm level" and "critical anomaly," specifying which types and contents of alarm instructions should be sent to the source, grid, load, and storage components under these conditions, as well as the logical coordination between these instructions. For example, a strategy might be associated with the following conditions: alarm level = "emergency," critical anomaly = "grid side (congestion)." The strategy might specify: 1) sending a "reduce output" adjustment instruction to the power plant on the source side (e.g., the sending-end power plant) that caused the congestion; 2) sending an "emergency discharge" instruction to a nearby, adjustable energy storage power station to alleviate the cross-sectional power flow; 3) sending a "start emergency demand response" notification to the receiving-end load aggregator; 4) highlighting the congested section and related protection devices on the grid-side monitoring system. The matching process uses the alarm level determined in step S310 and the critical anomaly determined in step S320 as composite query conditions to quickly retrieve the unique or optimal matching target alarm strategy from the strategy library. This demonstrates the digital and structured encapsulation of the collaborative handling experience of scheduling experts, and the realization of automated invocation.

[0106] Step S340: Based on the target alarm strategy, generate and simultaneously send logically related alarm commands to the corresponding links of the target alarm strategy on the power supply side, grid side, load side and energy storage side to trigger coordinated alarms.

[0107] After acquiring the target alarm strategy, the system automatically parses the content of each instruction defined in the strategy, the receiving objects (specific plants, systems, or control entities), and the logical order or timing relationship between the instructions. Subsequently, the system generates specific, executable alarm instruction messages or control commands, and sends them almost synchronously to the control systems or monitoring terminals of all relevant links specified by the strategy via the dispatch data network or dedicated communication channel. The logical correlation between instructions is reflected in: time coordination (sent almost simultaneously, or sequentially within a very short time window), unified objectives (jointly addressing the same critical anomaly, such as alleviating congestion on the same section), and complementary actions (e.g., one side reduces power transmission, while the other side increases absorption capacity or reduces demand). For example, for the aforementioned grid congestion emergency strategy, the system will simultaneously send corresponding adjustment instructions and alarm notifications to the designated automatic generation control (AGC) system, energy storage management system (EMS), and load aggregator platform. This achieves coordinated alarming from single-link, isolated alarms to multi-link, linked responses, directly transforming panoramic situational awareness into coordinated control actions.

[0108] Accordingly, a second aspect of the present invention provides an electronic device, including: at least one processor; and a memory connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to cause the at least one processor to perform the above-described power system source-grid-load-storage panoramic dispatching operation monitoring and collaborative alarm method.

[0109] Accordingly, a third aspect of the present invention provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the above-described method for panoramic dispatching, operation monitoring, and coordinated alarming of power system source-grid-load-storage systems.

[0110] The embodiments of the present invention aim to protect a method for panoramic dispatching and operation monitoring and coordinated alarming of power system sources, grids, loads and storage, which has the following effects: 1. By doing so, we solved the problem of... and achieved...

[0111] 2. By doing so, we solved the problem of... and achieved...

[0112] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0113] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0114] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0115] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for panoramic dispatching, operation monitoring, and collaborative alarming of power system sources, grid, load, and storage, characterized in that, Includes the following steps: Collect real-time operational data from the power source side, grid side, load side, and energy storage side of the power system; Dynamic weight adaptive clustering is performed on several real-time running data to generate a panoramic running status quantification result. The dynamic weight adaptive clustering is used to adaptively adjust the weights of several real-time running data according to the system running stage and generate panoramic running status labels and panoramic running status indices through cluster analysis. Based on the quantitative results of the panoramic operation status, coordinated alarms are triggered for the power supply side, grid side, load side, and energy storage side.

2. The method for panoramic dispatching, operation monitoring, and coordinated alarming of power system sources, grid, load, and storage as described in claim 1, is characterized in that, The step of performing dynamic weight adaptive clustering processing on several sets of real-time operational data to generate a panoramic operational status quantification result includes: The corresponding operational characteristic quantities are extracted from the real-time operational data of the power supply side, grid side, load side and energy storage side respectively; According to the system operation stage, weights are dynamically assigned to several operational feature quantities, and a current weighted feature vector is generated based on the real-time operational data. Clustering is performed on multiple historical weighted feature vectors formed from historical data to establish a tag library containing several panoramic operation status labels; The current weighted feature vector generated based on the real-time operation data is matched with the tag library to output the corresponding panoramic operation status tag, and the quantized distance between the current weighted feature vector and the cluster center of the matched panoramic operation status tag is calculated as the panoramic operation status index.

3. The method for panoramic dispatching, operation monitoring, and coordinated alarming of power system sources, grid, load, and storage according to claim 2, is characterized in that, The step of extracting corresponding operational characteristic quantities from the real-time operational data of the power supply side, grid side, load side, and energy storage side includes: Based on the real-time operating data of the power supply side, power supply side characteristic quantities are calculated, including power output fluctuation rate and power output prediction deviation. Based on the real-time operating data of the power grid side, the characteristic quantities of the power grid side are calculated, including the power flow margin and voltage stability margin of the key section. Based on the real-time operating data of the load side, load side characteristic quantities are calculated, including demand response potential and probability of uncontrollable load mutations. Based on the real-time operating data of the energy storage side, the characteristic quantities of the energy storage side are calculated, including the state of charge and adjustable power. The operating characteristic is obtained by combining the power supply side characteristic, the grid side characteristic, the load side characteristic, and the energy storage side characteristic.

4. The method for panoramic dispatching, operation monitoring, and coordinated alarming of power system source-grid-load-storage systems according to claim 3, characterized in that, The step of dynamically assigning weights to several operational features based on the system's operational phase, and generating a current weighted feature vector based on the real-time operational data, includes: Based on the total system load and total renewable energy output in the real-time operating data, the current system operation stage is identified, which includes the peak load stage, the renewable energy generation stage, and the normal balance stage. Based on the system operation phase, a preset benchmark weight configuration is obtained, which includes the initial weight values ​​of the power supply side characteristic, the grid side characteristic, the load side characteristic, and the energy storage side characteristic. Based on the real-time values ​​of the power supply side characteristics, the grid side characteristics, the load side characteristics, and the energy storage side characteristics, the initial weight values ​​in the benchmark weight configuration are dynamically and collaboratively corrected to generate the final weight of each characteristic. The final weights are used to weight the values ​​of the power source side characteristic, the grid side characteristic, the load side characteristic, and the energy storage side characteristic, respectively, and all the weighted values ​​are combined into a vector in a predetermined order to generate the current weighted feature vector.

5. The method for panoramic dispatching, operation monitoring, and coordinated alarming of power system sources, grid, load, and storage according to claim 4, characterized in that, The method involves dynamically and collaboratively correcting the initial weight values ​​in the benchmark weight configuration based on the real-time values ​​of the power supply-side characteristics, the grid-side characteristics, the load-side characteristics, and the energy storage-side characteristics, to generate the final weight for each characteristic, including: The deviation of the real-time value of the output fluctuation rate in the power supply side characteristic quantity from its preset standard value is calculated and used as the power supply side fluctuation deviation. The negative deviation of the real-time value of the power flow margin of the key section in the power grid side characteristic quantity relative to its preset safety threshold is calculated as the power grid congestion deviation. The growth deviation of the real-time value of the uncontrollable load mutation probability among the load-side characteristic quantities relative to its historical baseline value is calculated as the load mutation deviation. The deviation of the real-time value of the state of charge in the energy storage side characteristic quantity from the boundary of its preset optimal range is calculated as the energy storage state deviation. Based on the power supply side fluctuation deviation, the grid congestion deviation, the load change deviation, and the energy storage state deviation, the weight correction coefficients corresponding to the power supply side characteristic, the grid side characteristic, the load side characteristic, and the energy storage side characteristic are calculated respectively through a preset collaborative correction function. The initial weight value of each feature is multiplied by the corresponding weight correction coefficient to obtain the final weight of the corresponding feature.

6. The method for panoramic dispatching, operation monitoring, and coordinated alarming of power system sources, grid, load, and storage according to claim 2, is characterized in that, The method involves clustering multiple historical weighted feature vectors formed based on historical data to establish a tag library containing several panoramic operational status labels, including: Obtain multiple historical weighted feature vectors classified and stored according to the system operation stage within a historical time period, with each historical weighted feature vector corresponding to a specific historical moment; Clustering algorithms are used to perform cluster analysis on the set of historical weighted feature vectors belonging to the same stage of system operation, dividing the feature space into several clusters and determining the center vector of each cluster. For each cluster, based on the magnitude of the characteristic components in its central vector that represent the power supply side fluctuation level, grid side congestion level, load side sudden change risk and energy storage side support capacity, the cluster is assigned the panoramic operation status label according to the preset label generation rules. Establish a mapping relationship between the center vector of the cluster and the corresponding panoramic operation status label, and the label library is composed of the center vector of all clusters and their mapped panoramic operation status labels.

7. The method for panoramic dispatching, operation monitoring, and coordinated alarming of power system sources, grid, load, and storage according to claim 6, is characterized in that, The calculation of the quantized distance between the current weighted feature vector and the cluster centers of the matched panoramic operational status labels as the panoramic operational status index includes: Obtain the cluster center vector corresponding to the panoramic operation status label; Based on all historical weighted feature vectors belonging to the cluster center vector, calculate the statistical distribution of the distance between them and the cluster center vector, and determine the distance benchmark for measuring the degree of tension based on the statistical distribution. Calculate the weighted Euclidean distance between the current weighted feature vector and the cluster center vector; The calculated weighted Euclidean distance is compared with the distance benchmark. By using a preset normalization mapping function, the weighted Euclidean distance is converted into a continuous scalar within a preset numerical range, thus obtaining the panoramic operating status index that characterizes the degree to which the overall operating state of the system deviates from its typical pattern.

8. The method for panoramic dispatching, operation monitoring, and coordinated alarming of power system sources, grid, load, and storage according to claim 7, is characterized in that, The calculation of the statistical distribution of the distance between the vector and the cluster center vector, and the determination of a distance benchmark for measuring the degree of tension based on the statistical distribution, includes: Calculate the set of historical distances between all historical weighted feature vectors belonging to the cluster center vector and the cluster center vector, and calculate its statistical mean and standard deviation based on the set of historical distances; Based on the numerical distribution characteristics of the historical distance set, historical distance values ​​that represent significant anomalies in the operating status within the cluster are selected from the historical distance set and used as distance references to represent typical abnormal fluctuations within the cluster. Based on the preset importance level of the system operation mode represented by the panoramic operation status label corresponding to the cluster, the preliminary benchmark value determined based on the statistical mean and the distance reference is adjusted. Based on the different requirements for operational stability at the current system operation stage, the distance benchmark value after importance adjustment is finally calibrated to generate the distance benchmark used for current situation assessment.

9. The method for panoramic dispatching, operation monitoring, and coordinated alarming of power system sources, grid, load, and storage according to claim 7, is characterized in that, The step of comparing the calculated weighted Euclidean distance with the distance benchmark, and converting the weighted Euclidean distance into a continuous scalar within a preset numerical range using a preset normalization mapping function, to obtain the panoramic operational status index characterizing the degree to which the overall system operating state deviates from its typical pattern, includes: The weighted Euclidean distance is compared with the distance reference to determine the degree of relative deviation of the weighted Euclidean distance from the distance reference. Based on the different numerical intervals in which the relative deviation is located, the weighted Euclidean distance is mapped to a first mapping value located in the first preset numerical interval through a piecewise linear mapping function. The first mapping value is correlated and adjusted with the safety margin level of the typical operating mode represented by the panoramic operating status label to obtain the second mapping value; The second mapping value is restricted to a second preset numerical closed interval, and the restricted value is output as the panoramic operation status index.

10. The method for panoramic dispatching, operation monitoring, and coordinated alarming of power system source-grid-load-storage systems according to any one of claims 1-9, characterized in that, Based on the quantitative results of the panoramic operation status, the system triggers coordinated alarms for the power supply side, grid side, load side, and energy storage side, including: The panoramic operation status index is compared with the preset multi-level alarm threshold, and the current alarm level of the power system is determined based on the comparison result; The severity of anomalies in the power supply-side characteristic quantities, grid-side characteristic quantities, load-side characteristic quantities, and energy storage-side characteristic quantities that constitute the panoramic operation status label is analyzed to determine the key anomaly links. Based on the alarm level and the key abnormal link, the corresponding target alarm strategy is matched from the preset collaborative alarm strategy library; According to the target alarm strategy, alarm commands with logical correlation are generated and sent simultaneously to the corresponding links of the target alarm strategy in the power supply side, grid side, load side and energy storage side to trigger the coordinated alarm.

Images