Intelligent power supply unified management method and system

Abstract

The present invention relates to the technical field of power management data analysis, and in particular to a method and system for unified intelligent power management. The method comprises the following steps: integrating and collecting heterogeneous power data based on a power management system to generate integrated power data; performing multi-level power perception feature analysis based on the integrated power data to generate multi-level power perception feature data; performing local multi-element power perception driven associated feature analysis based on local multi-level power perception feature data to generate local power perception driven associated feature data; performing local dynamic power scheduling demand intelligent analysis based on local power perception driven associated feature data to generate local dynamic power scheduling demand data; designing global intelligent power management decisions based on local dynamic power scheduling demand data; and executing intelligent power unified management operations based on global intelligent power management decisions. The present invention analyzes the characteristics of heterogeneous power supplies and the characteristics of each power usage scenario to achieve intelligent unified power management.

Description

Technical Field

[0001] The present invention relates to the technical field of power management data analysis, and in particular to an intelligent power unified management method and system. Background Art

[0002] With the rapid development of modern smart devices, renewable energy systems and Internet of Things technologies, power management systems have gradually become a core component of power infrastructure. Especially in the context of diversified energy structures and increased uncertainty in electricity demand, power management faces many challenges. Power management usually relies on fixed power dispatching modes and manual decisions, and it is difficult to dynamically adapt to power grid fluctuations and changes in load demand. Intelligent power management can automatically manage the dispatching and allocation of power through intelligent technical means, improve the level of intelligence and power management efficiency. However, due to the diversity of power types, the existing intelligent power unified management methods make the management methods of different power sources relatively independent, and cannot achieve global unified and efficient scheduling. It is difficult to meet the high-frequency power dispatching needs, and there are still problems such as energy waste and unstable operation; and it is difficult to effectively deal with complex and changeable power consumption scenarios. With the increase in the scale of new energy access and the diversification of power demand, it is impossible to carry out targeted unified power management for power consumption scenarios. Summary of the invention

[0003] Based on this, the present invention provides an intelligent power supply unified management method to solve at least one of the above technical problems.

[0004] To achieve the above object, a smart power unified management method includes the following steps:

[0005] Step S1: performing initial power management data collection based on the power management system to generate initial power management data; performing heterogeneous power data integration processing based on the initial power management data to generate integrated power data;

[0006] Step S2: performing a multi-level power-aware cluster analysis based on the integrated power data to generate multi-level power-aware cluster data; performing a multi-level power-aware feature analysis based on the multi-level power-aware cluster data to generate multi-level power-aware feature data;

[0007] Step S3: performing local distribution division processing according to the multi-level power perception feature data to generate local multi-level power perception feature data; performing local multi-element power perception drive correlation feature analysis based on the local multi-level power perception feature data to generate local power perception drive correlation feature data;

[0008] Step S4: perform local power drive feature analysis based on local multi-level power perception feature data to generate local power drive feature data; design an intelligent monitoring engine for local power scheduling demand prediction based on local power perception drive associated feature data to obtain a local power scheduling demand prediction engine; transmit the local power drive feature data to the local power scheduling demand prediction engine in real time to perform local dynamic power scheduling demand intelligent analysis to generate local dynamic power scheduling demand data; perform automated decision analysis of global intelligent power management based on the local dynamic power scheduling demand data to generate a global intelligent power management decision; perform intelligent power unified management adjustment processing on the initial power management data based on the global intelligent power management decision to generate intelligent power unified management data, and execute intelligent power unified management operations through the intelligent power unified management data.

[0009] The present invention can obtain the operation information of various power supply devices in real time, including important data such as energy consumption, state, and load, by collecting the initial power management data. The automation of this data collection improves the real-time performance of the system and ensures that the operation status of the power supply can be accurately reflected. For different types of power supply systems, the problem of management independence caused by different protocols and architectures in the traditional management system is solved through the integrated processing of heterogeneous power supply data. The integrated data forms a unified power management data pool, so that the system can realize unified scheduling and management of different power supplies on one platform. Through the multi-level perception clustering analysis of the integrated power supply data, the system can comprehensively obtain the performance of the power supply at different levels (energy consumption, state, efficiency, etc.). The overall operation of the power supply can be captured more accurately, and then effectively managed and scheduled. The abnormal data is automatically screened out through cluster analysis, which improves the accuracy of data processing and avoids the interference of abnormal power supply data on the overall management decision. The screened data ensures the reliability of subsequent analysis and decision-making. The multi-level power perception feature data generated by feature analysis provides basic data support for subsequent power scheduling. These feature data can effectively reflect the core characteristics of the power supply and help the system make more accurate decisions and scheduling. The multi-sensory driven correlation characteristics of local power sources are analyzed to dig out the key characteristics and correlations of power sources in local use. This helps to improve the system's efficiency in using power sources under local loads and optimize its energy distribution. Through correlation feature analysis, the usage patterns and mutual influences between various types of power sources are discovered, and then accurate predictions are made for local scheduling needs. This correlation analysis helps the system to make more accurate power scheduling, avoid resource waste, and improve energy efficiency. By designing an intelligent monitoring engine for local power scheduling demand prediction, the system can monitor the operating status and usage of power sources in real time, and make accurate demand predictions based on historical data and current status. The engine can significantly improve the intelligence level of local power scheduling and ensure that the system can adjust power distribution according to actual needs. Through real-time intelligent analysis of local power scheduling needs, the system can dynamically adjust the power usage strategy, respond quickly, and ensure the balance between load and power supply. This real-time dynamic scheduling enables the system to maintain stable operation when power demand fluctuates greatly, avoiding the problem of excessive energy consumption or insufficient supply. Based on the local power scheduling demand data, the system further makes global intelligent power management decisions, comprehensively considering the operating status and needs of all power sources. This global optimization decision analysis ensures that the power dispatch of the entire system reaches the optimal state, effectively improving the efficiency and overall stability of power management. The initial power management data is intelligently and uniformly managed and adjusted, and the system can perform fully automatic unified management of heterogeneous power supplies, reducing the need for manual intervention. This automated execution not only improves the efficiency of power dispatch, but also ensures the accuracy and consistency of management operations.

[0010] This specification provides an intelligent power unified management method system for executing the intelligent power unified management method as described above. The intelligent power unified management system includes:

[0011] An integrated power data collection module is used to collect initial power management data based on the power management system to generate initial power management data; perform heterogeneous power data integration processing based on the initial power management data to generate integrated power data;

[0012] A multi-level power-aware feature analysis module is used to perform multi-level power-aware cluster analysis based on the integrated power data to generate multi-level power-aware cluster data; perform multi-level power-aware feature analysis based on the multi-level power-aware cluster data to generate multi-level power-aware feature data;

[0013] A local power-sensing driving correlation feature analysis module is used to perform local distribution division processing based on multi-level power-sensing feature data to generate local multi-level power-sensing feature data; perform local multi-level power-sensing driving correlation feature analysis based on the local multi-level power-sensing feature data to generate local power-sensing driving correlation feature data;

[0014] The intelligent power unified management module is used to perform local power drive feature analysis based on local multi-level power perception feature data to generate local power drive feature data; design an intelligent monitoring engine for local power scheduling demand prediction based on local power perception drive associated feature data to obtain a local power scheduling demand prediction engine; transmit the local power drive feature data in real time to the local power scheduling demand prediction engine to perform intelligent analysis of local dynamic power scheduling demand to generate local dynamic power scheduling demand data; perform automated decision analysis of global intelligent power management based on local dynamic power scheduling demand data to generate global intelligent power management decisions; perform intelligent power unified management adjustment processing on the initial power management data based on the global intelligent power management decision to generate intelligent power unified management data, and execute intelligent power unified management operations through the intelligent power unified management data.

[0015] The beneficial effect of the present application is that the intelligent power unified management method of the present invention realizes the unified integrated management of data of different types of power supplies through the collection and analysis of initial power management data and the design of heterogeneous power management protocol stacks, and can integrate decentralized power management systems into a whole, thereby enhancing the coordination of the system. Through multi-level power perception feature data, local scheduling demand prediction engine, and global intelligent management decision analysis, it can effectively generate global intelligent power management decisions and realize global unified scheduling of power supplies. This global scheduling strategy can ensure the coordinated operation of multiple heterogeneous power supplies, greatly improve the efficiency of power scheduling, and avoid the problems of unreasonable scheduling and energy waste caused by decentralized management. Through multi-level power perception clustering analysis and feature analysis, the system can fully perceive the power consumption and state characteristics, respond to the diverse needs of different power usage scenarios, and screen out abnormal power nodes and power supplies through abnormal clustering data, thereby improving the accuracy and robustness of the system. By introducing a local power scheduling demand prediction engine, the driving characteristics of local power are analyzed in real time, and future scheduling needs are predicted. The scheduling prediction of the XGBoost model can dynamically adjust the power usage strategy to ensure efficient resource scheduling when the power usage scenario changes. According to the local power scheduling needs and the automated decision-making analysis of global data, management decisions can be automatically generated and power scheduling can be executed according to the real-time power usage scenario needs, which improves the system's ability to cope with complex scenarios, significantly reduces energy consumption, and improves system stability. In this way, automated unified power management is performed for heterogeneous power data and complex power usage scenarios, reducing unnecessary power consumption and achieving more efficient and intelligent unified power management. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of the steps of a smart power supply unified management method of the present invention;

[0017] Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart;

[0018] Figure 3 for Figure 1 Detailed implementation steps of step S4 in FIG.

[0019] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0020] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0021] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0022] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0023] To achieve this, please refer to Figures 1 to 3 The present invention provides a smart power supply unified management method, comprising the following steps:

[0024] Step S1: performing initial power management data collection based on the power management system to generate initial power management data; performing heterogeneous power data integration processing based on the initial power management data to generate integrated power data;

[0025] Step S2: performing a multi-level power-aware cluster analysis based on the integrated power data to generate multi-level power-aware cluster data; performing a multi-level power-aware feature analysis based on the multi-level power-aware cluster data to generate multi-level power-aware feature data;

[0026] Step S3: performing local distribution division processing according to the multi-level power perception feature data to generate local multi-level power perception feature data; performing local multi-element power perception drive correlation feature analysis based on the local multi-level power perception feature data to generate local power perception drive correlation feature data;

[0027] Step S4: perform local power drive feature analysis based on local multi-level power perception feature data to generate local power drive feature data; design an intelligent monitoring engine for local power scheduling demand prediction based on local power perception drive associated feature data to obtain a local power scheduling demand prediction engine; transmit the local power drive feature data to the local power scheduling demand prediction engine in real time to perform local dynamic power scheduling demand intelligent analysis to generate local dynamic power scheduling demand data; perform automated decision analysis of global intelligent power management based on the local dynamic power scheduling demand data to generate a global intelligent power management decision; perform intelligent power unified management adjustment processing on the initial power management data based on the global intelligent power management decision to generate intelligent power unified management data, and execute intelligent power unified management operations through the intelligent power unified management data.

[0028] The present invention can obtain the operation information of various power supply devices in real time, including important data such as energy consumption, state, and load, by collecting the initial power management data. The automation of this data collection improves the real-time performance of the system and ensures that the operation status of the power supply can be accurately reflected. For different types of power supply systems, the problem of management independence caused by different protocols and architectures in the traditional management system is solved through the integrated processing of heterogeneous power supply data. The integrated data forms a unified power management data pool, so that the system can realize unified scheduling and management of different power supplies on one platform. Through the multi-level perception clustering analysis of the integrated power supply data, the system can comprehensively obtain the performance of the power supply at different levels (energy consumption, state, efficiency, etc.). The overall operation of the power supply can be captured more accurately, and then effectively managed and scheduled. The abnormal data is automatically screened out through cluster analysis, which improves the accuracy of data processing and avoids the interference of abnormal power supply data on the overall management decision. The screened data ensures the reliability of subsequent analysis and decision-making. The multi-level power perception feature data generated by feature analysis provides basic data support for subsequent power scheduling. These feature data can effectively reflect the core characteristics of the power supply and help the system make more accurate decisions and scheduling. The multi-sensory driven correlation characteristics of local power sources are analyzed to dig out the key characteristics and correlations of power sources in local use. This helps to improve the system's efficiency in using power sources under local loads and optimize its energy distribution. Through correlation feature analysis, the usage patterns and mutual influences between various types of power sources are discovered, and then accurate predictions are made for local scheduling needs. This correlation analysis helps the system to make more accurate power scheduling, avoid resource waste, and improve energy efficiency. By designing an intelligent monitoring engine for local power scheduling demand prediction, the system can monitor the operating status and usage of power sources in real time, and make accurate demand predictions based on historical data and current status. The engine can significantly improve the intelligence level of local power scheduling and ensure that the system can adjust power distribution according to actual needs. Through real-time intelligent analysis of local power scheduling needs, the system can dynamically adjust the power usage strategy, respond quickly, and ensure the balance between load and power supply. This real-time dynamic scheduling enables the system to maintain stable operation when power demand fluctuates greatly, avoiding the problem of excessive energy consumption or insufficient supply. Based on the local power scheduling demand data, the system further makes global intelligent power management decisions, comprehensively considering the operating status and needs of all power sources. This global optimization decision analysis ensures that the power dispatch of the entire system reaches the optimal state, effectively improving the efficiency and overall stability of power management. The initial power management data is intelligently and uniformly managed and adjusted, and the system can perform fully automatic unified management of heterogeneous power supplies, reducing the need for manual intervention. This automated execution not only improves the efficiency of power dispatch, but also ensures the accuracy and consistency of management operations.

[0029] As an embodiment of the present invention, refer to Figure 1 The above is a schematic diagram of the steps of a smart power unified management method of the present invention. In this embodiment, the smart power unified management method includes the following steps:

[0030] Step S1: performing initial power management data collection based on the power management system to generate initial power management data; performing heterogeneous power data integration processing based on the initial power management data to generate integrated power data;

[0031] In an embodiment of the present invention, the power management system is used to perform real-time monitoring and data collection on the existing management data of various power supplies, including information such as power supply output voltage, current, energy consumption, operating status, and working status. The collected data of different power supply systems are parsed and standardized according to a predetermined protocol or standard to facilitate subsequent integration operations. For example, the output data units of different power supplies may be different, and the system will uniformly convert them into a standard format. After the standardization process is completed, the data from different power supplies are integrated. By designing the management protocol stack of heterogeneous power supplies, it is ensured that the data between heterogeneous power supplies can be seamlessly integrated. The integrated power data generated at this time contains unified feature data of multiple power supplies, providing a basis for subsequent analysis.

[0032] Step S2: performing a multi-level power-aware cluster analysis based on the integrated power data to generate multi-level power-aware cluster data; performing a multi-level power-aware feature analysis based on the multi-level power-aware cluster data to generate multi-level power-aware feature data;

[0033] In an embodiment of the present invention, a hierarchical clustering algorithm is used to perform cluster analysis on data such as power consumption and operating status based on integrated power data. In this process, abnormal data is identified and removed by a preset threshold, thereby generating preliminary multi-level power-aware clustering data. For abnormal data, deviation analysis and anomaly recognition algorithms are used to screen the preliminary clustering results. Abnormal clustering data, i.e., abnormal power data, is removed, and the abnormal power data is fed back. Feature analysis is performed on the cleaned multi-level power-aware clustering data to extract energy consumption features, state features, abnormal behavior features, etc. from power data at different levels to generate multi-level power-aware feature data.

[0034] Step S3: performing local distribution division processing according to the multi-level power perception feature data to generate local multi-level power perception feature data; performing local multi-element power perception drive correlation feature analysis based on the local multi-level power perception feature data to generate local power perception drive correlation feature data;

[0035] In an embodiment of the present invention, the characteristics of each local power system are divided and extracted based on the multi-level power perception feature data. Different power systems may exhibit different characteristics in local areas. Through this local feature division, the differences between the power systems can be effectively distinguished. Based on the locally distributed multi-level power perception feature data, a local multi-power perception driven associated feature analysis is performed to analyze the correlation between different local power systems, such as the correlation between local power energy consumption and power status in various power usage scenarios, thereby mining the key feature data that affects the power scheduling requirements. The generated local power perception driven associated feature data provides a basis for subsequent scheduling demand prediction.

[0036] Step S4: perform local power drive feature analysis based on local multi-level power perception feature data to generate local power drive feature data; design an intelligent monitoring engine for local power scheduling demand prediction based on local power perception drive associated feature data to obtain a local power scheduling demand prediction engine; transmit the local power drive feature data to the local power scheduling demand prediction engine in real time to perform local dynamic power scheduling demand intelligent analysis to generate local dynamic power scheduling demand data; perform automated decision analysis of global intelligent power management based on the local dynamic power scheduling demand data to generate a global intelligent power management decision; perform intelligent power unified management adjustment processing on the initial power management data based on the global intelligent power management decision to generate intelligent power unified management data, and execute intelligent power unified management operations through the intelligent power unified management data.

[0037] In an embodiment of the present invention, the local power supply sensing driving associated feature data is used to further analyze the driving features in each local power supply system. Through this analysis, the operating characteristics and driving requirements of different local power supply nodes are clarified, and local power supply driving feature data are generated. Based on the local power supply sensing driving associated feature data, an intelligent monitoring engine is designed. The monitoring engine can process and analyze the local power supply driving feature data in real time and predict the scheduling requirements of the power supply. The core of the monitoring engine includes a machine learning-based model (such as XGBoost), which can continuously learn the operating status of the power supply system and adjust the prediction model. The local power supply driving feature data collected in real time is input into the intelligent monitoring engine for analysis. The intelligent monitoring engine processes these data and generates local dynamic power supply scheduling demand data, which can reflect the scheduling demand changes of each local power supply system in real time. Based on the local dynamic power supply scheduling demand data, a global automated decision analysis is further performed. Through this automated analysis, a global intelligent power supply management decision is generated in combination with the actual needs of each local area and the global power supply management strategy. According to the global intelligent power supply management decision, the system automatically adjusts the initial power supply management data, generates new intelligent power supply unified management data, including the management strategy and scheduling optimization of each power supply system, and ensures that the entire power supply system operates efficiently under optimized conditions through unified management operations.

[0038] Preferably, step S1 comprises the following steps:

[0039] Step S11: performing initial power management data collection based on the power management system to generate initial power management data;

[0040] Step S12: performing initial power management analysis processing according to the initial power management data to generate initial power management analysis data;

[0041] Step S13: Designing a heterogeneous power management protocol stack according to the initial power management analysis data to generate a heterogeneous power management protocol stack;

[0042] Step S14: performing heterogeneous power data integration processing according to the heterogeneous power management protocol stack to generate integrated power data.

[0043] Through the collection of initial power management data, the system of the present invention can obtain information such as the operating status and energy consumption of different types of power supplies in real time, ensuring the integrity and accuracy of the basic management data. The initial power management analysis analyzes the historical management data of the power supply from multiple dimensions (including energy consumption, load, working status, etc.), ensuring that the collected initial power management data can fully reflect the operating status of the power supply, and converting the power management data into a standardized and structured data format. This data normalization ensures that different types of power supply data can be uniformly processed and analyzed in the same system, solving the problem of differences in heterogeneous power supply data structures. Different types of power supply data use different communication protocols and data formats. The heterogeneous power management protocol stack design can unify these differences and ensure that different power supplies can be seamlessly integrated and managed in the same system. This design solves the problem that heterogeneous power supplies are difficult to manage uniformly in traditional systems. Through the protocol stack design, it adapts to the new power supply connected in the future, regardless of the communication protocol and data structure it adopts, and provides good scalability and compatibility for the subsequent upgrade of the system. The unified protocol stack reduces the need for developers to write separate interfaces for each power supply device, reducing the development complexity and operation and maintenance costs of the system. The unified standard generated by the protocol stack can integrate the data of different types of power supplies, solve the problem of incompatibility of different power supply data, and make management more centralized and efficient. After the integrated processing, the use of power supplies can be analyzed and optimized from a global perspective, which improves the comprehensiveness of power management. After the heterogeneous power supply data is integrated, the system can flexibly dispatch different power supplies and make more flexible and intelligent decisions based on different power supply characteristics and real-time data, thereby improving energy utilization efficiency and reducing the waste of power resources.

[0044] In an embodiment of the present invention, multiple data acquisition devices are deployed in the power management system to monitor the power management of various heterogeneous power systems. The power systems include data management of photovoltaic power generation systems, wind power generation systems, battery energy storage systems, etc. The key monitoring parameters of each system include power management conditions such as power output voltage, current, power, temperature, energy consumption, and operating status. Through IoT devices or sensors, the operating data in different power systems are collected in real time into the power management system, the data streams sent by each power system are processed, and various power parameters are automatically recorded. The collected power management data is classified according to time sequence and power type, and the data is stored in a database, and the initial power management data is generated, including detailed information such as the working status, energy consumption, and output of each power system over a period of time. The collected initial power management data is parsed and processed. Different power systems may use different communication protocols and data formats, so the data must be uniformly parsed. The parsing module converts the original data (such as binary format, specific communication protocol data) in the initial power management data into a standardized format through specific conversion rules. During the parsing process, possible errors or abnormal data are detected and processed. For invalid or damaged data packets, the system will execute an error correction mechanism or mark anomalies to ensure the accuracy of subsequent data processing and generate initial power management parsing data. For example, the initial power management data may exist in binary format, JSON format, XML format, etc. The system converts these data formats into a standardized JSON format for subsequent processing. Each data packet contains a CRC checksum to detect errors during data packet transmission. If an error is detected, the system will re-request data or mark an abnormality. The normal range of voltage is set to 200V-500V. If the collected voltage data exceeds this range, the system will mark the data as abnormal. The output data unit of the photovoltaic system is W, the wind power system is kW, and the energy storage system is Wh. The output power of all power systems is converted to kW and recorded in the parsed data. Since different power systems or different power information use different communication protocols (such as Modbus, CAN, IEC 61850, etc.), a compatible heterogeneous power management protocol stack is designed to analyze the communication protocol required for each power system according to the characteristics of the parsed data. According to the network protocol stack model (such as the OSI seven-layer model), a management protocol stack suitable for the power supply system is designed. The design of the protocol stack needs to include the physical layer, data link layer, transport layer, application layer, etc. to ensure that different power supply systems can communicate with each other and exchange data.To ensure that the protocol stack can correctly handle the communication between heterogeneous power systems, compatibility testing is performed. The testing process includes verifying whether the protocol stack can correctly parse the protocol data of various power systems, ensuring that the protocol stack is compatible with multiple protocol formats, and can run efficiently in the power management system. After the design and testing are completed, a management protocol stack containing all heterogeneous power systems is generated, and various types of data are presented in a unified data structure (such as JSON format), so that the power management system can seamlessly communicate and interact with a variety of different power systems. After the heterogeneous power management protocol stack is deployed, the power management system communicates with different power systems through the protocol stack, receives data from each system, and collects and processes data from different power systems in a timely manner. The power data from different sources is integrated into a unified integrated power data, which contains detailed power data on the operating status of various power systems, and generates integrated power data.

[0045] Preferably, step S2 comprises the following steps:

[0046] Step S21: extracting the power energy consumption perception data identifier according to the integrated power data to generate the power energy consumption perception data;

[0047] Step S22: extracting the power state sensing data identifier according to the integrated power data to generate power state sensing data;

[0048] Step S23: using a preset hierarchical clustering algorithm to perform an initial multi-level power-aware clustering analysis on the power consumption awareness data and the power state awareness data to generate initial multi-level power-aware clustering data;

[0049] Step S24: performing abnormal clustering data screening processing based on the initial multi-level power-aware clustering data to generate multi-level power-aware clustering data;

[0050] Step S25: performing multi-level power-aware feature analysis based on the multi-level power-aware clustering data to generate multi-level power-aware feature data.

[0051] The present invention extracts power energy consumption perception data from integrated power data, accurately identifies the energy consumption characteristics of various types of power supplies, and helps the system to monitor the energy consumption of power supplies in real time. By extracting energy consumption data, the system can identify high-energy consumption power supplies and take targeted adjustment measures to optimize energy consumption distribution, which is helpful to improve the efficiency of overall energy consumption management. The extraction of power state perception data can reflect the operating status of various types of power supplies in real time, including the load, working mode and health status of the power supply, and provide a precise state monitoring basis for the system. After extracting the state perception data, the system can quickly identify the abnormal state or potential fault of the power supply, ensure the stable operation of the power supply, and reduce downtime. Through the preset hierarchical clustering algorithm, the energy consumption and state perception data are preliminarily clustered and analyzed. The system can classify the power supplies according to the characteristics of different power supplies (such as energy consumption and state), so that the system can distinguish different types of power supplies, which is convenient for subsequent management and scheduling. The preliminary clustering analysis lays the foundation for subsequent precise management and exception processing, and at the same time improves the data processing efficiency of the system. By screening out abnormal clustering data, power supply data that does not conform to normal rules during operation can be effectively eliminated, preventing abnormal data from affecting the management and scheduling decisions of the system. Significantly improve the accuracy and reliability of system data, ensure that subsequent analysis is based on normal data, and thus improve the correctness of power scheduling and decision-making. Multi-level power perception feature analysis can deeply explore the characteristics of power supply, such as load mode, energy consumption trend, etc., and accurately understand the operating characteristics of power supply from multiple dimensions.

[0052] As an embodiment of the present invention, refer to Figure 2 As shown, Figure 1 Detailed implementation step flow diagram of step S2 in the embodiment, step S2 in this embodiment includes:

[0053] Step S21: extracting the power energy consumption perception data identifier according to the integrated power data to generate the power energy consumption perception data;

[0054] In an embodiment of the present invention, on the basis of integrating power supply data, energy consumption parameters of various power supply systems are identified, and these parameters include: hourly energy consumption, total energy consumption, power consumption efficiency, etc. The energy consumption data of each power supply system will be distinguished by data identifiers to ensure that data from different sources can be correctly classified and processed. For example, each power supply system is assigned a unique ID, such as the photovoltaic system ID is "PV001", the wind power generation system ID is "WT001", and the energy storage system ID is "BAT001". Each energy consumption data is assigned an independent identifier, such as the identifier of hourly energy consumption is "Hour_Consumption", and the identifier of total energy consumption is "Total_Consumption". Energy consumption data is extracted according to specific identifiers of different power supply systems (such as power supply ID, equipment code, etc.), and an independent identifier is added to each energy consumption data to ensure that all energy consumption data can be correctly matched to the corresponding power supply system. After the identification extraction process, power supply energy consumption perception data containing energy consumption related features is generated. These data will provide basic support for subsequent power supply clustering and scheduling decisions.

[0055] Step S22: extracting the power state sensing data identifier according to the integrated power data to generate power state sensing data;

[0056] In an embodiment of the present invention, in addition to energy consumption data, the operating status of the power supply (such as startup, shutdown, failure, overload, etc.) is also a key management parameter. These status information are extracted from the integrated power supply data, and the operating status of each power supply is identified. A specific identifier is defined for each state (such as normal, abnormal, failure, etc.) to ensure that different power supply states can be uniformly identified. This identification extraction process automatically extracts and classifies the status code or alarm information sent when the power supply system is running. After the state identification is extracted, the power supply state perception data is generated, which reflects the real-time operation status of the power supply. These data provide key information for subsequent cluster analysis and help identify the operating modes of different power supplies.

[0057] Step S23: using a preset hierarchical clustering algorithm to perform an initial multi-level power-aware clustering analysis on the power consumption awareness data and the power state awareness data to generate initial multi-level power-aware clustering data;

[0058] In an embodiment of the present invention, a hierarchical clustering algorithm, such as agglomerative hierarchical clustering or divisive hierarchical clustering, is used to perform preliminary clustering analysis based on power supply energy consumption perception data and power supply status perception data. Hierarchical clustering gradually classifies data into different hierarchical structures by calculating the similarity between data points (such as Euclidean distance or cosine similarity). In the initial clustering stage, energy consumption perception data and status perception data are used as input, and multi-level clustering analysis is performed on each power supply system. The purpose of preliminary clustering is to classify power supplies with similar energy consumption and status characteristics into a group and identify power supply systems with similar operating modes. For example, agglomerative hierarchical clustering is selected, and Euclidean distance is used as the similarity metric, and hourly energy consumption data for the past 24 hours is collected as clustering input. The hourly energy consumption data of the photovoltaic system are: 3.2kWh, 3.5kWh, 2.8kWh, etc.; the hourly energy consumption of the wind power generation system is: 8.0kWh, 7.5kWh, 6.5kWh, etc. The number of state changes in the past 24 hours is used as the clustering feature, and the clustering threshold is set to classify the power sources with a Euclidean distance less than 2.0 into one category. The minimum number of clusters is set to 3 categories to ensure that the clustering results can reflect the different operating modes of the power system. After preliminary clustering, the initial multi-level power-aware clustering data is generated, which provides a basis for subsequent abnormal screening and further feature analysis.

[0059] Step S24: performing abnormal clustering data screening processing based on the initial multi-level power-aware clustering data to generate multi-level power-aware clustering data;

[0060] In an embodiment of the present invention, after the initial cluster analysis, the abnormal data in the clustering results are screened. These anomalies are caused by equipment failures, data transmission errors, etc., and the abnormal data usually appear as isolated points or abnormal clusters that are significantly deviated from other data points in the cluster. Abnormal data is identified using a preset threshold or statistical method (such as outlier detection or the 3σ principle). For example, the 3σ principle is used to set the standard deviation range of the data in each cluster to ±3σ. If the energy consumption or state characteristic data of a power supply exceeds this range, it is marked as abnormal. After the abnormal data is identified, these data are removed from the clustering results to avoid interference of abnormal data on subsequent analysis. After the abnormal data is screened out, more accurate multi-level power-aware clustering data is generated, which reflects the normal operating mode of the power supply system and eliminates potential abnormal influences.

[0061] Step S25: performing multi-level power-aware feature analysis based on the multi-level power-aware clustering data to generate multi-level power-aware feature data.

[0062] In an embodiment of the present invention, further feature analysis is performed on the basis of the multi-level power-aware clustering data that has been cleaned of abnormal data. The goal of the feature analysis is to extract key features that can characterize the operating status and energy consumption of the power system from the clustering results, such as average energy consumption, peak energy consumption, equipment failure rate, etc. An in-depth analysis is performed on each level in the clustering data to extract data reflecting the operating characteristics of power systems at different levels, such as extracting corresponding feature information for stratified user behavior energy consumption, environmental impact energy consumption, current distribution, and power status. Multi-level power-aware feature data is generated based on the extracted key features, which can comprehensively reflect the operating status, energy consumption pattern, etc. of the power system, providing an important basis for subsequent scheduling demand prediction and optimization.

[0063] Preferably, step S24 comprises the following steps:

[0064] Performing a multi-level clustering deviation data analysis based on the initial multi-level power-aware clustering data to generate multi-level clustering deviation data, wherein the multi-level clustering deviation data includes multi-level clustering deviation node data and multi-level clustering deviation quantification data;

[0065] Perform deviation node aggregation processing according to multi-level clustering deviation node data to generate deviation node aggregation data;

[0066] Perform abnormal power identification processing based on deviation node aggregation data and multi-level clustering deviation quantification data to generate abnormal power data;

[0067] The initial multi-level power-aware clustering data is screened for abnormal power according to the abnormal power data to generate multi-level power-aware clustering data, and the abnormal power data is transmitted to the terminal to perform abnormal power feedback operations.

[0068] The present invention can accurately identify the deviation points in the power system, especially the multi-level clustering deviation nodes and quantitative data, through the deviation analysis of the initial multi-level power perception clustering data, so that the system can more accurately detect and quantify the abnormal conditions of each node in the power system, and reduce misjudgment. The deviation node data and the deviation quantification data further refine the state of the power system, decompose the abnormal conditions into specific deviation amounts and deviation nodes, which is helpful for more accurate subsequent problem location and improves the perception ability of the power management system. Through the aggregation processing of the deviation nodes, the system can uniformly process the data of multiple deviation nodes, avoid the isolated analysis of individual abnormal nodes, and enable the system to integrate the information of multiple deviation nodes to form more representative aggregated data, thereby providing a more comprehensive basis for abnormal detection. The aggregation of deviation nodes helps to improve the efficiency of data analysis, avoid the error of a single node from being amplified, ensure that the system can effectively extract useful information from a large amount of data, and reduce the occupation of system resources. By combining the deviation node aggregation data with the deviation quantification data to identify abnormal power, the abnormal points in the power system are identified, and the abnormal power nodes that affect the operation of the power system can be found faster and more accurately. The use of a combination of aggregated data and quantitative data for anomaly identification can effectively reduce the probability of false alarms, improve the power management system's ability to judge abnormal situations, and ensure that the identified anomalies are actual problems. By screening out abnormal power supplies, the multi-level power-aware clustering data is further optimized, and abnormal data that may affect the accuracy of data analysis is removed, thereby improving the reliability and accuracy of the overall data. Abnormal power data is transmitted to the terminal, and feedback operations are performed to quickly respond to and process identified abnormal situations. Adjustments can be made in a timely manner to ensure the stable operation of the power management system. Through the screening and feedback mechanism of abnormal data, anomalies in the power system are processed in real time, ensuring that subsequent analysis and decisions are based on more accurate clustering data, further improving the overall efficiency of the power management system.

[0069] In an embodiment of the present invention, all clustering results are extracted from the initial multi-level power-aware clustering data, and the data deviation in each cluster group is calculated, including the degree of deviation of various features such as energy consumption and state. A deviation value is generated for each data point in each cluster according to its distance or similarity to the cluster center. The system marks the data points far from the cluster center as "deviation nodes" to form multi-level clustering deviation node data. To further quantify the degree of deviation, the deviation quantification data of each deviation node is calculated. The deviation quantification data can be generated by calculating the Euclidean distance, Mahalanobis distance or other distance measurement methods of each data point. The generated multi-level clustering deviation data describes the degree of deviation of each node and its relative position to other nodes. For example, the Euclidean distance is used as the main deviation measurement standard, and the distance threshold is set to 3.0. If the Euclidean distance between a node and the cluster center is greater than 3.0, it is marked as a "deviation node", and the average values ​​of the key features of the photovoltaic system (such as energy consumption, voltage, and temperature) at the cluster center are set to be: energy consumption 3.5kWh, output voltage 400V, and temperature 30°C. Assume that the characteristic values ​​of a node are: energy consumption 5.0kWh, output voltage 450V, temperature 40°C. By calculating its deviation value, it is 4.8, which exceeds the set threshold of 3.0, so it is marked as a "deviation node". After obtaining the multi-level clustering deviation node data, all deviation nodes are aggregated. This process includes classifying the nodes with deviation degrees in the same power management data and aggregating them into a group. The algorithm used in the aggregation process performs aggregation processing by device number or identifier classification. Through these aggregations, the complexity of analyzing a single node is reduced, and more meaningful abnormal trends are provided through the aggregated deviation group to generate aggregate data containing all deviation nodes. The deviation node aggregation data groups similar abnormal nodes to form a more obvious abnormal pattern. For example, the user end and the power transmission end of the same power management have abnormalities and are assigned to the same abnormal group. Based on the deviation node aggregation data and the deviation quantification data, each aggregation group is anomaly identified. The identification method includes threshold judgment or statistical analysis (such as the 3σ criterion) to identify aggregation nodes that deviate from the normal range. Different deviation thresholds are set for different types of power supplies. When the deviation quantification data of the deviation node aggregate data exceeds this preset threshold, the node is marked as an abnormal power supply. When one or more deviation nodes are identified as abnormal, the system generates abnormal power supply data, including detailed information of each abnormal power supply, such as power supply ID, deviation type, deviation degree, identification time, etc. Based on the identified abnormal power supply data, the data related to it in the initial multi-level power supply-aware clustering data is screened out. The screening process will record these abnormal data separately to ensure that they do not interfere with the normal power supply-aware clustering data.After filtering out the abnormal power data, the remaining cluster data is cleaned to ensure the integrity and consistency of the data. The cleaned multi-level power-aware cluster data will be used for subsequent power scheduling and optimization. After filtering and cleaning, multi-level power-aware cluster data without abnormal data is generated. All identified abnormal power data are transmitted to the designated terminal (such as the operation and maintenance center, control system) in real time. These data will include specific abnormal information of the power supply, such as deviation, abnormal type, identification time, etc. After receiving the abnormal power data, the terminal can start the feedback operation. The feedback operation can include alarm notification, manual inspection, automatic scheduling adjustment or shutdown of abnormal power supply. The feedback system will automatically execute the corresponding response mechanism according to the severity of the abnormal data, such as sending alarms to operation and maintenance personnel, or performing automatic adjustments to avoid further problems caused by power anomalies. The execution of all feedback operations will be recorded for subsequent analysis. The system can also further optimize the abnormal identification and processing process based on historical feedback data.

[0070] Preferably, step S3 comprises the following steps:

[0071] Step S31: performing local distribution division processing according to the multi-level power sensing feature data to generate local multi-level power sensing feature data;

[0072] Step S32: performing a local multi-level power sensing feature correlation analysis based on the local multi-level power sensing feature data to generate local multi-level power sensing correlation data;

[0073] Step S33: performing local multi-element power-aware driving correlation feature analysis based on the local multi-level power-aware correlation data to generate local power-aware driving correlation feature data.

[0074] The present invention divides the local distribution of multi-level power perception feature data and performs targeted management according to the power characteristics of different regions, which can not only better adapt to the power demand of each region, but also improve the accuracy of power scheduling. Local distribution division divides the overall data into smaller, manageable units, reduces the dependence on global data processing, thereby reducing the computational complexity of the system in large-scale data processing, improves processing efficiency, quickly responds to local power conditions, and optimizes and adjusts power scheduling or management strategies in a timely manner, thereby improving the dynamic response capability of the system. Correlation analysis of local multi-level power perception feature data can reveal the potential relationship between different power perception features, help understand the interaction of various variables in the power system, and provide an important basis for subsequent optimization. Correlation analysis can intelligently extract the intrinsic connection between features at different levels and help identify key factors in the local power system. Multivariate driven correlation feature analysis is performed on correlation data to reveal the main driving factors affecting the local power system. This in-depth analysis helps to understand the complex relationship between various parameters in the power system, find key variables that affect power scheduling or operating efficiency, and identify key driving factors that affect local power scheduling needs, so as to be more targeted in scheduling decisions and ensure the accuracy and efficiency of scheduling.

[0075] In an embodiment of the present invention, characteristic information about different power supplies is extracted from multi-level power supply perception feature data, and the characteristic information includes the energy consumption status, operating load, output efficiency, etc. of the power supply. According to the geographical location, device type, load condition, or grid connection condition of the power supply, the data of the entire power supply system is divided into several local areas. For example, local division is performed according to the geographical distribution location of each device type (photovoltaic energy storage device, power transmission equipment, etc.) of the power supply, and each area contains multiple power supply nodes. The data is divided according to predefined rules (for example, by region, device type, or voltage level). The power supply node data of each area is processed independently for subsequent local feature analysis. After the division is completed, the power supply data of each area is sorted to generate local multi-level power supply perception feature data, which represents the operating conditions of the power supply system in different local areas and provides basic data for the next step of feature correlation analysis. Based on the local distribution, the power supply perception feature data in each local area is subjected to correlation analysis. The correlation analysis reveals the relationship between different features (such as energy consumption, output power, and operating status). Correlation analysis tools (such as Pearson correlation coefficient, Spearman rank correlation coefficient, etc.) are used to measure the correlation between the variables in the local multi-level power perception feature data, quantify the linear or nonlinear association between the features, select Pearson correlation coefficient as the main analysis tool, and set the correlation threshold to ±0.6. If the absolute value of the correlation coefficient of two features exceeds 0.6, they are considered to have a strong correlation. In the case of some nonlinear associations, the Spearman rank correlation coefficient is used for supplementary analysis. The energy consumption state and output efficiency are calculated to have a correlation, and the Pearson correlation coefficient is -0.65, indicating that the energy consumption and output efficiency are negatively correlated. The operating load and energy consumption state are calculated to have a correlation coefficient of 0.72, indicating that the operating load and energy consumption state are positively correlated. The calculated correlation value indicates the strength of the relationship between each pair of features (usually ranging from -1 to 1). For example, a positive correlation indicates that the two features change in the same trend, and a negative correlation indicates that they change in the opposite trend. The local multi-level power perception correlation data is generated, which reveals the interaction between the key features in the local area and lays the foundation for the subsequent analysis of the driving correlation features. Based on the local multi-level power perception correlation data, further multi-element power perception driving correlation feature analysis is performed to analyze the dynamic change correlation relationship of power perception during power driving. Machine learning methods such as multi-element regression analysis, principal component analysis (PCA) or support vector machine (SVM) are used to analyze the correlation between multiple related features during power driving, and based on historical data, current feature performance and future demand forecasts, analyze which features have a significant impact on power scheduling, load fluctuations and operating efficiency.Through the feature-driven model, the driving features closely related to the power system scheduling can be identified. These features include the volatility of power demand, the failure rate of equipment, the impact of climate on new energy power sources, etc. These key related features will be screened and sorted for subsequent scheduling prediction and optimization control. For example, using multiple regression analysis, the target variable is set as "energy consumption state", and the independent variables are "output efficiency" and "operating load". The significance level threshold of the model is set to 0.05 to ensure that the analysis results are statistically significant. For nonlinear correlation features, principal component analysis (PCA) is used to reduce the dimension and extract the main driving features. The historical data of the past 30 days are selected for driving correlation feature analysis, with a time step of 1 hour, generating 720 data points. Using multiple regression analysis, the regression coefficient of output efficiency on energy consumption state is -0.35, the regression coefficient of operating load is 0.55, and the R² value of the model is 0.78, indicating that the model explains 78% of the energy consumption state changes. The PCA analysis of principal component 1 (representing the combined effect of output efficiency and operating load) explained 65% of the data variance, and principal component 2 explained 20% of the data variance. The key correlation features such as output efficiency (correlation: 0.65) and operating load (correlation: 0.55) were extracted and recorded to describe the correlation between the scheduling demand and each feature. For example, when the photovoltaic system is under high load (over 80%) and low output efficiency (below 70%), the energy consumption state shows high volatility. Local power supply perception-driven correlation feature data was generated to describe the correlation between the scheduling demand and each feature in the local power supply system.

[0076] Preferably, step S32 includes the following steps:

[0077] Performing local multi-level power-aware dependency feature analysis based on the local multi-level power-aware feature data to generate local multi-level power-aware dependency feature data;

[0078] Performing local multi-level power perception feature difference analysis based on the local multi-level power perception feature data to generate local multi-level power perception feature difference data;

[0079] Performing a local multi-level power-aware correlation covariance matrix analysis based on the local multi-level power-aware dependency feature data and the local multi-level power-aware feature difference data to generate a local multi-level power-aware correlation covariance matrix;

[0080] A local multi-level power-aware feature correlation analysis is performed according to the local multi-level power-aware correlation covariance matrix to generate local multi-level power-aware correlation data.

[0081] The present invention identifies the intrinsic dependency between different local power perception features through dependency feature analysis, which helps the power management system to better understand the interaction between various variables, thereby optimizing power scheduling and clearly identifying the key dependent variables in the local power system, which provides a more accurate basis for scheduling decisions, improves the accuracy and efficiency of power scheduling, foresees the mutual influence between different power variables, improves the responsiveness of the system in a dynamic environment, and can make timely adjustments to changes. According to the local multi-level power perception feature data, the local multi-level power perception feature difference analysis is performed, and the differences of similar features in different power systems can be identified. Understanding these differences can help the power management system adjust the power scheduling strategy locally, so as to better adapt to the specific conditions of each local system. Through difference analysis, inconsistent feature performance is identified, and detailed feature difference data support is provided, so that the power can be optimized and managed more accurately. According to the local multi-level power perception dependency feature data and the local multi-level power perception feature difference data, the local multi-level power perception correlation covariance matrix analysis is performed, which provides the comprehensive correlation of different perception features in the local power system, reveals the mutual dependence between each feature, and provides comprehensive information support for optimizing scheduling. The covariance matrix can quantify the correlation between features, so as to achieve more accurate power management. Performing local multi-level power-aware feature correlation analysis based on the local multi-level power-aware correlation covariance matrix can further refine and optimize the correlation of local multi-level power-aware features, help the system discover the most critical correlation features, provide more valuable reference information for scheduling decisions, analyze the correlation between different power features, and thus design a more reasonable power scheduling plan, improving the scientificity and rationality of scheduling decisions.

[0082] In an embodiment of the present invention, key feature data (such as energy consumption, load, voltage, temperature, etc.) are extracted from local multi-level power-aware feature data, and these data are used as input for dependency feature analysis. According to preset rules or dependency models, the dependency between power-aware features is determined, such as an increase in load may lead to an increase in energy consumption, voltage fluctuations, etc. Therefore, there may be dependencies between load and features such as energy consumption and voltage. Statistical methods or machine learning models (such as regression analysis, decision trees, etc.) are used to calculate the dependencies between various features. Features with high dependencies will be marked as key dependency features, and local multi-level power-aware dependency feature data are generated to describe the dependencies between power features, providing a basis for the system to understand the mutual influence between features and helping to optimize power scheduling in the future. Different subsets of the same type of features are selected for difference analysis. For example, the energy consumption, output power, temperature and other features of the same type of power supply in a certain period of time may show differences, and these data will be compared and analyzed. Statistical analysis methods (such as mean difference, variance analysis, etc.) are used to measure the differences between different power features. For each feature, the size of the difference value is calculated. Large differences may indicate abnormal phenomena or different operating modes in the power system. Based on the calculated difference metric, the features are divided into categories with large differences and small differences. The feature data with large differences will be marked as key monitoring objects, and local multi-level power-aware feature difference data will be generated, which shows the differences of the same type of features under different power nodes or operating conditions. Combined with the local multi-level power-aware dependent feature data and the local multi-level power-aware feature difference data, a correlation covariance matrix is ​​constructed. The covariance matrix is ​​used to quantify the strength of the correlation between each feature, especially between multiple dependent features or difference features. Each element of the covariance matrix represents the covariance value between two features, reflecting their linear correlation. A positive covariance indicates that the features change in the same direction, while a negative covariance indicates that the features change in the opposite direction. Based on the results of the covariance matrix, the features with low correlation or irrelevant are screened to simplify the subsequent analysis and processing. The features with high correlation will be further analyzed to generate an optimized feature subset, and a local multi-level power-aware correlation covariance matrix is ​​generated to provide the linear relationship between the features in the power system. Based on the results of the covariance matrix, a more in-depth analysis of the correlation between features is performed, and the Pearson correlation coefficient or other correlation indicators are used to evaluate the correlation strength between each pair of features. Highly correlated features are identified as key features, prompting the system to focus on the changes in these features during the scheduling process, and generate local multi-level power-aware correlation data. Based on these correlation data, power scheduling and load distribution are further optimized. The generated local multi-level power-aware correlation data reflects the intrinsic relationship between power features.

[0083] Preferably, step S33 includes the following steps:

[0084] Performing history-driven local multi-level power-aware correlation feature analysis on the local multi-level power-aware correlation data to generate local multi-level power-aware driven correlation feature data;

[0085] A local multi-level power-aware driving correlation feature analysis is performed based on the local multi-level power-aware driving correlation feature data to generate local power-aware driving correlation feature data.

[0086] The present invention performs a history-driven local multi-level power perception correlation feature analysis on local multi-level power perception correlation data, which can identify the long-term correlation between power perception features based on past power operation data. The historical analysis can help the system understand the changing trend of power characteristics from a longer time dimension, improve the accuracy and stability of the prediction, and provide a more accurate prediction basis for future power scheduling, especially for the prediction of power demand, and foresee potential power scheduling needs in advance, so as to make more reasonable resource allocation. According to the local multi-level power perception driven correlation feature data, the local multivariate power perception driven correlation feature analysis is performed, and the key correlation features in the power system are mined from multiple dimensions, revealing the complex correlation between each power perception feature and the actual power scheduling demand, providing comprehensive support for intelligent scheduling. Through the in-depth analysis of the correlation features, the key driving factors affecting power scheduling can be more accurately identified, which helps to make more accurate scheduling decisions based on real-time data, thereby improving the operating efficiency and reliability of the system, and can dynamically adjust and optimize the demand for local power based on the correlation features, reduce energy waste, and improve the overall performance of the power system.

[0087] In an embodiment of the present invention, historical records of local power perception correlation data are collected. These historical data may include energy consumption, power output, operating status, load fluctuation, etc. of the power supply. The historical data can be traced back to different time periods of system operation to observe long-term characteristic trends. The historical data is modeled using time series analysis tools (such as ARIMA models or LSTM neural networks) to identify long-term change trends and periodic characteristics. For example, some power characteristics may show periodic fluctuations during high load or low load periods. By analyzing the changes in the features in the historical data, it is identified which features have significantly changed in correlation compared to the current state, and local multi-level power perception driven correlation feature data is generated according to the correlation status of the feature changes driven by the history. The local multi-level power perception driven correlation feature data is used to further perform local multivariate power perception driven correlation feature analysis. During the analysis process, multivariate statistical methods (such as multivariate regression analysis, principal component analysis PCA, etc.) or machine learning models (such as random forests or neural networks) are used to identify the complex correlation relationship between features in the power driving process. After the multivariate correlation feature analysis, local power perception driven correlation feature data is generated. The local power perception driven correlation feature data reflects the changes and correlation changes of each perception feature in the power scheduling process.

[0088] Preferably, step S4 comprises the following steps:

[0089] Step S41: performing local power driving feature analysis according to the local multi-level power sensing feature data to generate local power driving feature data;

[0090] Step S42: performing local power-aware driving distributed node analysis according to the local power-aware driving associated feature data, generating local power-aware driving distributed node data, and designing a decision tree distributed node through the local power-aware driving distributed node data;

[0091] Step S43: Based on the preset XGBoost model, a distributed gradient boosting decision tree design of power drive and power scheduling requirements is performed on the decision tree distributed nodes and the local power perception drive associated feature data to obtain a distributed decision tree model of power scheduling requirements;

[0092] Step S44: designing an intelligent monitoring engine for local power dispatch demand prediction according to the distributed decision tree model of power dispatch demand, so as to obtain a local power dispatch demand prediction engine;

[0093] Step S45: transmitting the local power driving characteristic data to the local power dispatching demand prediction engine in real time to perform intelligent analysis of the local dynamic power dispatching demand, and generate local dynamic power dispatching demand data;

[0094] Step S46: performing automated decision analysis of global intelligent power management according to the local dynamic power scheduling demand data to generate a global intelligent power management decision;

[0095] Step S47: performing intelligent power unified management adjustment processing on the initial power management data according to the global intelligent power management decision, generating intelligent power unified management data, and executing the intelligent power unified management operation through the intelligent power unified management data.

[0096] The present invention performs local power drive feature analysis based on local multi-level power perception feature data, deeply explores the physical driving factors in the local power perception features, clarifies which factors have a key impact on the operation of the local power system, and provides accurate feature data support for subsequent power scheduling. By real-time monitoring and analyzing the local power drive features, the changes in the local power system can be quickly identified and responded to in a timely manner, thereby improving the efficiency of overall power management. According to the local power perception drive associated feature data, the local power perception drive distributed node is divided, and the distributed decision nodes adapted to different local power requirements can be identified and designed, the flexibility of node management is improved, resource allocation is optimized, the generation of distributed node data and the design of decision tree nodes are enabled, so that the system can realize distributed power scheduling on multiple nodes, improve the flexibility and accuracy of scheduling, provide highly customized decision support, ensure that the decision of power scheduling can effectively adapt to local conditions, and reduce power waste. Based on the preset XGBoost model, the distributed gradient boosting decision tree design of power drive and power scheduling requirements is performed on the decision tree distributed nodes and the local power perception drive associated feature data. By introducing the XGBoost model, accurate power scheduling demand prediction is performed based on the decision tree structure, and the gradient boosting algorithm can show strong accuracy and flexibility in the prediction of local power scheduling demand. The XGBoost model can efficiently process large-scale power perception data, quickly train models, and generate scheduling demand prediction results, ensuring that the system can respond to various power scheduling requirements in a timely manner and make dynamic adjustments during operation, thereby optimizing power scheduling and improving the overall efficiency of the power system. An intelligent monitoring engine is designed to predict local power scheduling requirements based on the distributed decision tree model of power scheduling requirements, and predict local power scheduling requirements in real time, so as to take scheduling measures in advance. The monitoring engine can dynamically update the prediction results based on real-time data to ensure that the power system responds quickly to changing requirements. The intelligent monitoring engine realizes the automatic prediction and control of local power scheduling, reduces manual intervention, and improves the automation level of power management. With the help of the distributed decision tree model, the monitoring engine can make more accurate predictions to ensure that the scheduling requirements of the local power system in different periods are met. The local power driving feature data is transmitted to the local power scheduling demand prediction engine in real time for intelligent analysis of local dynamic power scheduling requirements, realizing dynamic power scheduling demand analysis, ensuring that the system can quickly respond to changes in power demand and make timely scheduling adjustments. Intelligent analysis can accurately predict power demand at each moment, avoiding power waste or shortage due to inaccurate predictions, thereby improving the efficiency of power use. Real-time transmission and analysis can enable the system to respond more flexibly to changes in demand, ensure the rapid response capability of power scheduling, and avoid power system failures caused by delays.Automated decision analysis of global intelligent power management based on local dynamic power scheduling demand data can combine local scheduling requirements with global power management to achieve intelligent optimization of global power management. Global decisions are based on local dynamic scheduling requirements to ensure coordination between global and local power scheduling, avoid conflicts between local requirements and global plans, improve overall scheduling efficiency, and dynamically adjust global power management strategies based on actual needs and real-time data, thereby improving the adaptability of power management and the effectiveness of the overall system. Through intelligent power unified management decisions, the initial power management data can be comprehensively adjusted to ensure efficient and unified management of the entire power system, avoiding the dispersion and incoordination of power scheduling in various parts. Intelligent power unified management operations can achieve global intelligent control of the power system, ensuring that the system can maintain coordination and consistency in overall operation, thereby further optimizing power use and scheduling efficiency.

[0097] As an embodiment of the present invention, refer to Figure 3 As shown, Figure 1 Detailed implementation step flow chart of step S4 in the embodiment, step S4 in this embodiment includes:

[0098] Step S41: performing local power driving feature analysis according to the local multi-level power sensing feature data to generate local power driving feature data;

[0099] In the embodiment of the present invention, key features (such as energy consumption, load, output power, etc.) are extracted from the local multi-level power perception feature data. These features are the basis for analyzing the power drive characteristics. Multiple regression analysis or support vector machine and other methods are used to analyze the physical operation status of the power drive under the power operation status. The output power is used as the target variable, and the energy consumption and load are used as independent variables to establish a multiple regression model. The significance level of the model is set to 0.05. If the regression coefficient is significantly lower than the threshold, it is regarded as a significant influencing feature, that is, the power drive feature. The power drive feature includes voltage fluctuations and equipment operation status under a specific load mode, etc., and generates local power drive feature data, including the physical operation status of the power drive under each power usage scenario.

[0100] Step S42: performing local power-aware driving distributed node analysis according to the local power-aware driving associated feature data, generating local power-aware driving distributed node data, and designing a decision tree distributed node through the local power-aware driving distributed node data;

[0101] In an embodiment of the present invention, based on the local power supply perception drive associated feature data, the features of different power supply nodes are distributedly analyzed to identify the scheduling requirements and behavior patterns of each power supply node in different power consumption scenarios. Distributed node data is generated for each power supply node, and a decision tree structure is designed based on these data. Each node of the decision tree represents the driving behavior of a power supply system, ensuring that different power supply nodes can work independently or collaboratively according to their respective characteristics. The generated distributed node data contains the scheduling requirements, driving characteristics, and correlation with other nodes of each node. For example, each power supply node is regarded as an independent distributed node, and its scheduling requirements (such as energy consumption, load, state, etc.) are extracted based on the local power supply drive feature data, and the correlation between nodes is identified based on the associated feature data. The correlation threshold is set to ±0.6. If the scheduling requirement correlation of two nodes exceeds 0.6, they are deemed to have a collaborative work requirement. Each decision tree node represents the state of a power supply node. When designing a decision tree, the maximum depth is set to 5 and the minimum number of split samples is set to 10 to ensure the generalization ability of the model.

[0102] Step S43: Based on the preset XGBoost model, a distributed gradient boosting decision tree design of power drive and power scheduling requirements is performed on the decision tree distributed nodes and the local power perception drive associated feature data to obtain a distributed decision tree model of power scheduling requirements;

[0103] In an embodiment of the present invention, a preset XGBoost model is used, and the local power-sensing drive-related feature data is used as input, combined with the feature data of distributed nodes, and the model is gradually optimized by the gradient boosting method. By training the XGBoost model, the prediction accuracy of the model for power scheduling requirements is gradually improved, and each training step is verified according to the actual power scheduling demand feedback data. The model performance is evaluated using 10-fold cross-validation, and the evaluation index is set to mean square error (MSE). If the MSE is lower than 0.2, it is considered that the model has good prediction accuracy, and the model parameters are adjusted to generate a distributed decision tree model for power scheduling requirements. The model can predict the scheduling requirements in various power consumption scenarios based on real-time data and dynamically optimize the power scheduling strategy.

[0104] Step S44: designing an intelligent monitoring engine for local power dispatch demand prediction according to the distributed decision tree model of power dispatch demand, so as to obtain a local power dispatch demand prediction engine;

[0105] In an embodiment of the present invention, an intelligent monitoring engine is designed based on a distributed decision tree model. The engine can analyze the operating status of a local power supply in real time and predict scheduling requirements. The intelligent monitoring engine is integrated into a power management system to perform real-time monitoring and scheduling prediction tests. The engine's prediction capabilities are verified under different load conditions and optimized and adjusted. After the design of the intelligent monitoring engine is completed, the system can predict the scheduling requirements of the local power supply through the engine and achieve dynamic scheduling optimization.

[0106] Step S45: transmitting the local power driving characteristic data to the local power dispatching demand prediction engine in real time to perform intelligent analysis of the local dynamic power dispatching demand, and generate local dynamic power dispatching demand data;

[0107] In an embodiment of the present invention, driving characteristic data (such as energy consumption, load, state change, etc.) of a local power supply is transmitted to an intelligent monitoring engine in real time through an IoT device to ensure low latency and high reliability of data during transmission. The monitoring engine performs dynamic analysis based on the real-time transmitted data to predict upcoming scheduling needs. Through the prediction model in the engine, the system can adjust the power scheduling strategy in real time to ensure stable operation of the system, and generate local dynamic power scheduling demand data based on the analysis results as a basis for intelligent scheduling decisions.

[0108] Step S46: performing automated decision analysis of global intelligent power management according to the local dynamic power scheduling demand data to generate a global intelligent power management decision;

[0109] In an embodiment of the present invention, local dynamic power scheduling demand data is combined with a global power management strategy, that is, the dynamic power scheduling demand under each power usage scenario is globally connected and analyzed, and the global power scheduling is optimized to generate a global power management decision to ensure that the power system can be coordinated and scheduled according to local and global needs to achieve the best balance between power efficiency and system stability, and generate a global intelligent power management decision as the basis for global power scheduling.

[0110] Step S47: performing intelligent power unified management adjustment processing on the initial power management data according to the global intelligent power management decision, generating intelligent power unified management data, and executing the intelligent power unified management operation through the intelligent power unified management data.

[0111] In the embodiment of the present invention, according to the global intelligent power management decision, the system uniformly manages and adjusts the initial power management data, including adjusting load distribution, optimizing power use strategy, reconfiguring the scheduling plan of each power system, etc., and generates intelligent power unified management data, including the scheduling plan, load distribution and operating parameters of each power system. For example, the load demand of each power user and the load distribution of each power generation node are adjusted, and the intelligent power management operation is performed according to the generated unified management data to ensure that the scheduling and management of the power system are operated with the optimal strategy.

[0112] Preferably, step S43 includes the following steps:

[0113] Based on the preset XGBoost model, the distributed nodes of the decision tree are designed with distributed gradient boosting decision tree architecture for power drive and power scheduling requirements, and the distributed decision tree model architecture for power scheduling requirements is generated;

[0114] According to the local power perception drive correlation characteristic data, the fluctuation characteristics of power drive and power dispatch demand are analyzed to generate power drive dispatch demand fluctuation characteristic data;

[0115] According to the power drive scheduling demand fluctuation characteristic data, the power dispatch demand distributed decision tree model architecture is used to perform decision tree gradient feature learning processing of power drive and power dispatch demand to generate a distributed decision tree model of power dispatch demand.

[0116] The present invention designs a distributed gradient boosting decision tree architecture for power drive and power scheduling requirements for decision tree distributed nodes based on a preset XGBoost model. The XGBoost model can efficiently process large-scale data, not only has high computational efficiency, but also can automatically optimize hyperparameters in scheduling demand prediction, thereby accelerating the model training process and improving prediction accuracy. The design of the distributed decision tree model architecture has good scalability and can adapt to power management systems of different scales, ensuring that it can work efficiently even in distributed power systems of large-scale nodes. In addition, the XGBoost model is designed based on a gradient boosting tree model, which can gradually optimize the prediction of power scheduling requirements, reduce overfitting, and improve the generalization ability of the model, ensuring that the system can make accurate scheduling decisions in different scenarios. The volatility of the power perception driving characteristics is analyzed, and the system can capture the fluctuation characteristics of the power scheduling requirements. This is very important for coping with dynamic changes in power demand, especially in complex power systems, where volatility analysis can help identify potential scheduling challenges. The identification of the fluctuation characteristics of power scheduling requirements can foresee possible demand fluctuations in advance, and make corresponding adjustments in the scheduling strategy to ensure the stable operation of the power system and reduce the occurrence of power shortages or overloads. The power-driven dispatch demand fluctuation characteristic data provides real-time information on power demand changes, allowing the dispatch strategy to be adjusted dynamically according to changes in demand, so as to better adapt to changes in power demand in different time periods or regions. Gradient feature learning of the power-driven dispatch demand fluctuation characteristic data can continuously optimize the distributed decision tree model according to the fluctuation law of power dispatch demand, help the model to continuously update the decision basis, and improve the accuracy of dispatch demand prediction. The introduction of gradient feature learning ensures that the decision tree model can continuously evolve during the operation of the power system, so that it can adapt to the changing power demand situation. This continuously optimized model significantly improves the flexibility and effectiveness of power dispatch. The distributed decision tree model of power dispatch demand can cope with more complex and uncertain power dispatch scenarios, enabling the system to make robust dispatch decisions under various uncertain factors, ensuring the safety and efficiency of power management.

[0117] In the embodiment of the present invention, an XGBoost model is pre-designed, which is a machine learning model based on gradient boosting, and is used to process large-scale power scheduling data. The XGBoost model parameter settings are as follows: learning rate: set to 0.1, to control the step size of each update, to avoid overfitting of the model while ensuring the training speed; the maximum depth of the tree: set to 6, to ensure that the model has sufficient complexity on each power node to adapt to different power characteristics; subsample ratio: set to 0.8, to randomly select 80% of the data for training each tree, to reduce the risk of overfitting; feature column sampling ratio: set to 0.7, to randomly select 70% of the features for training each tree, to increase the generalization ability of the model. Minimum sample split number: set to 10, to ensure that at least 10 sample data are included when the node is split, to avoid too fine a split resulting in excessive model complexity. The driving features are extracted from the perception data of the local power source, and the basic architecture of the decision tree model is initialized. For different power nodes, an independent decision tree structure is configured for each node. For example, the decision tree of the photovoltaic node has a depth of 3 layers, with output power, load, and temperature as the main features; the decision tree of the wind node has a depth of 4 layers, with wind speed, output power, and equipment status as the main features. The decision tree of each node is adjusted through the training of the XGBoost model to adapt to the specific power characteristics and scheduling requirements. Multiple independent decision trees are combined into a distributed decision tree model architecture, and the efficient management of power drive and scheduling requirements is achieved through the synergy of multiple power nodes. This architecture can cope with the complex scheduling requirements between power nodes, and optimize the overall performance through the collaborative work of distributed nodes. A complete distributed decision tree model architecture for power scheduling requirements is generated, which is used to process the scheduling requirements of each power node in real time and optimize power scheduling globally. The fluctuation information of power scheduling requirements is extracted from the local power perception drive association feature data, including data fluctuations in energy consumption, power output, load, etc. The fluctuation characteristics are the key reference factors in the power scheduling process, reflecting the changing trend of power load. Use time series analysis methods (such as moving average, Fourier transform, etc.) to analyze the fluctuation characteristics of power scheduling demand and identify the changing rules of power load in different time periods. These fluctuation data will be marked as key features by the system, and the fluctuation characteristic data of power-driven scheduling demand will be generated according to the analysis results. Through the gradient boosting algorithm of XGBoost, the distributed decision tree model is feature learned. The input data is the previously generated power-driven scheduling demand fluctuation characteristic data. The parameters of the decision tree model are gradually adjusted to enable it to better adapt to the changing needs of power scheduling. The decision tree model is continuously trained in the actual power scheduling process. By comparing the error between the predicted results and the actual scheduling demand, the gradient weight of the model is adjusted. For example, using the gradient boosting algorithm of XGBoost, the input data is the fluctuation characteristic data generated in the previous step.The ratio of the training set to the validation set is set to 8:2 to ensure that the model can be effectively learned and validated. The feature weights are adjusted according to the prediction error in each iteration. If the prediction error of the PV node output power exceeds 10%, the weight of the output power feature is increased, and the increment is set to 0.1. The maximum depth of the decision tree is gradually adjusted during the training process. If the mean square error (MSE) of the validation set continues to decrease, the depth is increased (up to 10); if the MSE increases, the depth is reduced (to a minimum of 3). The splitting rule is set as follows: if the energy consumption fluctuation feature exceeds the set threshold (for example, ±15%), the node is split to create a new decision path. By comparing the error between the prediction result and the actual scheduling demand, a loss function (such as square loss) is used to adjust the gradient weight of the model. An early stopping strategy is set. When the MSE of the validation set does not improve in 5 consecutive iterations, training is stopped to prevent overfitting. The model performance is evaluated using a cross-validation method (such as 5-fold cross-validation). The threshold of MSE is set to 0.05. If the error is lower than this threshold, the model is considered to have a high prediction accuracy. The distributed decision tree model is optimized, including adjusting the depth of the tree, node splitting rules and weight distribution, etc. Through this continuous learning and optimization, the model can more accurately predict future power scheduling needs. After the learning process is completed, a distributed decision tree model for power scheduling needs is generated. The model can intelligently predict power scheduling needs based on the fluctuation characteristics of power scheduling needs, and optimize the scheduling efficiency of power through the synergy of distributed nodes.

[0118] This specification provides an intelligent power unified management method system for executing the intelligent power unified management method as described above. The intelligent power unified management system includes:

[0119] An integrated power data collection module is used to collect initial power management data based on the power management system to generate initial power management data; perform heterogeneous power data integration processing based on the initial power management data to generate integrated power data;

[0120] A multi-level power-aware feature analysis module is used to perform multi-level power-aware cluster analysis based on the integrated power data to generate multi-level power-aware cluster data; perform multi-level power-aware feature analysis based on the multi-level power-aware cluster data to generate multi-level power-aware feature data;

[0121] A local power-sensing driving correlation feature analysis module is used to perform local distribution division processing based on multi-level power-sensing feature data to generate local multi-level power-sensing feature data; perform local multi-level power-sensing driving correlation feature analysis based on the local multi-level power-sensing feature data to generate local power-sensing driving correlation feature data;

[0122] The intelligent power unified management module is used to perform local power drive feature analysis based on local multi-level power perception feature data to generate local power drive feature data; design an intelligent monitoring engine for local power scheduling demand prediction based on local power perception drive associated feature data to obtain a local power scheduling demand prediction engine; transmit the local power drive feature data in real time to the local power scheduling demand prediction engine to perform intelligent analysis of local dynamic power scheduling demand to generate local dynamic power scheduling demand data; perform automated decision analysis of global intelligent power management based on local dynamic power scheduling demand data to generate global intelligent power management decisions; perform intelligent power unified management adjustment processing on the initial power management data based on the global intelligent power management decision to generate intelligent power unified management data, and execute intelligent power unified management operations through the intelligent power unified management data.

[0123] The beneficial effect of the present application is that the intelligent power unified management method of the present invention realizes the unified integrated management of data of different types of power supplies through the collection and analysis of initial power management data and the design of heterogeneous power management protocol stacks, and can integrate decentralized power management systems into a whole, thereby enhancing the coordination of the system. Through multi-level power perception feature data, local scheduling demand prediction engine, and global intelligent management decision analysis, it can effectively generate global intelligent power management decisions and realize global unified scheduling of power supplies. This global scheduling strategy can ensure the coordinated operation of multiple heterogeneous power supplies, greatly improve the efficiency of power scheduling, and avoid the problems of unreasonable scheduling and energy waste caused by decentralized management. Through multi-level power perception clustering analysis and feature analysis, the system can fully perceive the power consumption and state characteristics, respond to the diverse needs of different power usage scenarios, and screen out abnormal power nodes and power supplies through abnormal clustering data, thereby improving the accuracy and robustness of the system. By introducing a local power scheduling demand prediction engine, the driving characteristics of local power are analyzed in real time, and future scheduling needs are predicted. The scheduling prediction of the XGBoost model can dynamically adjust the power usage strategy to ensure efficient resource scheduling when the power usage scenario changes. According to the local power scheduling needs and the automated decision-making analysis of global data, management decisions can be automatically generated and power scheduling can be executed according to the real-time power usage scenario needs, which improves the system's ability to cope with complex scenarios, significantly reduces energy consumption, and improves system stability. In this way, automated unified power management is performed for heterogeneous power data and complex power usage scenarios, reducing unnecessary power consumption and achieving more efficient and intelligent unified power management.

[0124] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0125] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for unified management of intelligent power supply, characterized in that: The following steps are involved: Step S1: performing initial power management data collection based on the power management system to generate initial power management data; Perform heterogeneous power data integration processing based on initial power management data to generate integrated power data; Step S2: performing a multi-level power-aware cluster analysis based on the integrated power data to generate multi-level power-aware cluster data; performing a multi-level power-aware feature analysis based on the multi-level power-aware cluster data to generate multi-level power-aware feature data; Step S3: performing local distribution division processing according to the multi-level power perception feature data to generate local multi-level power perception feature data; performing local multi-element power perception drive correlation feature analysis based on the local multi-level power perception feature data to generate local power perception drive correlation feature data; Step S4: performing local power driving feature analysis according to the local multi-level power sensing feature data to generate local power driving feature data; An intelligent monitoring engine is designed to predict local power scheduling requirements based on local power perception-driven associated feature data to obtain a local power scheduling requirement prediction engine; local power driving feature data is transmitted in real time to the local power scheduling requirement prediction engine for intelligent analysis of local dynamic power scheduling requirements to generate local dynamic power scheduling requirement data; automated decision analysis of global intelligent power management is performed based on the local dynamic power scheduling requirement data to generate a global intelligent power management decision; initial power management data is adjusted and processed for intelligent unified power management based on the global intelligent power management decision to generate intelligent unified power management data, and intelligent unified power management operations are executed through the intelligent unified power management data.

2. The intelligent power supply unified management method according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: performing initial power management data collection based on the power management system to generate initial power management data; Step S12: performing initial power management analysis processing according to the initial power management data to generate initial power management analysis data; Step S13: Designing a heterogeneous power management protocol stack according to the initial power management analysis data to generate a heterogeneous power management protocol stack; Step S14: performing heterogeneous power data integration processing according to the heterogeneous power management protocol stack to generate integrated power data.

3. The intelligent power supply unified management method according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: extracting the power energy consumption perception data identifier according to the integrated power data to generate the power energy consumption perception data; Step S22: extracting the power state sensing data identifier according to the integrated power data to generate power state sensing data; Step S23: using a preset hierarchical clustering algorithm to perform an initial multi-level power-aware clustering analysis on the power consumption awareness data and the power state awareness data to generate initial multi-level power-aware clustering data; Step S24: performing abnormal clustering data screening processing based on the initial multi-level power-aware clustering data to generate multi-level power-aware clustering data; Step S25: performing multi-level power-aware feature analysis based on the multi-level power-aware clustering data to generate multi-level power-aware feature data.

4. The intelligent power supply unified management method according to claim 3, characterized in that: Step S24 includes the following steps: Performing a multi-level clustering deviation data analysis based on the initial multi-level power-aware clustering data to generate multi-level clustering deviation data, wherein the multi-level clustering deviation data includes multi-level clustering deviation node data and multi-level clustering deviation quantification data; Perform deviation node aggregation processing according to multi-level clustering deviation node data to generate deviation node aggregation data; Perform abnormal power identification processing based on deviation node aggregation data and multi-level clustering deviation quantification data to generate abnormal power data; The initial multi-level power-aware clustering data is screened for abnormal power according to the abnormal power data to generate multi-level power-aware clustering data, and the abnormal power data is transmitted to the terminal to perform abnormal power feedback operations.

5. The intelligent power supply unified management method according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: performing local distribution division processing according to the multi-level power sensing feature data to generate local multi-level power sensing feature data; Step S32: performing a local multi-level power sensing feature correlation analysis based on the local multi-level power sensing feature data to generate local multi-level power sensing correlation data; Step S33: performing local multi-element power-aware driving correlation feature analysis based on the local multi-level power-aware correlation data to generate local power-aware driving correlation feature data.

6. The intelligent power supply unified management method according to claim 5, characterized in that: Step S32 includes the following steps: Performing local multi-level power-aware dependency feature analysis based on the local multi-level power-aware feature data to generate local multi-level power-aware dependency feature data; Performing local multi-level power perception feature difference analysis based on the local multi-level power perception feature data to generate local multi-level power perception feature difference data; Performing a local multi-level power-aware correlation covariance matrix analysis based on the local multi-level power-aware dependency feature data and the local multi-level power-aware feature difference data to generate a local multi-level power-aware correlation covariance matrix; A local multi-level power-aware feature correlation analysis is performed according to the local multi-level power-aware correlation covariance matrix to generate local multi-level power-aware correlation data.

7. The intelligent power supply unified management method according to claim 5, characterized in that: Step S33 includes the following steps: Performing history-driven local multi-level power-aware correlation feature analysis on the local multi-level power-aware correlation data to generate local multi-level power-aware driven correlation feature data; A local multi-level power-aware driving correlation feature analysis is performed based on the local multi-level power-aware driving correlation feature data to generate local power-aware driving correlation feature data.

8. The intelligent power supply unified management method according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: performing local power driving feature analysis according to the local multi-level power sensing feature data to generate local power driving feature data; Step S42: performing local power-aware driving distributed node analysis according to the local power-aware driving associated feature data, generating local power-aware driving distributed node data, and designing a decision tree distributed node through the local power-aware driving distributed node data; Step S43: Based on the preset XGBoost model, a distributed gradient boosting decision tree design of power drive and power scheduling requirements is performed on the decision tree distributed nodes and the local power perception drive associated feature data to obtain a distributed decision tree model of power scheduling requirements; Step S44: designing an intelligent monitoring engine for local power dispatch demand prediction according to the distributed decision tree model of power dispatch demand, so as to obtain a local power dispatch demand prediction engine; Step S45: transmitting the local power driving characteristic data to the local power dispatching demand prediction engine in real time to perform intelligent analysis of the local dynamic power dispatching demand, and generate local dynamic power dispatching demand data; Step S46: performing automated decision analysis of global intelligent power management according to the local dynamic power scheduling demand data to generate a global intelligent power management decision; Step S47: performing intelligent power unified management adjustment processing on the initial power management data according to the global intelligent power management decision, generating intelligent power unified management data, and executing the intelligent power unified management operation through the intelligent power unified management data.

9. The intelligent power supply unified management method according to claim 8, characterized in that: Step S43 includes the following steps: Based on the preset XGBoost model, the distributed nodes of the decision tree are designed with distributed gradient boosting decision tree architecture for power drive and power scheduling requirements, and the distributed decision tree model architecture for power scheduling requirements is generated; According to the local power perception drive correlation characteristic data, the fluctuation characteristics of power drive and power dispatch demand are analyzed to generate power drive dispatch demand fluctuation characteristic data; According to the power drive scheduling demand fluctuation characteristic data, the power dispatch demand distributed decision tree model architecture is used to perform decision tree gradient feature learning processing of power drive and power dispatch demand to generate a distributed decision tree model of power dispatch demand.

10. An intelligent power supply unified management method system, characterized in that: Used to execute the intelligent power unified management method as claimed in claim 1, the intelligent power unified management system comprises: The integrated power data mobile phone module is used to collect initial power management data based on the power management system and generate initial power management data; perform heterogeneous power data integration processing based on the initial power management data and generate integrated power data; A multi-level power-aware feature analysis module is used to perform multi-level power-aware cluster analysis based on the integrated power data to generate multi-level power-aware cluster data; perform multi-level power-aware feature analysis based on the multi-level power-aware cluster data to generate multi-level power-aware feature data; A local power-sensing driving correlation feature analysis module is used to perform local distribution division processing based on multi-level power-sensing feature data to generate local multi-level power-sensing feature data; perform local multi-level power-sensing driving correlation feature analysis based on the local multi-level power-sensing feature data to generate local power-sensing driving correlation feature data; The intelligent power unified management module is used to perform local power drive feature analysis based on local multi-level power perception feature data to generate local power drive feature data; design an intelligent monitoring engine for local power scheduling demand prediction based on local power perception drive associated feature data to obtain a local power scheduling demand prediction engine; transmit the local power drive feature data in real time to the local power scheduling demand prediction engine to perform intelligent analysis of local dynamic power scheduling demand to generate local dynamic power scheduling demand data; perform automated decision analysis of global intelligent power management based on local dynamic power scheduling demand data to generate global intelligent power management decisions; perform intelligent power unified management adjustment processing on the initial power management data based on the global intelligent power management decision to generate intelligent power unified management data, and execute intelligent power unified management operations through the intelligent power unified management data.

Images