Data analysis method and system based on power business agent

Abstract

The invention provides a data analysis method and system based on a power business agent, and the method comprises the steps: firstly obtaining an initial data set of a plurality of business nodes in a power business system, carrying out the feature analysis of the initial data set, extracting equipment state features and scene condition correlation features, and generating a business feature set; then inputting the service feature set into a pre-constructed power service agent analysis model, generating a power service analysis result containing an equipment state fault tendency and a service scene adaptation state, and determining a potential type and fault penetration data of a power equipment operation fault according to the analysis result; and finally, based on the potential type and the fault penetration data, generating a service optimization instruction containing an equipment positioning identifier, and sending the service optimization instruction to the target power equipment management terminal to trigger a management operation, thereby effectively integrating the data, accurately analyzing the service state, and improving the operation stability and the service processing efficiency of the power equipment.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to a data analysis method and system based on an intelligent agent for power business. Background Technology

[0002] In the power sector, with the rapid development of smart grids and the dramatic increase in the number of power devices, power business systems have become increasingly complex. Currently, power business systems contain numerous business nodes, typically involving various stages such as power generation, transmission, substation, distribution, and consumption. Each business node generates a large amount of data. Power equipment operation data reflects the real-time status of the equipment, while business scenario-related data is closely linked to power business processes, covering environmental condition information and user demand information.

[0003] However, existing power business data analysis methods have significant shortcomings. On the one hand, the processing of power equipment operation data and business scenario-related data is often isolated, lacking in-depth exploration of the inherent connections between the two, making it difficult to comprehensively and accurately grasp the overall operational status of the power business. On the other hand, traditional technologies lack flexibility and intelligence when processing complex and ever-changing power business data, failing to dynamically adjust analysis strategies according to different business scenarios and equipment states. This results in inaccurate predictions of power equipment operational failures and assessments of business scenario adaptability, making it difficult to meet the needs of efficient and safe operation of the power business. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a data analysis method based on a power business intelligent agent, the method comprising: Obtain an initial data set from multiple business nodes in the power business system. The initial data set includes power equipment operation data and business scenario-related data. The power equipment operation data is continuously collected equipment status information, and the business scenario-related data includes environmental condition information and user demand information related to the power business process. The initial dataset is subjected to feature parsing processing to extract equipment status features from power equipment operation data and scenario condition association features from business scenario association data, generating a business feature set containing equipment status features and scenario condition association features. The equipment status features are used to describe the changing pattern of equipment status over time, and the scenario condition association features are used to describe the interrelationship between different conditions in the business scenario. The set of business features is input into a pre-built power business intelligent agent analysis model to generate power business analysis results that include equipment status failure tendency and business scenario adaptation status. The power business analysis results represent the correlation degree and change trend of each dimension feature in the form of feature vectors. Based on the power business analysis results, determine the potential types of power equipment operation failures and the failure penetration data of failure tendencies in business scenarios; Based on the potential types and fault penetration data, a service optimization instruction containing the device location identifier is generated, and the service optimization instruction is sent to the target power equipment management terminal to trigger the corresponding management operation.

[0005] In another aspect, embodiments of the present invention also provide a data analysis system based on a power business intelligent agent, including a processor and a machine-readable storage medium connected to the processor. The machine-readable storage medium is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the machine-readable storage medium to implement the above-described method.

[0006] Based on the above, this embodiment of the invention acquires an initial data set from multiple business nodes in the power business system, comprehensively integrates power equipment operation data and business scenario-related data, and performs feature parsing processing on the initial data set. This enables accurate extraction of equipment status features and scenario condition-related features, deeply revealing the changing patterns of equipment status over time and the interrelationships between different conditions in the business scenario. This leads to a deeper and more detailed understanding of the power business. The business feature set is input into a pre-built power business intelligent agent analysis model to generate power business analysis results represented in the form of feature vectors. This effectively presents the correlation and changing trends of features in each dimension, achieving intelligent and accurate analysis of the power business status. Based on the analysis results, the potential types of power equipment operation faults and fault penetration data in the business scenario are determined. Finally, based on the potential types of power equipment operation faults and fault penetration data in the business scenario, a business optimization instruction containing equipment location identifier is generated and sent to the target power equipment management terminal. This can trigger corresponding management operations in a timely manner, effectively improving the stability of power equipment operation and the efficiency of power business processing, and ensuring the safe and reliable operation of the power system. Attached Figure Description

[0007] Figure 1 This is a schematic diagram of the execution flow of the data analysis method based on power business intelligent agents provided in an embodiment of the present invention.

[0008] Figure 2 This is a schematic diagram of exemplary hardware and software components of a data analysis system based on a power business intelligent agent provided in an embodiment of the present invention. Detailed Implementation

[0009] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1This is a flowchart illustrating a data analysis method based on a power business intelligent agent according to an embodiment of the present invention. The following is a detailed description of the data analysis method based on a power business intelligent agent.

[0010] Step S110: Obtain the initial data set of multiple business nodes in the power business system. The initial data set includes power equipment operation data and business scenario related data. The power equipment operation data is continuously collected equipment status information, and the business scenario related data is environmental condition information and user demand information related to the power business process.

[0011] In the actual operation of the power business, numerous business nodes are distributed throughout the power business system. These nodes play a crucial role in data collection throughout the entire process of power generation, from production to end-user consumption. Specifically, taking the power supply system of a large city as an example, in the generation stage, the generator monitoring equipment in each power plant is an important business node. It collects various operational data of the generators in real time, such as voltage, current, power, temperature, and speed. This operational data reflects the real-time operating status of the generators and is essential for ensuring the stability and safety of power generation. In the transmission stage, various sensors installed on transmission lines act as business nodes, responsible for collecting information such as current, voltage, temperature, and tension of the lines to monitor their operating status and promptly identify potential safety hazards. In the distribution stage, monitoring devices in substations are also business nodes, monitoring parameters such as voltage, current, and power factor to ensure the stability and reliability of power distribution.

[0012] Operating data of power equipment is continuously collected through sensors installed on key parts of the equipment. These sensors can capture changes in the equipment's status in real time. For example, temperature sensors are installed on the stator windings of generators to monitor the temperature changes of the windings in real time; oil temperature sensors are installed in the oil sump of transformers to obtain oil temperature information in a timely manner.

[0013] The business scenario-related data includes environmental condition information and user demand information. Environmental condition parameters, such as temperature, humidity, wind speed, and precipitation, significantly impact the operation of power equipment. High temperatures can hinder heat dissipation in power equipment, affecting its performance and lifespan; strong winds can cause transmission lines to sway, increasing the risk of line faults. Power system environmental parameters, such as grid voltage stability and frequency stability, reflect the overall operating status of the power system. Unstable grid voltage can damage power equipment and affect users' normal electricity consumption. Regarding user demand information, user electricity consumption time parameters reflect the differences in user demand at different times. For example, industrial users typically have higher electricity demand during the day on weekdays, while residential users have relatively higher demand at night. Electricity capacity parameters reflect the scale of user electricity demand; different sizes of businesses and households have different requirements for electricity capacity. Power supply reliability requirements reflect users' expectations for power supply stability; some enterprises with extremely high power supply requirements, such as hospitals and data centers, have strict requirements for power supply reliability.

[0014] Step S120: Perform feature parsing processing on the initial dataset, extract equipment status features from the power equipment operation data and scenario condition association features from the business scenario association data, and generate a business feature set containing equipment status features and scenario condition association features. The equipment status features are used to describe the changing pattern of equipment status over time, and the scenario condition association features are used to describe the interrelationship between different conditions in the business scenario.

[0015] Step S121: Perform time series segmentation processing on the power equipment operation data, dividing the continuous equipment status information into multiple operation data segments with fixed time intervals. Each operation data segment contains equipment status parameters at multiple time points.

[0016] After acquiring continuous power equipment operation data, time series segmentation is required to more effectively analyze the changes in equipment status over time. First, the data acquisition frequency and time span must be determined. The acquisition frequency depends on the sensor performance and settings, determining the number of data acquisitions per unit time. The time span is determined based on actual business needs and analytical objectives, covering the equipment operation period to be analyzed. The time interval for each data segment is determined based on the acquisition frequency and the preset segment length. The segment length needs to comprehensively consider the power equipment's operating cycle and business requirements. For example, for equipment with short and rapidly changing operating cycles, the segment length can be set shorter to capture changes in equipment status more precisely; for equipment with longer and relatively stable operating cycles, the segment length can be appropriately extended.

[0017] Step S1211: Determine the collection frequency and time span of power equipment operation data. Based on the collection frequency and the preset segment time length, determine the time interval of each operation data segment. The segment time length is set according to the operation cycle of the power equipment and business requirements.

[0018] The sampling frequency is determined by the sensor's performance and system settings; different types of sensors may have different sampling frequencies. For example, for parameters sensitive to change, such as generator voltage and current, a higher sampling frequency may be used to ensure timely detection of minute changes; while for parameters that change relatively slowly, such as transformer oil temperature, a relatively lower sampling frequency is acceptable. The time span is determined based on specific analytical needs and may be a day, a week, a month, or even longer. The preset segment length needs to comprehensively consider the operating cycle of the power equipment and business requirements. For example, for a small generator with a clear daily start-stop pattern, the segment length can be set to one operating cycle, i.e., one day. Based on the sampling frequency and segment length, the time interval for each operating data segment can be calculated. For example, if the sampling frequency is once per hour and the segment length is one day, then the time interval for each operating data segment is one hour.

[0019] Step S1212: Extract the equipment status information of each operating data segment from the power equipment operation data in chronological order, so that adjacent operating data segments are continuous in time and do not overlap. Each operating data segment contains the identifiers of the start time point and the end time point.

[0020] After determining the time intervals for the operational data segments, the equipment status information for each segment is extracted from the power equipment operational data in chronological order. For example, from the generator operational data, parameters such as voltage, current, and temperature for each operational data segment are extracted sequentially at set time intervals. This ensures that adjacent operational data segments are temporally continuous and do not overlap, thus guaranteeing data integrity and continuity. Simultaneously, start and end time markers are added to each operational data segment to facilitate subsequent data management and analysis. For example, an operational data segment might start at 8:00 AM and end at 9:00 AM, thus clearly defining the time range corresponding to that data segment during data analysis.

[0021] Step S1213: Perform an integrity check on the device status parameters in each running data segment, and remove time points with missing data or faults. If the proportion of missing data exceeds a preset threshold, the time interval of the running data segment is re-divided so that the number of valid data points in each running data segment meets the analysis requirements.

[0022] After extracting the device status information from the operational data segments, it is necessary to perform an integrity check on the device status parameters within each segment. During data acquisition, sensor malfunctions, communication interruptions, and other issues may occur, leading to missing or incorrect data. For example, if the voltage data at a certain point in time shows an abnormal value or is empty, it indicates a potential problem with the data at that point. Time points with missing or faulty data need to be removed from the operational data segments. Then, the data missing percentage for each operational data segment is calculated. If the missing percentage exceeds a preset threshold, it means the data quality for that segment does not meet the analysis requirements. In this case, the time intervals of the operational data segments need to be redefined, for example, by shortening the time intervals, to increase the number of valid data points and ensure that each segment contains sufficient valid data for subsequent analysis.

[0023] Step S1214: Number the processed running data segments according to time sequence to generate a set of running data segments containing running data segment number, time interval, start time point, end time point and device status parameters.

[0024] After completing the data integrity check and processing, the processed running data segments are numbered chronologically. For example, the first running data segment is numbered 1, the second is numbered 2, and so on. The generated set of running data segments includes the segment number, time interval, start time, end time, and device status parameters for each segment. For example, the segment number allows for quick location of a specific running data segment, the start and end times define the time range of that segment, and the device status parameters provide the foundation for feature extraction.

[0025] Step S1215: Determine the feature analysis weight of each running data segment based on the time interval of the running data segment and the change range of the equipment status parameters.

[0026] The feature analysis weights of operational data segments are determined based on the time interval and the magnitude of changes in equipment status parameters. Operational data segments with shorter time intervals and larger changes in equipment status parameters indicate more drastic changes in the equipment's operating status within that time period. These segments are more valuable for analyzing equipment status changes and providing early warnings of faults, and therefore their feature analysis weights can be set relatively higher. For example, if a generator's voltage and current fluctuate significantly within a certain operational data segment, and the time interval of this segment is short, then the feature analysis weight for this segment can be set higher than for other operational data segments. By appropriately setting the feature analysis weights, subsequent analyses can focus more on operational data segments that have a greater impact on changes in equipment status.

[0027] Step S122: Perform trend analysis processing on each running data segment, calculate the rate of change and direction of change of equipment status parameters at adjacent time points, and generate equipment status features that reflect the change of equipment status over time. These equipment status features include multiple dimensions of rate of change and direction parameters.

[0028] After obtaining the processed set of operational data segments, trend analysis is performed on each segment. For each device status parameter within each segment, the parameter difference between adjacent time points is calculated sequentially to obtain the parameter change for each time interval. For example, for the generator voltage parameter within a certain operational data segment, the voltage difference between two adjacent time points is calculated; this voltage difference represents the voltage change within that time interval. If the voltage value at the later time point is greater than the voltage value at the earlier time point, the change is positive; otherwise, it is negative.

[0029] Step S1221: For each device status parameter within each running data segment, calculate the parameter difference between adjacent time points in sequence to obtain the parameter change amount for each time interval.

[0030] Taking generator voltage parameters as an example, within an operating data segment, there are voltage data points at multiple time points. Assuming the voltage at the first time point is one value and the voltage at the second time point is another, subtracting the voltage value at the first time point from the voltage value at the second time point gives the voltage change between these two time points. By performing the above calculation on adjacent time points, the voltage change for each time interval within the operating data segment can be obtained. The same method is used to calculate the changes in other equipment status parameters, such as current, temperature, and speed, for each time interval.

[0031] Step S1222: Divide the parameter change by the time interval length to obtain the parameter change rate, which represents the magnitude of change of the equipment status parameter per unit time.

[0032] After obtaining the parameter change amount for each time interval, dividing it by the corresponding time interval length yields the parameter change rate. For example, if the voltage change between two points in time is a fixed value and the time interval is one hour, dividing the change amount by one hour gives the hourly voltage change rate. This rate reflects the magnitude of voltage change per unit time. By comparing the change rates of the same parameter within different operating data segments, or the change rates of different parameters within the same operating data segment, we can understand how quickly the equipment status parameters change.

[0033] Step S1223: Determine the direction of change based on the positive or negative value of the parameter change. A positive value indicates that the parameter is increasing, a negative value indicates that the parameter is decreasing, and a zero value indicates that the parameter remains stable.

[0034] The direction of parameter change is determined by the sign of the calculated change. A positive value indicates that the parameter is increasing during the time interval; a negative value indicates that the parameter is decreasing; and a zero value indicates that the parameter remains stable. For example, a positive change in generator voltage indicates that the voltage is increasing during the time interval, possibly due to a decrease in generator load or adjustments to the excitation system; a negative value indicates that the voltage is decreasing, possibly due to an increase in load or a generator malfunction; and a zero value indicates that the voltage remains constant, suggesting that the generator's operating state is relatively stable.

[0035] Step S1224: Standardize the rate of change and direction of change of each device status parameter, arrange the standardized rate of change and direction of change of all device status parameters in the order of the parameters' dimensions, and generate device status features. Each dimension of the device status feature corresponds to the rate of change and direction of change information of a device status parameter. The device status feature is used to characterize the dynamic change trend of the device status in the time series.

[0036] Since the dimensions and value ranges of different device state parameters may vary, it is necessary to standardize the rate of change and direction of change for each device state parameter to eliminate the impact of these differences on subsequent analysis. There are various methods for standardization, such as normalization, which maps the rate of change and direction of change to a defined interval. Then, the standardized rates of change and directions of change for all device state parameters are arranged in dimensional order. For example, the standardized rates of change and directions of change for voltage parameters are arranged first, followed by current parameters, and so on. The resulting device state feature is a multi-dimensional vector, with each dimension corresponding to the rate of change and direction of change of a device state parameter, effectively displaying the dynamic trend of device state changes over time.

[0037] Step S123: Perform conditional classification processing on the business scenario-related data to generate conditional classification results that include category identifiers, parameter lists, business link correspondences, and parameter importance ranking.

[0038] The business scenario-related data contains a wealth of information. To better analyze the impact of this data on the power business, it is necessary to perform condition classification processing. First, environmental condition information and user demand information are separated from the business scenario-related data. Environmental condition information includes natural environmental parameters such as temperature, humidity, wind speed, and precipitation, as well as power system environmental parameters such as grid voltage stability and frequency stability. User demand information includes parameters such as user electricity consumption time periods, electricity capacity, and power supply reliability requirements.

[0039] Step S1231: Separate environmental condition information and user demand information from the business scenario associated data. The environmental condition information includes natural environment parameters and power system environment parameters, while the user demand information includes user electricity consumption time parameters, electricity capacity parameters, and power supply reliability requirements parameters.

[0040] In business scenario-related data, environmental condition information and user demand information are separated through data filtering and classification methods. For example, data collected from various business nodes is classified according to its source and nature. If the data concerns the natural environment or power system environment, such as temperature and humidity data provided by weather stations or voltage stability data provided by power grid monitoring systems, it is classified as environmental condition information. If the data concerns users' electricity needs, such as information submitted by users regarding electricity usage periods, electricity capacity, and power supply reliability requirements, it is classified as user demand information.

[0041] Step S1232: Perform parameter filtering operations on the separated environmental condition information and user demand information respectively, retain parameters that have a clear correspondence with at least one link in the power business process, and obtain the initial parameter set after filtering.

[0042] The separated environmental condition information and user demand information are then filtered for parameters. For environmental condition information, each parameter is checked to see if it has a clear correspondence with a specific stage in the power business process. For example, temperature parameters may be related to the heat dissipation of power equipment; high temperatures may cause heat dissipation difficulties, affecting equipment performance. If a parameter does not have a clear correspondence with the power business process, it is discarded. User demand information is similarly filtered, retaining parameters that are clearly related to the generation, transmission, and distribution stages of the power business process. For example, user electricity consumption time parameters are related to the formulation of generation plans, and electricity capacity parameters are related to the load capacity of transmission lines. Through these filtering operations, an initial set of filtered parameters is obtained.

[0043] Step S1233: Perform intra-group correlation test on the parameters in each initial parameter set to determine whether the parameters jointly affect the same business result or are driven by the same business factor. Parameters that pass the intra-group correlation test are classified into the same condition category, and a unique category identifier is assigned to each condition category. The category identifier contains category source information and category order information.

[0044] After obtaining the initial set of filtered parameters, an intra-group correlation test is performed on the parameters within each set. By analyzing the correlation and causal relationships between parameters, it is determined whether they jointly affect the same business outcome or are driven by the same business factor. For example, in the initial set of environmental condition information parameters, temperature and humidity may jointly affect the heat dissipation effect of power equipment, and they can be classified into the same condition category. A unique category identifier is assigned to each condition category, which contains category source information and category order information. For example, the category identifier can be represented by "Environment" or "User" to indicate the source, followed by a sequential number, such as "Environment 1", "User 2", etc., which can effectively identify the source and order of each condition category.

[0045] Step S1234: Establish the correspondence between each condition category and the power business process link. The content of this correspondence is the name of the specific business link affected by the condition category and the direction of the impact.

[0046] After categorizing the conditions, establish the correspondence between each condition category and the relevant stages of the power business process. For each condition category, analyze its impact on each stage of the power business process. For example, the "Environment 1" category (including temperature and humidity parameters) may affect the power generation and transmission stages of power equipment. In the power generation stage, high temperature and high humidity environments may reduce generator efficiency; in the transmission stage, it may increase the resistance of transmission lines, leading to increased transmission losses. Clearly define the specific business stages affected by each condition category and the direction of its impact; for example, "Environment 1" has a negative impact on both the power generation and transmission stages.

[0047] Step S1235: Based on the degree of direct impact of the parameters on the corresponding business process operation, perform an importance sorting operation on the parameters in each condition category to generate a condition classification result containing category identifier, parameter list, business process correspondence, and parameter importance sorting.

[0048] After establishing the correspondence between condition categories and business process steps, the parameters within each condition category are ranked by importance based on their direct impact on the operational effectiveness of the corresponding business steps. For example, in the "Environment 1" category, temperature parameters may have a greater impact on the heat dissipation of electrical equipment, and therefore their importance may be higher than that of humidity parameters. By ranking the parameters by importance, a condition classification result is generated that includes a category identifier, a parameter list, the correspondence between business steps, and the ranking of parameter importance.

[0049] Step S124: Based on the condition classification results, generate scenario condition association features that describe the condition associations in the business scenario. The elements of the scenario condition association features represent the association strength between different condition categories.

[0050] After obtaining the condition classification results, scenario condition association features describing the condition associations in the business scenario are further generated. First, the parameter list for each condition category in the condition classification results is extracted, and the parameter with the highest importance in each parameter list is selected as the key parameter for that condition category.

[0051] Step S1241: Extract the parameter list for each condition category in the condition classification results, and select the parameter with the highest importance from each parameter list as the key parameter for that condition category.

[0052] The parameter list for each condition category in the condition classification results is extracted, and the parameter ranked first in importance from each list is selected as the key parameter for that condition category. In the previously generated condition classification results, each condition category has a corresponding parameter list and a ranking of parameter importance. For example, for the "Environment 1" condition category, its parameter list might include parameters such as temperature and humidity. Based on the previous importance ranking, temperature is ranked first, so it is selected as the key parameter for the "Environment 1" condition category. Similarly, the same operation is performed for other condition categories. For example, for the "User 2" condition category, if its parameter list includes parameters such as user electricity usage time and electricity capacity, and electricity capacity is ranked first in importance, then electricity capacity becomes the key parameter for the "User 2" condition category. In this way, a key parameter is determined for each condition category, and these key parameters can, to a certain extent, represent the impact of the corresponding condition category on the power business.

[0053] Step S1242: Form a parameter set from the key parameters of all condition categories, and statistically analyze the frequency and direction of change of any two key parameters of different condition categories in the parameter set in historical business data to obtain the correlation calculation results.

[0054] The key parameters of all condition categories are aggregated to form a parameter set. Based on historical business data, the key parameters of any two different condition categories in this parameter set are analyzed. For each pair of key parameters, the frequency with which they change simultaneously in the historical business data is counted. For example, for the key parameter temperature of the "Environment 1" condition category and the key parameter electricity capacity of the "User 2" condition category, observe in the historical business data which time periods temperature and electricity capacity change simultaneously, and record the number of such time periods to determine their frequency of simultaneous change. Simultaneously, the consistency of the direction of change of these two key parameters should also be considered. If, in certain time periods, the temperature rises while the electricity capacity also increases, or the temperature falls while the electricity capacity also decreases, then their direction of change is considered consistent in these time periods; otherwise, they are inconsistent. Through the analysis of all different key parameter pairs, correlation calculation results are obtained, which reflect the association between key parameters of different condition categories.

[0055] Step S1243: Obtain pre-stored business rule data, determine the degree of influence between any two condition categories based on the business rule data, and obtain the degree of influence determination result. The business rule data contains records of the common influence of different condition categories on the same business process.

[0056] The pre-stored business rule data is derived from long-term power business practices and experience, and includes records of the combined impact of different condition categories on the same business process. For example, the business rule data might record the combined impact of the "Environment 1" condition category (with temperature as the key parameter) and the "User 2" condition category (with electricity capacity as the key parameter) on the power generation process. When the temperature rises, it may lead to a decrease in generator efficiency, while an increase in electricity capacity will increase the generator load. These two condition categories work together on the power generation process, affecting the stability and efficiency of power generation. Based on this business rule data, the degree of influence between any two condition categories is analyzed. For example, if the business rules indicate that changes in temperature and electricity capacity will significantly affect the output power of the power generation process, then it can be determined that the influence between the "Environment 1" and "User 2" condition categories is relatively large. By analyzing all different condition category pairs, the degree of influence is determined.

[0057] Step S1244: The correlation calculation result and the influence degree judgment result are weighted and fused to obtain the correlation strength between the two condition categories.

[0058] To comprehensively consider the correlation of key parameters and their degree of influence among different condition categories, a weighted fusion of the correlation calculation results and the influence determination results is necessary. First, the weights of the correlation calculation results and the influence determination results are determined. These weights can be determined based on actual business needs and experience. For example, if the correlation of key parameters is considered to have a greater impact on the association between condition categories, then the correlation calculation results can be assigned a higher weight; conversely, if the degree of influence between condition categories is more important, then the influence determination results can be assigned a higher weight. Then, the correlation calculation results and the influence determination results are fused according to their respective weights. For example, assuming the correlation calculation results are represented by a standardized mapped value, and the influence determination results are also represented by a standardized mapped value, multiplying the correlation calculation result by its weight and the influence determination result by its weight, and then adding the two products, the result is the association strength between the two condition categories. This weighted fusion method can more comprehensively and accurately reflect the association relationships between different condition categories.

[0059] Step S1245: Arrange the association strength between any two condition categories in the order of the category identifiers to form a two-dimensional matrix, and generate scene condition association features. The row index and column index of the two-dimensional matrix correspond to the category identifiers of the condition categories, and the elements of the two-dimensional matrix are the association strength between the corresponding condition categories.

[0060] After obtaining the association strength between any two condition categories, they are arranged according to their category identifiers. A two-dimensional matrix is ​​constructed using the category identifiers as row and column indices. For example, if the condition categories are "Environment 1," "User 2," and "Environment 3," then the first row and first column of the two-dimensional matrix are indices for "Environment 1," the second row and second column for "User 2," and so on. Each element in the matrix corresponds to the association strength between the two corresponding condition categories. For instance, the element in the first row and second column of the matrix represents the association strength between the condition categories "Environment 1" and "User 2." In this way, the association strengths between all condition categories are presented in the form of a two-dimensional matrix, forming a scenario condition association feature. This scenario condition association feature can intuitively display the association relationships between different condition categories in a business scenario.

[0061] Step S125: Combine the device status features and scene condition association features according to the correspondence between time series and condition categories to generate a business feature set.

[0062] Equipment status features reflect the changing patterns of power equipment status over time, while scenario condition association features describe the relationships between different condition categories within a business scenario. These two features are combined according to the correspondence between time series and condition categories. For example, at a specific point in time, equipment status features contain information on the rate and direction of change of various parameters of the power equipment at that time, while scenario condition association features reflect the strength of the association between different condition categories at that time. Based on the association between condition categories and power equipment, equipment status features and scenario condition association features are matched and combined. If a certain condition category affects a specific power equipment, then the scenario condition association feature corresponding to that condition category is combined with the equipment status feature of that power equipment. Through this method, equipment status features and scenario condition association features at different time points are combined to ultimately generate a business feature set, which integrates information from both equipment status and business scenario conditions.

[0063] Step S130: Input the set of business features into the pre-built power business intelligent agent analysis model to generate power business analysis results that include equipment status failure tendency and business scenario adaptation status. The power business analysis results represent the correlation degree and change trend of each dimension feature in the form of feature vectors.

[0064] Step S131: Obtain the dimensional distribution information of device status features and the dimensional distribution information of scene condition associated features in the business feature set. The dimensional distribution information includes the specific description content and arrangement order of each feature.

[0065] The business feature set includes device status features and scenario condition association features, and their dimensional distribution information needs to be obtained separately. For device status features, the dimensional distribution information includes a detailed description of the device status parameters corresponding to each dimension, such as the rate and direction of voltage change, the rate and direction of current change, etc., as well as the order of these dimensions. For example, the first dimension of the device status feature might correspond to the generator's voltage change rate, the second dimension to the generator's current change rate, and so on. For scenario condition association features, the dimensional distribution information includes a detailed description of the association strength between the corresponding condition categories for each dimension, as well as the order of these dimensions. For example, the first dimension of the scenario condition association feature might correspond to the association strength between the condition categories "Environment 1" and "User 2," the second dimension to the association strength between the condition categories "Environment 1" and "Environment 3," and so on. By obtaining this dimensional distribution information, the specific composition and arrangement of each feature in the business feature set can be clearly understood.

[0066] Step S132: Perform a format unification operation on the business feature set according to the dimensional distribution information, and convert the device status features and scene condition association features into a continuous sequence structure of the same data type to form a unified input feature sequence.

[0067] Based on the acquired dimensional distribution information, the business feature set undergoes a format unification operation. Since equipment status features and scenario condition association features may have different data types and structures, they need to be converted into continuous sequence structures of the same data type to facilitate processing by the subsequent power business intelligent agent analysis model. For example, equipment status features might be stored in vector form, while scenario condition association features might be stored in matrix form. By rearranging and transforming the equipment status features and scenario condition association features, they are unified into a continuous sequence structure, ensuring data type consistency. This unified input feature sequence allows the power business intelligent agent analysis model to process them more efficiently.

[0068] Step S133: Input the structurally unified input feature sequence into the input processing unit of the power business intelligent agent analysis model, adjust the numerical range of each dimension feature in the input feature sequence, and obtain the standardized feature sequence.

[0069] The input processing unit of the power business intelligent agent analysis model is primarily responsible for preprocessing the input feature sequence. Since the numerical ranges of features across different dimensions in the input feature sequence can vary significantly, this can affect the model's training and analysis performance. Therefore, the input processing unit adjusts the numerical ranges of the features across each dimension. For example, it uses normalization or standardization methods to map the values ​​of each feature dimension to a defined interval, ensuring that different features have similar numerical ranges. This eliminates the influence of different dimensions and numerical values, improving the model's stability and accuracy. After processing by the input processing unit, a standardized feature sequence is obtained, which is more suitable as input for subsequent processing in the power business intelligent agent analysis model.

[0070] Step S134: Input the standardized feature sequence into the multilayer perceptron structure of the power business intelligent agent analysis model, and perform cross-interaction analysis on the equipment state features and scene condition correlation features through hierarchical calculation, and output the correlation weight sequence reflecting the degree of interaction between the equipment state features and scene condition correlation features.

[0071] Step S1341: Determine the layer number setting for the hierarchical calculation method. This layer number setting is positively correlated with the number of dimensions of the standardized feature sequence.

[0072] The number of layers in a multilayer perceptron architecture is determined by the number of dimensions of the standardized feature sequence. Generally, the more dimensions the standardized feature sequence has, the more complex the information to be processed, and thus the more layers the multilayer perceptron architecture needs to have. For example, if the standardized feature sequence contains many device state feature dimensions and scene condition association feature dimensions, then a larger number of layers is needed to more fully analyze the interactions between these features. By reasonably determining the number of layers, the multilayer perceptron architecture can effectively process the standardized feature sequence.

[0073] Step S1342: Input the standardized feature sequence into the first layer computing node of the multilayer perceptron structure. The first layer computing node performs preliminary combination calculation on the input device state features and scene condition association features, and outputs the first layer combined features.

[0074] The standardized feature sequence is input into the first-layer computation node of the multilayer perceptron structure. In this node, preliminary combination calculations are performed on the input device state features and scene condition association features. For example, certain dimensions of the device state features are correlated and calculated with certain dimensions of the scene condition association features. New feature combinations are obtained through defined calculation rules. This process can be understood as a preliminary fusion of the input features, aiming to uncover the initial correlation between the device state features and the scene condition association features. After processing by the first-layer computation node, the first-layer combined features are output, which contain information from the preliminary combination of the device state features and scene condition association features.

[0075] Step S1343: Input the first layer combined features into the next layer computing node of the multilayer perceptron structure. The next layer computing node performs nonlinear transformation processing on the received combined features to enhance the nonlinear interaction between features and outputs the next layer combined features.

[0076] The first-layer combined features are passed to the next-layer computation node in the multilayer perceptron structure. The next-layer computation node performs a nonlinear transformation on the received combined features. This nonlinear transformation can introduce more complex relationships and enhance the nonlinear interactions between features. For example, by using a nonlinear activation function to process the first-layer combined features, the relationships between features are no longer simple linear relationships, thus revealing more hidden and complex correlations between features. After processing by the next-layer computation node, the next-layer combined features are output. These next-layer combined features contain the feature information after the nonlinear transformation, further enriching the correlation information between device state features and scene condition association features.

[0077] Step S1344: Repeat the feature input and nonlinear transformation processing steps until the preset number of layers is reached, and output the final combined features.

[0078] Following the aforementioned feature input and nonlinear transformation processing steps, the combined features from the previous layer are continuously input into the next layer's computation node for processing. Each layer's processing further enhances the nonlinear interactions between features, uncovering more feature correlation information. This process continues until a preset number of layers is reached. Once the preset number of layers is reached, the final combined feature is output. This final combined feature, obtained after multiple layers of computation and nonlinear transformation processing, fully reflects the complex cross-interactions between device state features and scene condition correlation features.

[0079] Step S1345: Extract the interaction strength values ​​between the device state features and the scene condition association features from the final combined features, arrange them according to the dimensional order of the standardized feature sequence, and form an association weight sequence that reflects the degree of interaction between the device state features and the scene condition association features.

[0080] After obtaining the final combined features, the interaction strength values ​​between the device state features and the scene condition association features of each dimension are extracted. These values ​​represent the degree of interaction between different dimensional features. For example, the interaction strength value between a certain dimension of the device state features and a certain dimension of the scene condition association features reflects the tightness of the correlation between these two dimensions. These interaction strength values ​​are arranged according to the dimensional order of the standardized feature sequence. For example, if the first dimension of the standardized feature sequence corresponds to a parameter of the device state features, then the interaction strength value between that parameter and the dimensions of the scene condition association features is placed first in the association weight sequence. Through the above arrangement, an association weight sequence is formed, which effectively reflects the degree of interaction between the device state features and the scene condition association features.

[0081] Step S135: Extract the temporal arrangement information of equipment status features from the standardized feature sequence, input the temporal arrangement information of the equipment status features into the long short-term memory network structure of the power business intelligent agent analysis model, and use the long short-term memory network structure combined with the associated weight sequence to deduce the temporal change law of equipment status features, and output a time-series prediction feature vector containing a description of the trend of equipment status change.

[0082] The temporal arrangement information of equipment state features is extracted from the standardized feature sequence, recording the order of equipment state features at different time points. This temporal arrangement information is then input into the Long Short-Term Memory (LSTM) network structure of the power business intelligent agent analysis model. The LTM network structure has a memory function, enabling it to handle long-term dependencies in time-series data. Combined with the previously obtained association weight sequence, the LTM network structure can better analyze the temporal variation patterns of equipment state features. For example, the association weight sequence reflects the degree of interaction between equipment state features and scenario condition association features; the LTM network structure can use this association information to predict changes in equipment state features at future time points. Through processing by the LTM network structure, a time-series predicted feature vector containing a description of equipment state change trends is output. This vector contains information on the changing trends of equipment state over a future period, which is of great significance for equipment fault early warning and condition assessment.

[0083] Step S136: Extract the category correspondence information of the scenario condition association features from the standardized feature sequence, input the scenario condition association features into the fully connected network structure of the power business intelligent agent analysis model according to the category correspondence information, and use the fully connected network structure combined with the association weight sequence to judge the combination matching relationship of the scenario condition association features, and output the scenario adaptation feature vector containing the business scenario condition adaptation description.

[0084] The category correspondence information of scenario condition association features is extracted from the standardized feature sequence. This information indicates the correspondence between different dimensions of the scenario condition association features and the condition categories. The scenario condition association features are then input into the fully connected network structure of the power business intelligent agent analysis model according to the category correspondence information. The fully connected network structure can comprehensively connect and compute the input features. Combined with the association weight sequence, the fully connected network structure can analyze the combination matching relationship of scenario condition association features. For example, the association weight sequence reflects the degree of association between different condition categories. Based on this association information, the fully connected network structure can determine whether the combination of different scenario condition association features is suitable for the needs of the power business. Through the processing of the fully connected network structure, a scenario adaptation feature vector containing a description of the business scenario condition adaptation is output. This scenario adaptation feature vector describes the adaptation status of the business scenario conditions.

[0085] Step S137: Combine the time-series prediction feature vector and the scenario adaptation feature vector according to the cross-mapping relationship between the time-series prediction feature vector and the category-corresponding information to generate power business analysis results.

[0086] The time-series prediction feature vector contains information on the temporal trend of equipment status changes, while the scenario adaptation feature vector describes the adaptation to business scenario conditions. These two vectors are combined according to the cross-mapping relationship between time-series information and category-corresponding information. For example, at a specific point in time, based on the time-series information and category-corresponding information, the equipment status change trend information at that point is combined with the corresponding business scenario condition adaptation information. By combining information from different time points and different condition categories, a power business analysis result is generated. This result represents the correlation and trend of features in each dimension in the form of feature vectors, integrating information on equipment status failure tendency and business scenario adaptation status.

[0087] Step S140: Based on the power business analysis results, determine the potential types of power equipment operation failures and the failure penetration data of failure tendencies in the business scenario.

[0088] Step S141: Identify abnormal feature combinations based on the power business analysis results, and perform clustering processing on the abnormal feature combinations. Abnormal feature combinations with similar change patterns are grouped into the same fault feature category, and a unique identifier is assigned to each fault feature category.

[0089] The power business analysis results contain various characteristic information about equipment status and business scenarios. By analyzing this characteristic information, abnormal feature combinations are identified. For example, if the rate and direction of change of certain parameters in the equipment status characteristics deviate from normal patterns, and the correlation strength between certain condition categories in the scenario condition correlation characteristics also changes abnormally, then these abnormal feature combinations can be identified. The identified abnormal feature combinations are then clustered. The clustering is based on the change pattern of the abnormal feature combinations. If certain abnormal feature combinations have similar change trends in the time series or similar correlations in the feature dimensions, they are grouped into the same fault feature category. Each fault feature category is assigned a unique identifier for convenient subsequent management and analysis. For example, abnormal feature combinations with similar abnormal voltage and current change patterns are grouped into one fault feature category and assigned a specific identifier.

[0090] Step S142: Based on the mapping relationship between fault feature categories and predefined fault types, determine the potential type of power equipment operation fault corresponding to each fault feature category.

[0091] A mapping relationship between fault feature categories and predefined fault types has been established in advance. The above mapping relationship is summarized based on a large amount of historical fault data and expert experience. For each fault feature category, according to its characteristic manifestations and change patterns, the corresponding predefined fault type is searched in the mapping relationship. For example, a certain fault feature category is manifested as the continuous increase of the temperature parameter of the device and abnormal current fluctuations, and at the same time, the correlation intensity of some condition categories in the business scenario also shows abnormal changes. By querying the mapping relationship, it is found that the predefined fault type corresponding to the above feature combination may be a fault in the device's cooling system. Through the above method, the potential types of power equipment operation faults corresponding to each fault feature category are determined.

[0092] Step S142: Determine the business processes affected by the equipment state fault tendency corresponding to the potential type, and the fault propagation path in the business scenario, where the fault propagation path is used to reflect the specific link nodes and transmission methods of the fault transferring from the equipment side to the business processes.

[0093] After determining the potential types of power equipment operation faults, it is necessary to further analyze which business processes will be affected by the fault tendencies of these potential types. Different fault types have different impacts on business processes. For example, if the potential fault type is a fault in the generator, the affected business processes may include the power generation process itself, and may also spread to the power transmission process and the power distribution process. Because a generator fault may lead to a decrease in power generation, which in turn affects the load of the power transmission line and the stability of power distribution.

[0094] At the same time, it is necessary to determine the fault propagation path in the business scenario. The fault propagation path reflects the specific link nodes and transmission methods of the fault transferring from the equipment side to the business processes. Taking the generator fault as an example, the fault may first affect the output power of the generator, and then transfer the above influence to the substation through the power transmission line, thereby affecting the power distribution process. In this process, the generator, the power transmission line, the substation, etc. are all specific link nodes on the fault propagation path, and the changes in power, voltage fluctuations, etc. are the ways of fault transmission. By analyzing the fault propagation path, the scope of influence and transmission mechanism of the fault can be understood more clearly.

[0095] Step S143: According to the fault propagation path and the affected business processes, determine the fault penetration scope and penetration direction of the fault tendency in the business scenario, and generate fault penetration data including the fault penetration scope and penetration direction.

[0096] Based on the fault propagation path and the affected business processes, the scope and direction of fault penetration within the business scenario are further determined. The scope of fault penetration refers to the range of business processes that a fault may affect. For example, if the fault propagation path involves multiple processes such as power generation, transmission, and distribution, then the scope of fault penetration covers all of these processes. The direction of fault penetration refers to the direction in which the fault propagates within the business scenario, typically from the equipment end to the business processes. For example, a fault starting from a generator and propagating along transmission lines to substations and the distribution system constitutes the direction of fault penetration.

[0097] By conducting a detailed analysis of the fault propagation path and affected business processes, and comprehensively considering various factors such as the severity of the fault and the interrelationships between business processes, the scope and direction of fault penetration are determined. This information is then compiled and recorded to generate fault penetration data containing the scope and direction of fault penetration. This data is crucial for assessing the impact of faults and developing response strategies. For example, based on the scope and direction of fault penetration, it is possible to determine which business processes require priority preventative measures to minimize the impact of the fault on the entire power business system.

[0098] Step S150: Generate a service optimization instruction containing the device location identifier based on the potential type and fault penetration data, and send the service optimization instruction to the target power equipment management terminal to trigger the corresponding management operation.

[0099] For example, step S151: Obtain the equipment fault feature information recorded in the potential type, which includes the abnormal identification of the equipment operating parameters corresponding to the fault type.

[0100] After identifying the potential types of power equipment malfunctions, fault characteristic information is extracted from the records of these potential types. This information, accumulated during fault diagnosis and analysis, includes abnormal identifiers of equipment operating parameters corresponding to the fault type. For example, for a specific fault type, the corresponding fault characteristic information might indicate that the equipment's voltage parameters are outside the normal range, the temperature is too high, or the speed is unstable. These abnormal identifiers of equipment operating parameters are crucial for identifying faulty equipment and assessing the severity of the fault. By acquiring this information, the specific situation of the faulty equipment can be understood more accurately.

[0101] Step S152: Based on the abnormal identification of the equipment operating parameters, extract the physical location information of the corresponding equipment from the power equipment asset database. The physical location information includes a description of the equipment installation location and a unique number. Use the physical location information as the equipment location identifier.

[0102] The power equipment asset database stores detailed information on all power equipment, including its physical location. Based on anomaly indicators in the equipment's operating parameters, queries and matches are performed in the database to extract the corresponding physical location information. This physical location information includes a detailed description of the equipment's installation location, such as which substation, floor, and specific location it is installed in, as well as a unique identification number. This unique number ensures accurate identification and location of the equipment. The extracted physical location information is used as the equipment's location identifier. For example, maintenance personnel can quickly locate the faulty equipment using this identifier for timely repair and handling.

[0103] Step S153: Analyze the impact path of the fault recorded in the fault penetration data in the business scenario. The impact path includes the specific nodes of the fault transmission from the device end to the business process.

[0104] Fault penetration data meticulously records the impact path of faults within business scenarios, serving as a crucial basis for analyzing the scope of fault impact and developing response strategies. In-depth analysis of fault penetration data reveals the specific nodes in the fault's propagation from the equipment end to business processes. For example, a fault might originate from a generator, travel through transmission lines to a substation, then affect distribution transformers, and finally impact the user end. Analyzing these specific nodes effectively elucidates the fault's propagation process and scope of impact within the business scenario. This helps identify which business processes are most affected and which require priority for preventative measures.

[0105] Step S154: Determine the management operation content to be performed based on the impact path. The management operation content includes status adjustment actions on the device side or process intervention actions on the business process.

[0106] Based on the analysis of the impact path of the fault in the business scenario, the management operations that need to be performed are determined. These management operations mainly include status adjustment actions on the equipment side and process intervention actions on the business side. For equipment status adjustment actions, depending on the type and severity of the fault, it may be necessary to repair the faulty equipment, replace parts, or adjust the equipment's operating parameters. For example, if the fault is caused by the damage of a certain component, then that component needs to be replaced promptly; if the equipment's operating parameters are outside the normal range, then the parameters need to be adjusted.

[0107] For intervention actions in business processes, adjustments and optimizations may be necessary depending on the affected business processes and the extent of the impact of the fault. For example, if a fault affects the stability of the transmission line, it may be necessary to adjust the load distribution of the transmission line or activate the backup transmission line; if a fault affects the power quality of the distribution line, it may be necessary to adjust the operation strategy of the distribution system to ensure normal power supply for users. By identifying these management operations, faults can be handled in a targeted manner, reducing their impact on the power business system.

[0108] Step S155: Associate and combine the device location identifier with the management operation content to generate a business optimization instruction that includes the device location identifier and management operation content.

[0109] The previously identified equipment location identifiers and management operation details are associated and combined. The equipment location identifier clarifies the specific location of the faulty equipment, while the management operation details describe the specific measures to be taken in response to the fault. Combining these two generates a business optimization instruction that includes the equipment location identifier and management operation details. For example, the business optimization instruction might explicitly state, "The generator located at position XX on floor XX of substation XX (equipment unique number: XXXX) has failed and requires immediate repair of the generator's cooling system and adjustment of the load distribution on the transmission line." This business optimization instruction effectively informs the target power equipment management terminal of the location of the faulty equipment to be addressed and the specific management operation details, facilitating timely implementation of appropriate measures by the management terminal.

[0110] Step S156: Send the service optimization instruction to the target power equipment management terminal to trigger the corresponding management operation.

[0111] After generating a business optimization instruction, it is sent to the target power equipment management terminal. The target power equipment management terminal is responsible for managing and controlling power equipment; it can receive business optimization instructions and trigger corresponding management operations based on the instruction content. The business optimization instructions are accurately transmitted to the target power equipment management terminal via the communication network. Upon receiving the instruction, the management terminal can parse and process it, locate the faulty equipment based on the equipment location identifier, and perform maintenance and adjust business processes according to the requirements of the management operation. For example, the management terminal may notify maintenance personnel to go to the location of the faulty equipment for repair, while simultaneously adjusting the operating parameters and processes of relevant business processes. Through this method, timely handling of power equipment faults and optimization of business processes are achieved, improving the stability and reliability of the power business system.

[0112] Figure 2The illustration shows exemplary hardware and software components of a power business intelligent agent-based data analysis system 100 that can implement the ideas of this application, according to some embodiments of this application. For example, a processor 120 can be used in the power business intelligent agent-based data analysis system 100 and to perform the functions in this application.

[0113] The data analysis system 100 based on power business intelligent agents can be a general-purpose server or a special-purpose server; both can be used to implement the data analysis method based on power business intelligent agents of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the load.

[0114] For example, a data analysis system 100 based on a power business intelligent agent may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the data analysis system 100 based on a power business intelligent agent may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The data analysis system 100 based on a power business intelligent agent also includes an input / output I / O interface 150 between the computer and other input / output devices.

[0115] For ease of explanation, only one processor is described in the data analysis system 100 based on power business intelligent agents. However, it should be noted that the data analysis system 100 based on power business intelligent agents in this application may also include multiple processors. Therefore, the steps executed by one processor as described in this application may also be executed jointly or individually by multiple processors. For example, if the processor of the data analysis system 100 based on power business intelligent agents executes steps A and B, it should be understood that steps A and B may also be executed jointly by two different processors or individually by one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor jointly execute steps A and B.

[0116] Furthermore, this embodiment of the invention also provides a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the above-mentioned data analysis method based on power business intelligent agents is implemented.

[0117] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A data analysis method based on a power business intelligent agent, characterized in that, The method includes: Obtain an initial data set from multiple business nodes in the power business system. The initial data set includes power equipment operation data and business scenario-related data. The power equipment operation data is continuously collected equipment status information, and the business scenario-related data includes environmental condition information and user demand information related to the power business process. The initial dataset is subjected to feature parsing processing to extract equipment status features from power equipment operation data and scenario condition association features from business scenario association data, generating a business feature set containing equipment status features and scenario condition association features. The equipment status features are used to describe the changing pattern of equipment status over time, and the scenario condition association features are used to describe the interrelationship between different conditions in the business scenario. The set of business features is input into a pre-built power business intelligent agent analysis model to generate power business analysis results that include equipment status failure tendency and business scenario adaptation status. The power business analysis results represent the correlation degree and change trend of each dimension feature in the form of feature vectors. Based on the power business analysis results, determine the potential types of power equipment operation failures and the failure penetration data of failure tendencies in business scenarios; Based on the potential types and fault penetration data, a service optimization instruction containing the device location identifier is generated, and the service optimization instruction is sent to the target power equipment management terminal to trigger the corresponding management operation.

2. The data analysis method based on power business intelligent agents according to claim 1, characterized in that, The initial dataset is subjected to feature parsing processing to extract equipment status features from power equipment operation data and scenario condition association features from business scenario association data, generating a business feature set containing equipment status features and scenario condition association features, including: The power equipment operation data is processed by time series segmentation, dividing the continuous equipment status information into multiple operation data segments with fixed time intervals. Each operation data segment contains equipment status parameters at multiple time points. Trend analysis is performed on each running data segment to calculate the rate of change and direction of change of equipment status parameters at adjacent time points, and to generate equipment status features that reflect the change of equipment status over time. The equipment status features include multiple dimensions of rate of change and direction parameters. The business scenario-related data is processed by conditional classification to generate conditional classification results that include category identifiers, parameter lists, business link correspondences, and parameter importance rankings. Based on the condition classification results, scenario condition association features describing the condition associations in the business scenario are generated, and the elements of the scenario condition association features represent the association strength between different condition categories. The device status features and scenario condition association features are combined according to the correspondence between time series and condition categories to generate a set of business features.

3. The data analysis method based on power business intelligent agents according to claim 2, characterized in that, The step of performing time-series segmentation processing on the power equipment operation data, dividing continuous equipment status information into multiple operation data segments with fixed time intervals, includes: The frequency and time span for collecting power equipment operation data are determined, and the time interval for each operation data segment is determined based on the collection frequency and the preset segment time length, wherein the segment time length is set according to the operation cycle of the power equipment and business requirements; The equipment status information of each operating data segment is extracted sequentially from the power equipment operation data in chronological order so that adjacent operating data segments are continuous in time and do not overlap. Each operating data segment includes the identifiers of the start time point and the end time point. The integrity of the device status parameters in each running data segment is checked, and time points with missing data or faults are removed. If the proportion of missing data exceeds the preset threshold, the time interval of the running data segment is re-divided so that the number of valid data points in each running data segment meets the analysis requirements. The processed running data segments are numbered in chronological order to generate a set of running data segments containing the running data segment number, time interval, start time point, end time point, and device status parameters; The feature analysis weights for each running data segment are determined based on the time interval of the running data segment and the magnitude of changes in equipment status parameters.

4. The data analysis method based on power business intelligent agents according to claim 2, characterized in that, The process of performing trend analysis on each segment of operational data, calculating the rate of change and direction of change of equipment status parameters at adjacent time points, and generating equipment status characteristics reflecting the changes in equipment status over time includes: For each device status parameter within each running data segment, the parameter difference between adjacent time points is calculated sequentially to obtain the parameter change in each time interval; Divide the parameter change by the time interval length to obtain the parameter change rate, which represents the magnitude of change of the device status parameter per unit time. The direction of change is determined by the sign of the parameter change: a positive value indicates that the parameter is increasing, a negative value indicates that the parameter is decreasing, and a zero value indicates that the parameter remains stable. The rate of change and direction of change of each device status parameter are standardized. The standardized rate of change and direction of change of all device status parameters are arranged in order of the parameter dimensions to generate device status features. Each dimension of the device status features corresponds to the rate of change and direction of change information of a device status parameter. The device status features are used to characterize the dynamic change trend of device status over time.

5. The data analysis method based on power business intelligent agents according to claim 2, characterized in that, The conditional classification processing of the business scenario-related data generates a conditional classification result containing category identifiers, parameter lists, business process correspondences, and parameter importance rankings, including: Separate environmental condition information and user demand information from business scenario-related data. The environmental condition information includes natural environment parameters and power system environment parameters, while the user demand information includes user electricity consumption time parameters, electricity capacity parameters, and power supply reliability requirements parameters. Perform parameter filtering operations on the separated environmental condition information and user demand information respectively, retain the parameters that have a clear correspondence with at least one link in the power business process, and obtain the initial parameter set after filtering; Perform intra-group correlation tests on the parameters in each initial parameter set to determine whether the parameters jointly affect the same business result or are driven by the same business factor. Parameters that pass the intra-group correlation test are classified into the same condition category, and a unique category identifier is assigned to each condition category. The category identifier contains category source information and category order information. Establish a correspondence between each condition category and a step in the power business process. The correspondence includes the name of the specific business step affected by the condition category and the direction of its impact. Based on the degree of direct impact of parameters on the operational effectiveness of their corresponding business processes, an importance ranking operation is performed on the parameters within each condition category, generating a condition classification result that includes a category identifier, a parameter list, the corresponding business process relationship, and the parameter importance ranking.

6. The data analysis method based on power business intelligent agents according to claim 5, characterized in that, The step of generating scenario condition association features describing condition associations in a business scenario based on the condition classification results includes: Extract the parameter list for each condition category from the condition classification results, and select the parameter with the highest importance from each parameter list as the key parameter for that condition category; The key parameters of all condition categories are grouped into a parameter set. The frequency and direction of change of any two key parameters of different condition categories in the parameter set are statistically analyzed in historical business data to obtain the correlation calculation results. Obtain pre-stored business rule data, determine the degree of influence between any two condition categories based on the business rule data, and obtain the degree of influence determination result. The business rule data contains records of the common influence of different condition categories on the same business process. The correlation calculation result and the influence degree determination result are weighted and fused to obtain the correlation strength between the two condition categories; According to the category identifier order of the condition categories, the association strength between any two condition categories is arranged into a two-dimensional matrix to generate scene condition association features. The row index and column index of the two-dimensional matrix correspond to the category identifier of the condition category, and the elements of the two-dimensional matrix are the association strength between the corresponding condition categories.

7. The data analysis method based on power business intelligent agents according to claim 1, characterized in that, The step of inputting the set of business features into a pre-built power business intelligent agent analysis model to generate power business analysis results including equipment status failure tendency and business scenario adaptation status includes: Obtain the dimensional distribution information of device status features and the dimensional distribution information of scene condition associated features in the business feature set. The dimensional distribution information includes the specific description content and arrangement order of each feature. Based on the dimensional distribution information, a format unification operation is performed on the business feature set to convert the device status features and scene condition association features into a continuous sequence structure of the same data type, forming a unified input feature sequence. The standardized input feature sequence is input into the input processing unit of the power business intelligent agent analysis model, and the numerical range of each dimension feature in the input feature sequence is adjusted to obtain a standardized feature sequence. The standardized feature sequence is input into the multilayer perceptron structure of the power business intelligent agent analysis model. The cross-interaction analysis of equipment state features and scene condition correlation features is performed through hierarchical calculation. The correlation weight sequence that reflects the degree of interaction between the equipment state features and scene condition correlation features is output. The time arrangement information of equipment status features is extracted from the standardized feature sequence. The equipment status features are then input into the long short-term memory network structure of the power business intelligent agent analysis model according to the time arrangement information. The time change law of equipment status features is deduced by combining the long short-term memory network structure with the associated weight sequence, and a time-series prediction feature vector containing a description of the equipment status change trend is output. Extract the category correspondence information of the scenario condition association features from the standardized feature sequence, input the scenario condition association features into the fully connected network structure of the power business intelligent agent analysis model according to the category correspondence information, and use the fully connected network structure combined with the association weight sequence to judge the combination matching relationship of the scenario condition association features, and output the scenario adaptation feature vector containing the business scenario condition adaptation description. The power business analysis results are generated by combining the cross-mapping relationship between the time-series prediction feature vector and the scenario adaptation feature vector arranged by time and the corresponding information of the category.

8. The data analysis method based on power business intelligent agents according to claim 7, characterized in that, The process involves inputting the standardized feature sequence into the multilayer perceptron structure of the power business intelligent agent analysis model, performing cross-interaction analysis on equipment state features and scenario condition correlation features through hierarchical calculation, and outputting a correlation weight sequence reflecting the degree of interaction between the equipment state features and scenario condition correlation features, including: The number of layers in the hierarchical calculation method is determined, and the number of layers is positively correlated with the number of dimensions of the standardized feature sequence; The standardized feature sequence is input into the first layer computing node of the multilayer perceptron structure. The first layer computing node performs preliminary combination calculation on the input device state features and scene condition association features, and outputs the first layer combined features. The first-layer combined features are input into the next-layer computing node of the multilayer perceptron structure. The next-layer computing node performs nonlinear transformation processing on the received combined features to enhance the nonlinear interaction between features and outputs the next-layer combined features. Repeat the feature input and nonlinear transformation processing steps until the preset number of layers is reached, and output the final combined features; The interaction strength values ​​between the device state features and the scene condition association features of each dimension are extracted from the final combined features, and arranged in the dimensional order of the standardized feature sequence to form an association weight sequence that reflects the degree of interaction between the device state features and the scene condition association features.

9. The data analysis method based on power business intelligent agents according to claim 1, characterized in that, The fault penetration data, which determines the potential types and fault tendencies of power equipment operation faults in the business scenario based on the power business analysis results, includes: Based on the results of power business analysis, abnormal feature combinations were identified and clustered. Abnormal feature combinations with similar change patterns were grouped into the same fault feature category, and a unique identifier was assigned to each fault feature category. Based on the mapping relationship between fault feature categories and predefined fault types, determine the potential types of power equipment operation faults corresponding to each fault feature category; Identify the business processes affected by the potential type of equipment status failure tendency, and the failure propagation path in the business scenario. The failure propagation path is used to reflect the specific nodes and transmission methods of the failure from the equipment end to the business process. Based on the fault propagation path and the affected business processes, the fault penetration range and direction in the business scenario are determined, and fault penetration data containing the fault penetration range and direction is generated.

10. A data analysis system based on an intelligent agent for power business, characterized in that, The device includes a processor and a memory, the memory and the processor being connected. The memory is used to store programs, instructions or code, and the processor is used to execute the programs, instructions or code in the memory to implement the data analysis method based on power business intelligent agents as described in any one of claims 1-9.

Images