A method for data mining and processing of power distribution panoramic intelligent management and control decision based on the honey moon

Abstract

The application provides a power distribution panoramic intelligent management and control decision data mining and processing method based on a Hongmeng, relates to the technical field of power system intelligent decision and data mining, and comprises the following steps: performing feature analysis and state mapping on an optimized feature set to obtain an operating state vector; performing multi-objective optimization calculation based on the operating state vector to obtain a preliminary control parameter set; performing instruction coding and strategy fusion on the preliminary control parameter set to obtain a power distribution management and control instruction; and issuing the power distribution management and control instruction to a corresponding terminal and adopting a time sequence synchronization and hash check mechanism to ensure the sequence consistency and integrity of instruction transmission. The application constructs a full-process closed loop from data safe collection, efficient processing, intelligent mining to accurate decision and reliable execution, and realizes panoramic intelligent management and control and adaptive optimization of a power distribution system under data driving.

Description

Technical Field

[0001] This invention relates to the field of intelligent decision-making and data mining technology for power systems, and in particular to a method for data mining and processing of power distribution panoramic intelligent management and control decision-making based on HarmonyOS. Background Technology

[0002] With the continuous expansion of urban power distribution networks and the increasing penetration rate of distributed energy, traditional power distribution network management and control systems face severe challenges in data processing. Existing systems generally adopt a centralized data processing architecture, which has significant technical bottlenecks in areas such as multi-source heterogeneous data fusion, edge computing capability adaptation, and dynamic data calibration.

[0003] At the data acquisition level, the IEC61850 protocol data, MQTT telemetry data and custom JSON format data generated by the power distribution terminal lack a unified access standard, resulting in serious timing misalignment during protocol parsing. Actual operation data shows that the data packet loss rate is as high as 12% when multiple data sources are accessed concurrently, and the data preprocessing stage accounts for more than 65% of the total processing time, which may restrict data processing efficiency.

[0004] At the edge computing level, existing lightweight data mining models are limited by the memory capacity of edge terminals, which is usually less than or equal to 256MB, and cannot effectively support real-time inference of complex algorithms. When processing input data with more than 50 dimensions, the model inference latency increases sharply from 200ms to 1.2s, and the memory overflow probability reaches 23%, forcing the system to upload 70% of the raw data to the cloud for processing, which may cause the anomaly detection response time to exceed the safety threshold of 15 seconds.

[0005] At the feature engineering level, traditional methods are not adaptable enough to the time-varying characteristics of power distribution operation. When the output of new energy sources fluctuates by more than 30% of the rated capacity, the feature extraction model trained based on historical data will exhibit feature drift, and the load classification accuracy will drop by up to 28%. The system lacks a dynamic compensation mechanism based on real-time operation status and cannot adaptively calibrate the feature vector according to the actual operating conditions of the power grid, which may cause a significant deviation between control decisions and actual needs. Summary of the Invention

[0006] The technical problem to be solved by this invention is to provide a data mining and processing method for power distribution panoramic intelligent management and control decision-making based on HarmonyOS. It constructs a closed loop of the entire process from secure data collection, efficient processing, intelligent mining to accurate decision-making and reliable execution, realizing panoramic intelligent management and adaptive optimization of the power distribution system under data-driven conditions.

[0007] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0008] Firstly, a data mining and processing method for power distribution panoramic intelligent management and control decision-making based on HarmonyOS, the method comprising:

[0009] The system accesses and collects multi-source heterogeneous raw data from power distribution terminals via the HarmonyOS distributed soft bus; it performs protocol parsing and unified encapsulation on the multi-source heterogeneous raw data to obtain standardized data packets; and it performs encryption and device authentication on the standardized data packets to obtain a secure encrypted data stream.

[0010] The secure encrypted data stream is decrypted to obtain a decrypted data stream; the decrypted data stream is parsed using multi-source protocols to obtain preliminary structured data; the preliminary structured data is cleaned and spatiotemporally aligned to obtain a spatiotemporally consistent data sequence; the spatiotemporally consistent data sequence is fragmented and reassembled to obtain a standardized data stream.

[0011] Standardized data streams are input into a pre-trained lightweight data mining model, which performs load classification prediction, abnormal state detection, and new energy absorption capacity analysis at the HarmonyOS edge to obtain a basic feature set. A multi-dimensional feature space is constructed based on the basic feature set, and the multi-dimensional feature space is partitioned to obtain a partition configuration strategy. Dynamic compensation coefficients are generated according to the partition configuration strategy, and the basic feature set is calibrated in real time using the dynamic compensation coefficients to obtain an optimized feature set.

[0012] Feature parsing and state mapping are performed on the optimized feature set to obtain the operating state vector; multi-objective optimization calculation is performed based on the operating state vector to obtain the preliminary control parameter set; instruction encoding and strategy fusion are performed on the preliminary control parameter set to obtain the power distribution management and control instructions.

[0013] Power distribution control instructions are sent to the corresponding terminals, and a timing synchronization and hash verification mechanism is used to ensure the consistency and integrity of the instruction transmission order.

[0014] Secondly, the power distribution panoramic intelligent management and control decision-making data mining and processing system based on HarmonyOS includes:

[0015] The data acquisition module is used to access and acquire multi-source heterogeneous raw data from power distribution terminals via HarmonyOS distributed soft bus; it performs protocol parsing and unified encapsulation on the multi-source heterogeneous raw data to obtain standardized data packets; and it performs encryption and device authentication on the standardized data packets to obtain a secure encrypted data stream.

[0016] The standardization module is used to decrypt the secure encrypted data stream to obtain the decrypted data stream; perform multi-source protocol parsing on the decrypted data stream to obtain preliminary structured data; perform data cleaning and spatiotemporal alignment on the preliminary structured data to obtain a spatiotemporally consistent data sequence; and perform fragmentation and reassembly processing on the spatiotemporally consistent data sequence to obtain the standardized data stream.

[0017] The optimization module is used to input standardized data streams into a pre-trained lightweight data mining model, and perform load classification prediction, abnormal state detection, and new energy absorption capacity analysis at the HarmonyOS edge to obtain a basic feature set. Based on the basic feature set, a multi-dimensional feature space is constructed, and the multi-dimensional feature space is partitioned to obtain a partition configuration strategy. Dynamic compensation coefficients are generated according to the partition configuration strategy, and the basic feature set is calibrated in real time using the dynamic compensation coefficients to obtain an optimized feature set.

[0018] The fusion module is used to perform feature parsing and state mapping on the optimized feature set to obtain the operating state vector; perform multi-objective optimization calculation based on the operating state vector to obtain the preliminary control parameter set; and perform instruction encoding and strategy fusion on the preliminary control parameter set to obtain the power distribution management and control instructions.

[0019] The execution module is used to send power distribution control instructions to the corresponding terminals, and uses a timing synchronization and hash verification mechanism to ensure the consistency and integrity of the instruction transmission order.

[0020] Thirdly, a computing device includes:

[0021] One or more processors;

[0022] A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0023] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0024] The above-described solution of the present invention has at least the following beneficial effects:

[0025] HarmonyOS distributed soft bus adapts to various types of power distribution terminals, enabling comprehensive acquisition of multi-source heterogeneous data; unified protocol parsing and encapsulation eliminate format differences, forming regular and standardized data packets; encryption and identity authentication build security protection, ensuring the confidentiality and legitimacy of data transmission; decryption restores the original form of data, and multi-source protocol parsing achieves structured transformation; data cleaning improves data purity, and spatiotemporal alignment solves timing misalignment problems; fragmentation and reassembly optimize data structure, adapting to edge processing needs and forming a coordinated and consistent standardized data flow; lightweight models adapt to edge computing power and memory, enabling local real-time data processing; multi-dimensional feature space and Clear partitioning reveals the distribution patterns of features; dynamic compensation coefficients enable adaptive calibration of features, optimizing the feature set to suit real-time operating conditions; feature parsing and state mapping condense core operational information, forming a vector that comprehensively represents the system's status; multi-objective optimization covers core control requirements, with weights adapted to real-time scenarios; a unified instruction encoding format adapts to terminals, and strategy fusion coordinates multi-device operations, generating globally collaborative control instructions; a timing synchronization mechanism ensures consistency between instruction transmission and execution order; a dual verification mechanism strengthens data integrity protection; instruction issuance and closed-loop feedback ensure execution implementation, achieving full-process traceability and guaranteeing system operational stability. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating the data mining and processing method for intelligent power distribution panoramic management and control based on HarmonyOS, provided in an embodiment of the present invention.

[0027] Figure 2 This is a schematic diagram of a power distribution panoramic intelligent management and control decision data mining and processing system based on HarmonyOS, provided by an embodiment of the present invention. Detailed Implementation

[0028] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0029] like Figure 1 As shown, embodiments of the present invention propose a data mining and processing method for intelligent management and control of power distribution panorama based on HarmonyOS. The method includes the following steps:

[0030] Step 100: Access and collect multi-source heterogeneous raw data from the power distribution terminal via HarmonyOS distributed soft bus; perform protocol parsing and unified encapsulation on the multi-source heterogeneous raw data to obtain standardized data packets; perform encryption and device authentication on the standardized data packets to obtain a secure encrypted data stream;

[0031] Step 200: Decrypt the secure encrypted data stream to obtain a decrypted data stream; perform multi-source protocol parsing on the decrypted data stream to obtain preliminary structured data; perform data cleaning and spatiotemporal alignment on the preliminary structured data to obtain a spatiotemporally consistent data sequence; perform fragmentation and reassembly processing on the spatiotemporally consistent data sequence to obtain a standardized data stream.

[0032] Step 300: Input the standardized data stream into the pre-trained lightweight data mining model, perform load classification prediction, abnormal state detection, and new energy absorption capacity analysis at the HarmonyOS edge to obtain a basic feature set; construct a multi-dimensional feature space based on the basic feature set, and partition the multi-dimensional feature space to obtain a partition configuration strategy; generate dynamic compensation coefficients according to the partition configuration strategy, and use the dynamic compensation coefficients to calibrate the basic feature set in real time to obtain an optimized feature set;

[0033] Step 400: Perform feature parsing and state mapping on the optimized feature set to obtain the operating state vector; perform multi-objective optimization calculation based on the operating state vector to obtain the preliminary control parameter set; perform instruction encoding and strategy fusion on the preliminary control parameter set to obtain the power distribution management and control instructions.

[0034] Step 500: The power distribution control command is sent to the corresponding terminal, and a timing synchronization and hash verification mechanism is used to ensure the consistency and integrity of the command transmission order.

[0035] In this embodiment of the invention, the HarmonyOS distributed soft bus is used to achieve rapid interconnection of multi-source power distribution terminals, ensuring the real-time acquisition of raw data; protocol parsing and unified encapsulation standardize data formats, laying a unified foundation for subsequent processing; encryption processing and identity authentication are combined to enhance the security of data transmission; decryption operations restore the original validity of the data, ensuring the authenticity of the processed objects; multi-source protocol parsing enables data structure transformation, reducing the difficulty of data understanding; data cleaning removes redundant and interfering information, improving data quality; spatiotemporal alignment eliminates data dimensional deviations, forming a regular data sequence; fragmentation and reorganization optimize the data storage and retrieval structure, providing convenience for subsequent mining; a lightweight data mining model is adapted to the computing power of HarmonyOS edge computing, enabling localized deployment of mining tasks and reducing data transmission losses; load classification, anomaly detection, and renewable energy consumption are completed simultaneously. The system analyzes and efficiently extracts core operational features; constructs a rich feature representation dimension using a multi-dimensional feature space; optimizes feature processing logic through partitioned configuration strategies; dynamically compensates for feature data in real time, improving the adaptability and reliability of the feature set and providing precise support for decision-making; feature parsing and state mapping clearly present the operating status of the power distribution system, clarifying the core direction of control; multi-objective optimization calculations integrate diverse control requirements to form coordinated and unified preliminary control parameters; instruction coding and strategy fusion optimize the control instruction structure, enhancing the adaptability of instructions and terminal execution, and ensuring the coherence of control logic; power distribution control instructions are issued to target terminals, ensuring the targeted nature of control actions; a timing synchronization mechanism maintains the sequential stability of instruction transmission, avoiding execution logic errors; hash verification verifies instruction integrity, preventing data tampering during transmission and ensuring the effective implementation of control instructions.

[0036] In a preferred embodiment of the present invention, step 100 above involves accessing and collecting multi-source heterogeneous raw data from a power distribution terminal via the HarmonyOS distributed soft bus; performing protocol parsing and unified encapsulation on the multi-source heterogeneous raw data to obtain standardized data packets; and performing encryption and device authentication on the standardized data packets to obtain a secure encrypted data stream, including:

[0037] Step 101: Collect multi-source heterogeneous raw data through the HarmonyOS distributed soft bus. Specifically, this includes: First, configuring the communication interface of the HarmonyOS distributed soft bus to adapt and connect with the hardware interface of the power distribution terminal. The power distribution terminal includes various devices such as substation monitoring and control devices, photovoltaic inverters, charging pile controllers, and smart meters. A communication network with multiple candidate transmission paths is constructed through the distributed network topology of the soft bus, establishing a point-to-point communication link between the terminal and the management platform. The real-time data transmission service of the soft bus is then started, configuring the data transmission parameters to match the data generation frequency of the power distribution terminal, supporting various power distribution terminals to actively upload multi-source heterogeneous raw data according to their own operating status. During data transmission, the three-dimensional transmission parameters of each candidate transmission path are collected in real time. These three-dimensional transmission parameters include path delay value, signal strength value, and the integrity rate of transmitted data packets. The system sets a three-dimensional parameter benchmark value for the optimal transmission path, which is pre-calibrated based on the transmission performance requirements of the power distribution control system. It calculates the Euclidean distance between the three-dimensional transmission parameters of each candidate transmission path and the benchmark value, prioritizes each candidate path based on the distance value, selects the path with the smallest Euclidean distance as the primary transmission path, and uses the remaining paths as backup transmission paths. It achieves concurrent data reception through a multi-path dynamic switching mechanism of a soft bus, caches data received from the primary transmission path in real time, and continuously monitors the link status of the primary and backup paths. When the Euclidean distance of the primary path exceeds a preset threshold, it automatically switches to the optimal backup path. It performs integrity verification on the data during transmission to ensure comprehensive and complete collection of multi-source heterogeneous raw data, adapting to the access requirements of different types of power distribution terminals and ensuring continuous data acquisition with a low loss rate.

[0038] Step 102 involves identifying the multi-source heterogeneous raw data, distinguishing it from IEC61850 protocol data, MQTT telemetry data, and custom JSON format data. Specifically, this includes: constructing a multi-source data type identification model, which includes a feature extraction module, a feature matching module, and a type determination module; simultaneously building feature libraries for IEC61850 protocol data, MQTT telemetry data, and custom JSON format data. The IEC61850 protocol data feature library stores core features such as specific field identifiers of ASN.1 encoding, fixed frame structure features, and data type identifier bits; the MQTT telemetry data feature library records key information such as preset topic prefix strings, fixed byte sequences in message headers, and QoS level identifier fields; and the custom JSON format data feature library includes unique features such as the starting left curly brace identifier for key-value pairs, comma features separating fields, and double quotes identifiers for string values. The raw data to be processed is read byte by byte through a data reading interface, the first 16 bytes of the data are extracted as the header identifier field, the total number of bytes is counted as the data length feature, and the data is traversed... The separator character between consecutive fields in content recognition is used as the field separator feature. The extracted header identifier field is compared byte by byte with the header features in three feature libraries, and the percentage of matching bytes is counted. The extracted data length feature is matched with the typical range of data length in each feature library to determine whether it is within the length range of the corresponding data type. The extracted field separator feature is fully matched with the separator features in each feature library. The total matching degree is calculated by combining the matching results of the three features. A matching degree threshold is set. When the total matching degree of a feature library is higher than the threshold and significantly higher than the other two feature libraries, a type-specific verification is initiated. For IEC61850 protocol data, the existence of a specific identifier bit in the ASN.1 encoding is verified. For MQTT telemetry data, the topic prefix is ​​verified to be completely consistent with the preset string. For custom JSON format data, it is confirmed that the starting character is a left curly brace and contains at least one complete key-value pair structure. After passing the special verification, the data type is determined, and the classification of three different formats of raw data is completed, providing a clear and reliable basis for the subsequent differentiated parsing process.

[0039] Step 103 involves performing multi-protocol parsing on the IEC61850 protocol data, MQTT telemetry data, and custom JSON format data to identify and extract valid fields. Specifically, this includes: for the identified IEC61850 protocol data, using the abstract communication service interface protocol parsing method, parsing its manufacturing message specification frame structure, and extracting measurement value fields such as voltage, current, and power, as well as equipment status identifier fields; for the MQTT telemetry data, using a topic filtering mechanism to filter valid data topics, parsing the telemetry parameter fields in the message payload, and removing topic-irrelevant information; for the custom JSON format data, according to preset key-value pair mapping rules, traversing the data structure to extract key operational data fields, removing redundant comment information and invalid filler fields, ensuring that valid fields in various protocol data are extracted completely and accurately.

[0040] Step 104 involves semantic reconstruction and format standardization of the valid fields, encapsulating them into a standardized data packet with a unified time sequence identifier and a data source identifier. Specifically, this includes: establishing a unified semantic mapping rule base; standardizing the names of valid fields extracted from different protocols according to semantic associations; unifying the data types and numerical units of the fields to eliminate semantic ambiguity between different protocols; assigning a time sequence identifier of uniform precision to each data entry based on the global time synchronization mechanism of the HarmonyOS distributed soft bus to ensure consistency in the time dimension of the data; adding a unique data source identifier to each data entry, which corresponds one-to-one with the device number of the power distribution terminal to clarify the data's generating entity; and integrating and encapsulating the standardized fields, time sequence identifier, and data source identifier according to a preset data packet structure specification to form a standardized data packet with a unified structure and complete information.

[0041] Step 105 involves redundancy compression and length alignment of the standardized data packets to obtain standardized data packets. Specifically, this includes: using a lossless compression algorithm to compress redundant fields, duplicate data, and invalid padding in the standardized data packets, removing meaningless redundant information and reducing the data packet size; based on a preset standardized data length threshold, performing length alignment on the compressed data packets by supplementing with fixed-length invalid placeholders or splitting excessively long data packets to ensure all data packets have a uniform length of a preset number of bytes, guaranteeing a regular data packet structure; after redundancy compression and length alignment, a standardized data packet with a unified format, reduced size, and regular structure is formed, providing an efficient and adaptable data foundation for subsequent encryption, transmission, and data cleaning processes, and reducing resource consumption during data storage and transmission.

[0042] In a preferred embodiment of the present invention, step 200 above involves decrypting the secure encrypted data stream to obtain a decrypted data stream; performing multi-source protocol parsing on the decrypted data stream to obtain preliminary structured data; performing data cleaning and spatiotemporal alignment on the preliminary structured data to obtain a spatiotemporally consistent data sequence; and performing fragmentation and reassembly processing on the spatiotemporally consistent data sequence to obtain a standardized data stream, including:

[0043] Step 201 involves decrypting and verifying the integrity of the secure encrypted data stream to obtain a decrypted data stream. Specifically, this includes: using a pre-defined private key corresponding to an asymmetric encryption algorithm to decrypt the secure encrypted data stream. The asymmetric encryption algorithm is an encryption standard adapted to the security requirements of power distribution data transmission. During decryption, an integrity verification field attached to the encrypted data stream is extracted simultaneously. The extracted integrity verification field is compared with the real-time calculated verification value of the decrypted data stream. The real-time calculated verification value is obtained by calculating the decrypted data stream using a hash algorithm consistent with the encryption stage. If the comparison results match, the data is determined to be complete and valid, and the decrypted data stream is output. If the comparison results do not match, the data is determined to have been tampered with or lost. A retransmission request is immediately sent to the data sending terminal, and the source identifier and time sequence information of the abnormal data are recorded.

[0044] Step 202 involves performing multi-source protocol parsing on the decrypted data stream to identify the data source and format, obtaining preliminary structured data. Specifically, this includes: calling the multi-source data feature library constructed in step 102 to perform byte-by-byte feature matching on the decrypted data stream; first, identifying the data source identifier field contained in the data stream, which corresponds one-to-one with the unique device number of the power distribution terminal to determine the data's origin; then, determining the data's format type based on the feature matching results, distinguishing between IEC61850 protocol data, MQTT telemetry data, and custom JSON format data; employing dedicated parsing logic for different data formats; parsing the manufacturing message specification structure of IEC61850 protocol data through an abstract communication service interface to extract measurement fields such as voltage, current, and power, as well as device status fields; parsing MQTT telemetry data through topic filtering and message load parsing to extract telemetry parameters and acquisition time fields; extracting key operational data fields from custom JSON format data according to preset key-value pair mapping rules; and organizing the parsed fields according to a unified data structure specification, clarifying field names, data types, and numerical units to form preliminary structured data, achieving the structured transformation of different protocol data and providing a unified format foundation for subsequent data processing.

[0045] Step 203 involves outlier detection and null value removal on the preliminary structured data to obtain a valid data set; and timestamp calibration and device space topology matching on the valid data set to obtain a spatiotemporally consistent data sequence. Specifically, this includes: using a combination of statistical thresholding and data trend analysis to detect outliers in the preliminary structured data; setting normal value ranges and rate-of-change thresholds for each data field; marking data exceeding the range or rate of change exceeding the threshold as outliers and removing them; identifying null value fields in the preliminary structured data; directly removing data records containing null values ​​if the proportion of null values ​​is lower than a preset ratio; and sending a retransmission request to the corresponding terminal if the proportion of null values ​​is higher than a preset ratio to ensure the purity of the valid data set; obtaining the original timestamps of each data record based on the global time synchronization mechanism of the HarmonyOS distributed soft bus, calibrating them to a standard timestamp of uniform precision to eliminate timing misalignments caused by clock deviations between different terminals; and retrieving the device space topology of the power distribution system. The diagram shows how to obtain the installation location coordinates, electrical connection relationships, and equipment coverage parameters of each data source terminal. The installation location and effective data acquisition range of each terminal are abstracted into polygonal regions, and the vertex coordinate information of each polygonal region is extracted. A polygonal union area optimization algorithm is used to calculate the union area of ​​the polygonal region corresponding to the target terminal and the polygonal regions of adjacent associated terminals. The adjacent associated terminals are terminal devices with data interaction needs determined based on the electrical connection logic of the power distribution system. A union area matching threshold is set, which is pre-calibrated based on the power supply radius and data acquisition accuracy requirements of the power distribution system. Data from adjacent terminals with a union area greater than or equal to this threshold are selected as associated data. Each piece of data in the effective data set is bound to the spatial location information of the corresponding terminal and the selected associated data to determine the spatial attributes and spatial relationships of the data. Through timestamp calibration and optimized spatial topology matching, a spatiotemporally consistent data sequence with time consistency, spatial correlation, and matching is formed.

[0046] Step 204 involves performing transmission characteristic analysis on the spatiotemporally consistent data sequence to determine fragmentation configuration parameters; dynamically fragmenting the data sequence based on the fragmentation configuration parameters to obtain fragmented data units; adding redundancy check flags to the fragmented data units and implementing sequential encoding to obtain data units to be reassembled. Specifically, this includes: performing transmission characteristic analysis on the spatiotemporally consistent data sequence to extract parameters such as the total data volume, data transmission rate, data update frequency of each field, and memory capacity parameters of the edge terminal (the edge terminal memory capacity parameters are obtained in real-time based on the HarmonyOS device management interface); and determining fragmentation configuration parameters based on the above transmission characteristics, including fragmented data volume, number of fragments, and fragmentation method. Prioritization is implemented, ensuring that the amount of fragmented data does not exceed 70% of the available memory capacity of the edge terminal, thus guaranteeing that the fragmented data is adaptable to the computing power of the edge terminal. The data sequence, which is consistent in time and space, is dynamically fragmented according to the fragmentation configuration parameters, dividing the data sequence into multiple independent fragmented data units based on time order or data type. Each fragmented data unit contains a complete field structure and a time-space identifier. A dedicated redundancy check field is calculated for each fragmented data unit, based on the feature values ​​of all data within the fragment. All fragmented data units are sequentially numbered according to the fragmentation order, with the number including the fragment sequence number and the total number of fragments, forming a data unit to be reassembled that combines redundancy check functionality with sequence identification.

[0047] Step 205 involves reassembling and verifying the integrity of the data units to be reassembled to obtain a standardized data stream. Specifically, this includes: retrieving each data segment in sequence from the start number to the end number according to the sequential encoding of the data units to be reassembled; verifying the redundancy check field of each retrieved data segment by calculating the feature value of the current data segment and comparing it with the attached redundancy check field; if the comparison matches, the data segment is deemed complete and valid; if the comparison does not match, it is marked as an abnormal segment and a retransmission mechanism is initiated; reassembling the verified data segments according to the logical relationship corresponding to the sequential encoding to restore the original structure of the data sequence; verifying the integrity of the overall data after reassembly, and statistically analyzing the total data volume, field integrity, and temporal / spatial identifier consistency; if all meet the preset standards, the reassembly is deemed successful; and encapsulating the reassembled and verified data according to a unified format specification to ensure that the data field structure, temporal identifier, and spatial attributes are consistent, forming a standardized data stream.

[0048] In this embodiment of the invention, the decryption operation restores the original form of the data, ensuring the effectiveness of data processing; integrity verification checks for data tampering or missing data during transmission, ensuring the authenticity and integrity of the input data; the source and format of the decrypted data stream are identified, and multi-source protocol parsing is performed in a targeted manner to efficiently achieve data structure transformation, improving the orderliness and adaptability of data processing; outlier detection and null value removal remove invalid and interfering data, improving the purity of the data set; timestamp calibration unifies the time dimension benchmark of the data, and device spatial topology matching clarifies the spatial correlation of the data, jointly forming a spatiotemporally regular data sequence, providing a basis for subsequent multi-dimensional analysis. It provides consistent data support; transmission feature analysis accurately adapts to data transmission characteristics and determines fragmentation configuration parameters; dynamic fragmentation optimizes data storage and transmission structure to adapt to processing needs in different scenarios; redundancy check identifiers enhance the security of fragmented data, and sequential encoding clarifies the logical relationship between fragmented units, ensuring the integrity and orderliness of fragmented data; reassembly operations integrate fragmented data units to restore the complete logical structure of the data; integrity verification checks for data omissions or errors during the reassembly process to ensure the data is in a regular form; and finally, it forms a standardized data stream, reducing the processing complexity of subsequent data mining and decision-making processes and improving the efficiency of the overall data processing workflow.

[0049] In a preferred embodiment of the present invention, step 300 involves inputting a standardized data stream into a pre-trained lightweight data mining model, performing load classification prediction, abnormal state detection, and new energy absorption capacity analysis at the HarmonyOS edge to obtain a basic feature set; constructing a multi-dimensional feature space based on the basic feature set, and partitioning the multi-dimensional feature space to obtain a partitioning configuration strategy; generating dynamic compensation coefficients according to the partitioning configuration strategy, and using the dynamic compensation coefficients to calibrate the basic feature set in real time to obtain an optimized feature set, including:

[0050] Step 301 involves inputting the standardized data stream into a lightweight data mining model that fuses a pre-trained pruned decision tree and a deep forest to obtain a primary feature map. Specifically, this includes: selecting a fusion architecture of pruned decision trees and a deep forest adapted to the memory capacity limitations of the HarmonyOS edge; constructing a lightweight data mining model; determining the hierarchical structure and node splitting criteria of the pruned decision tree during construction; setting the initial number of layers in the decision tree and the feature splitting threshold for internal nodes; limiting the number of sub-forests in the deep forest and the number of decision trees contained in each sub-forest; controlling the depth of each decision tree to within 8 layers; and setting the number of sub-forests to 6 to 10 to ensure that the basic complexity of the fusion architecture is adapted to the computing power of the edge; employing a serial fusion logic, using the output of the pruned decision tree as one of the input features of the deep forest to form a two-level processing architecture for feature selection and feature enhancement; controlling the total number of parameters of the overall model within a preset threshold to lay the foundation for subsequent memory optimization.

[0051] The model training phase utilizes historical power distribution data covering different load scenarios, abnormal states, and fluctuations in renewable energy output for pre-training. Before training, the historical data is cleaned and labeled, and divided into training and validation sets, with the training set accounting for at least 70%. Labeling includes load type, abnormality level, and renewable energy absorption efficiency labels. The pruned decision tree is trained separately first, and the node splitting threshold is iteratively optimized using a gradient boosting strategy. Then, the trained pruned decision tree is embedded into a deep forest architecture for joint training to adjust the decision boundaries of the sub-forests. During training, the model's performance metrics and memory usage on the validation set are monitored in real time. A pruning algorithm removes redundant branch nodes from the decision tree that do not contribute to model performance improvement. The pruning process combines pre-pruning and post-pruning. In the pre-pruning stage, the decision tree is stopped from overgrowing based on the validation set accuracy. In the post-pruning stage, redundant branches that do not contribute to model performance are removed. For branches where the accuracy improvement is less than a preset threshold, the number of sub-forests in the deep forest is dynamically adjusted. If the memory usage approaches 256MB, the number of sub-forests is gradually reduced until the model's memory usage is strictly controlled within 256MB. Standardized data streams are input into the pre-trained model in fixed data block sizes (each block contains 100 to 200 consecutive data records). Through the fast feature selection mechanism of the pruning decision tree, key data dimensions strongly related to load status, anomaly detection, and renewable energy consumption are selected based on the node splitting threshold, while redundant feature dimensions are removed. Then, the deep forest processes the features in parallel through multiple sub-models. Each sub-forest independently learns to represent the selected features. The output results of each sub-forest are integrated through a voting mechanism to enhance the feature representation capability. Finally, a primary feature map containing information related to load characteristics, anomaly correlation, and renewable energy consumption is output.

[0052] Step 302: Based on the primary feature mapping, load pattern classification is performed using the hierarchical discriminant structure of the pruned decision tree to obtain a load state feature vector. Specifically, this includes: based on the load-related feature dimensions in the primary feature mapping, calling the hierarchical discriminant structure of the pruned decision tree. This structure sets discriminant nodes from coarse to fine according to feature dimensions such as electricity consumption time period, power amplitude, and fluctuation frequency. Each discriminant node has a preset feature threshold adapted to the power distribution scenario. The corresponding features in the primary feature mapping are sequentially input into each discriminant node, and the load type is classified through hierarchical logical judgment, including commercial load, industrial load, residential load, and new energy supporting load, etc. Key operating parameters are extracted for each load type, including rated power, peak and valley time distribution, fluctuation coefficient, etc. These parameters are arranged in a unified dimensional order to form a structured load state feature vector.

[0053] Step 303: Input the load state feature vector and the primary feature map into the multi-granularity scanning mechanism of the deep forest for abnormal pattern recognition to obtain an anomaly index vector. Specifically, this includes: preprocessing the load state feature vector and the primary feature map, unifying their temporal identification precision to the millisecond level, calibrating the numerical units of each feature dimension, and concatenating the dimensions in the order of load state feature vector first, followed by the primary feature map, keeping the total dimensions after concatenation below 80 to adapt to the memory processing capabilities of HarmonyOS edge computing; then... The vector input deep forest employs a multi-granularity scanning mechanism. This mechanism pre-sets three feature sliding windows of small, medium, and large sizes. The small window size is 4 to 6 dimensions, which is suitable for extracting short-period anomaly features such as power mutations. The medium window size is 8 to 10 dimensions, which is suitable for medium-period anomaly features such as continuous parameter exceedances. The large window size is 12 to 14 dimensions, which is suitable for long-period anomaly features such as data time-series breaks. The window scanning step size is set to 1 to 2 dimensions to ensure that all feature dimensions are traversed without omission. Each time the window slides, the corresponding local feature sub-vectors are extracted to form a multi-scale feature set.

[0054] Each sub-forest preloads a specialized anomaly detection model trained on historical anomaly data. Different sub-forests focus on identifying different types of anomaly patterns. Some sub-forests focus on power mutation identification, with a preset power mutation rate threshold of 30% to 50% of the corresponding equipment's rated power and a mutation duration threshold of 50 to 200 ms. When the power change rate in a local feature vector exceeds the threshold and the duration exceeds the threshold, it is determined to be a power mutation anomaly. Some sub-forests focus on continuous parameter over-limit identification, based on power grid operation procedures and equipment technical parameters, with preset safe operating threshold ranges for electrical parameters such as voltage, current, and frequency. When the value of a parameter in a local feature vector exceeds the threshold and the duration exceeds 300 ms, it is determined to be a continuous parameter over-limit anomaly. Some sub-forests focus on data time sequence break identification, with a preset maximum allowable interval threshold of 100 ms between adjacent data timestamps. When the timestamp interval corresponding to a local feature vector exceeds the threshold and there is no reasonable missing identifier, it is determined to be a data time sequence break anomaly. Each sub-forest performs one-by-one matching analysis on local feature vectors in the multi-scale feature set and outputs identification results including whether it is an anomaly, the anomaly type, and the anomaly confidence level.

[0055] Collect anomaly identification results from all sub-forests, categorize and count the number of sub-forests supporting the anomaly determination by anomaly type, and set a valid voting ratio threshold of 60% of the total number of sub-forests. If the support ratio for a certain anomaly type exceeds the threshold, it is determined to be an anomaly; if the ratio is 30% to 60%, it is marked as a suspected anomaly; and if it is below 30%, it is determined to be normal. For confirmed anomalies, integrate the identification details of all supporting sub-forests, eliminate contradictory information, and clarify the core features of the anomaly. Extract the anomaly occurrence time to the millisecond level and synchronize it with the global time. The anomaly parameter type is clearly defined as specific electrical parameters such as voltage, current, and power. The anomaly severity is quantified according to the scope of impact and severity level: minor anomalies are assigned a value of 0.2 to 0.4, moderate anomalies are assigned a value of 0.5 to 0.7, and severe anomalies are assigned a value of 0.8 to 1.0. At the same time, information such as the duration of the anomaly and the identification of associated equipment is extracted. Arrange this information in a preset 12-dimensional order to form a unified and complete anomaly indicator vector to comprehensively capture the abnormal state information of the power distribution system.

[0056] Step 304: Based on the load state feature vector and anomaly index vector, perform new energy output characteristic analysis through the regression analysis unit of the lightweight data mining model to obtain the absorption potential feature vector. Specifically, this includes: calling the lightweight regression analysis unit in the lightweight data mining model. This unit simplifies the regression calculation logic, reduces matrix operation complexity, and controls the data accuracy within a reasonable range during the calculation process, adapting to the computing power limitations of the HarmonyOS edge computing; determining the input features of the regression analysis, including parameters such as adjustable load capacity, load peak-to-valley difference, and load response rate in the load state feature vector; parameters such as the anomaly impact range, anomaly duration, and fault-free operating period in the anomaly index vector; and parameters such as historical new energy output data, output fluctuation coefficient, and predicted output in the primary feature mapping; and then... The input features are arranged in a preset order to form a regression analysis input vector, which is then input into a lightweight regression analysis unit. Through regression analysis, a correlation model between renewable energy output and load acceptance capacity is established. The adaptability of the load under different renewable energy output levels is analyzed, and key parameters such as the maximum absorption power, optimal access time, and absorption duration of renewable energy under the current load condition are calculated. Simultaneously, based on the abnormal information in the abnormal indicator vector, the absorption capacity parameters are corrected. If a severe abnormality exists, the maximum absorption power is reduced proportionally to the scope of the abnormality's impact; if a minor abnormality exists, the basic absorption parameters are maintained. Parameters such as the maximum absorption power, optimal access time, absorption duration, and absorption constraints are quantified and encoded to form a fixed 6-dimensional absorption potential feature vector. The system characterizes the absorption capacity and boundary conditions of renewable energy under the current power distribution conditions.

[0057] Step 305: Perform feature vectorization on the load state feature vector, anomaly index vector, and absorption potential feature vector to obtain vectorized features; then perform weighted fusion on the vectorized features to obtain a basic feature set. Specifically, this includes: using a unified vector coding standard to perform feature vectorization on the load state feature vector, anomaly index vector, and absorption potential feature vector, mapping the values ​​of each vector to a standardized range of 0 to 1 to eliminate the influence of differences in numerical ranges across different dimensions; and performing dimension padding on the 8-dimensional data of the load state feature vector, the 12-dimensional data of the anomaly index vector, and the 6-dimensional data of the absorption potential feature vector to ensure that all vectors are expanded to 20 dimensions, filling in missing dimensions. The weighting coefficients of each feature vector are determined based on the analytic hierarchy process (AHP) and mutual information method. The weighting coefficients of the load state feature vector are set to 0.4, the anomaly index vector to 0.3, and the absorption potential feature vector to 0.3. This weighting allocation is based on the degree of influence of each feature on power distribution control decisions. The three vectorized feature vectors are then summed according to the weighting coefficients. The value of each dimension is equal to the sum of the values ​​of the corresponding dimension in the three vectors multiplied by their respective weighting coefficients. The calculation results are then structured according to feature categories to determine the physical meaning, data type, and numerical range of each dimension, forming a basic feature set with unified dimensions and comprehensive information.

[0058] Step 306: Based on the aforementioned basic feature set, define the coordinate axes for each feature dimension and establish a multi-dimensional feature space. Specifically, this includes: defining an independent coordinate axis for each of the 20 feature dimensions contained in the basic feature set. The physical meaning of the coordinate axis is consistent with the physical attributes of the corresponding feature dimension, including power coordinate axes, current coordinate axes, voltage coordinate axes, fluctuation coefficient coordinate axes, anomaly level coordinate axes, and absorption capacity coordinate axes, etc. The value range of each coordinate axis is set by expanding the historical maximum and minimum values ​​of that feature dimension by 10% to ensure that all feature data points fall within the coordinate axis range. The accuracy of the coordinate axes is set according to the measurement accuracy of the feature data. The accuracy of continuous parameters such as voltage, current, and power is set to two decimal places, and the accuracy of discrete parameters such as anomaly level and absorption efficiency is set to integer levels. Combine the coordinate axes of all 20 feature dimensions in a preset order to establish a multi-dimensional feature space. Each coordinate point in this space corresponds to a feature vector in the basic feature set. The number of spatial dimensions is consistent with the number of feature dimensions in the basic feature set. Each dimension corresponds to a unique feature type identifier, providing a unified spatial framework for the distribution analysis and clustering of feature data.

[0059] Step 307 involves mapping the basic feature set to the multi-dimensional feature space to form a feature data point distribution. Specifically, this includes: extracting the 20-dimensional values ​​of each feature vector in the basic feature set; mapping the values ​​of each dimension to the corresponding positions on the coordinate axes in the multi-dimensional feature space to form individual feature data points; each feature data point carries the time sequence identifier, device source identifier, and feature category identifier of the original data. The time sequence identifier is consistent with the global time of the HarmonyOS distributed soft bus, and the device source identifier corresponds one-to-one with the unique number of the power distribution terminal; mapping all feature data points corresponding to all feature vectors to the multi-dimensional feature space in chronological order, recording the coordinate information, association identifier information, and mapping time of each data point in real time; after mapping, an overall distribution of feature data points is formed. This distribution intuitively reflects the degree of correlation, clustering trend, and dispersion between different features, and simultaneously generates a data point distribution heatmap with related visual description information, providing a quantifiable and analyzable data distribution foundation for subsequent clustering analysis.

[0060] Step 308 involves using a density-based spatial clustering method to perform cluster analysis on the distribution of the feature data points, identifying high-density regions of the data distribution, and generating a set of partition identifiers. Specifically, this includes: analyzing the distribution of feature data points using a density-based spatial clustering method, pre-setting a clustering density threshold and a neighborhood radius. The density threshold is set to a minimum of 15 to 20 data points per neighborhood, and the neighborhood radius is determined based on the average distance statistics of the feature data points to ensure that the clustering results accurately reflect the feature distribution pattern; traversing all feature data points in the multi-dimensional feature space, and calculating the neighborhood radius for each data point. The number of other data points is counted; if the number exceeds the preset density threshold, the data point is marked as a core point, and all data points in its neighborhood are grouped into the same cluster, while non-core points in the neighborhood are marked as boundary points; if the number does not reach the density threshold, it is marked as an isolated point; this process is repeated until all core points are clustered, and isolated points are recorded separately as abnormal data points; all high-density regions in the feature data distribution are identified, and each high-density region corresponds to a cluster; a unique partition identifier is assigned to each high-density region, which contains cluster number, core feature type and data point quantity information, and a set of partition identifiers is generated.

[0061] Step 309: Calculate the centroid position and boundary range of each partition in the partition identifier set to determine the spatial topological relationship between partitions. Specifically, this includes: for each partition in the partition identifier set, traversing the coordinates of all feature data points within that partition in each dimension, calculating the arithmetic mean of the coordinates in each dimension, and combining the means of all dimensions to form the centroid position of that partition. The coordinate precision of the centroid position is consistent with that of the feature data points. Simultaneously, determine the boundary range of each partition in each dimension. The boundary range of each dimension is defined by the maximum and minimum values ​​of all data points in that dimension, forming the partition's boundary range. Multidimensional bounding boxes are used to calculate the Euclidean distance between the centroids of different partitions. The spatial topological relationship between partitions is determined by combining the multidimensional bounding boxes of each partition. If the bounding boxes of two partitions overlap, they are considered to be in an overlapping relationship. If the bounding boxes do not overlap but the centroid distance is less than twice the neighborhood radius, they are considered to be in an adjacent relationship. If the centroid distance is greater than twice the neighborhood radius and the bounding boxes do not overlap, they are considered to be in a separate relationship. The centroid position, boundary range, and topological relationship with other partitions of each partition are recorded in a structured manner to form a partition spatial attribute table, providing spatial parameter support for the formulation of partition configuration strategies.

[0062] Step 310: Integrate the centroid location, boundary range, and spatial topology of the partitions to obtain a partition configuration strategy. Specifically, this includes: integrating the centroid location, boundary range, and spatial topology of each partition; supplementing auxiliary information such as feature type, number of data points, and average core feature for each partition; ranking the partitions according to the number of data points within the partition, the correlation between core features and power distribution control decisions, and determining partitions with more than 50 data points and core features such as power and voltage as key parameters as core partitions, with the rest being secondary partitions; formulating partition management rules, with core partitions using a priority processing mechanism and allocating higher computing resources in subsequent feature calibration, adjacent partitions using a collaborative analysis strategy, considering the influence of adjacent partitions simultaneously when a partition is calibrated, and separated partitions using an independent processing mechanism; and structurally integrating the above partition basic parameters, topology relationships, and management rules to form a partition configuration strategy that includes partition identifiers, spatial attributes, importance levels, and management rules.

[0063] Step 311: According to the partitioning configuration strategy, extract the feature data subsets corresponding to each partition; perform statistical analysis on the feature data subsets of each partition, and calculate the mean and variance of the feature values ​​within the partition respectively. Specifically, this includes: according to the partition identifier and boundary range in the partitioning configuration strategy, selecting all feature vectors whose coordinates fall within the boundary range of the partition from the basic feature set to form the feature data subset corresponding to the partition; performing dimensional statistical analysis on the feature data subsets of each partition, traversing the values ​​of the same dimension of all feature vectors within the subset, calculating the arithmetic mean of the values ​​of that dimension to reflect the central tendency of the feature in that dimension; simultaneously calculating the variance of the values ​​of that dimension to reflect the dispersion of the feature in that dimension; performing the above statistical operations on all 20 dimensions of the feature data subsets, recording the mean and variance values ​​of each dimension, and generating a partition feature statistical report. The report includes information such as partition identifier, mean of each dimension, variance of each dimension, and statistical time, providing quantitative data support for partition stability assessment.

[0064] Step 312: Based on the mean and variance of the feature values ​​within a partition, calculate the ratio of variance to mean as a partition stability index. Specifically, this includes: for each dimension of each partition, calculate the ratio of variance to mean. If the mean is zero, set the stability index for that dimension to a preset maximum value of 1.0, indicating that the feature of that dimension fluctuates extremely drastically; Calculate the arithmetic mean ratio of the variances of all dimensions within the same partition to obtain the overall stability index for that partition, with the index value ranging from 0 to 1.0. A larger index value indicates more drastic fluctuations in the feature values ​​within the partition and poorer stability, while a smaller value indicates more concentrated feature values ​​within the partition and better stability. This index quantifies the feature fluctuation state of each partition and correlates the stability index with feature drift risk. A value greater than 0.6 indicates a high drift risk partition, a value between 0.3 and 0.6 indicates a medium drift risk partition, and a value less than 0.3 indicates a low drift risk partition, providing a quantitative basis for the generation of dynamic compensation coefficients.

[0065] Step 313: Based on the partition stability index, obtain the dynamic compensation coefficient through a preset mapping relationship. Specifically, this includes: constructing a basic dataset of mapping relationships based on a large amount of power distribution system operation experimental data. The experimental data collection period is no less than 12 months, covering different load fluctuation scenarios such as commercial, industrial, and residential, extreme fluctuation scenarios of new energy output exceeding ±30%, abnormal operating conditions such as equipment short-circuit parameter exceeding limits and data time sequence breaks, and stable operating conditions of normal power grid operation. The sample size for each scenario is no less than 1000 sets, ensuring that the data covers the entire operating state of the power distribution system; designing multiple sets of comparative experimental schemes, firstly, the basic dataset is arranged according to the stability... Qualitative index values ​​are divided into intervals. Through statistical analysis of historical characteristic drift cases, stability indices of 0.3 and 0.6 are determined as critical thresholds, dividing the data into three drift risk intervals: low (0 to 0.3), medium (0.3 to 0.6), and high (0.6 to 1.0). For each interval's subset of characteristic data, different degrees of characteristic drift are artificially simulated, and the calibration effect of different compensation coefficients within the range of 0.1 to 1.0 is tested. The calibration effect evaluation indicators include the fit between the characteristic vector and the actual operating conditions, the consistency of subsequent load classification, and the adaptability to new energy consumption analysis.

[0066] By statistically analyzing the evaluation index results of each group of experiments, the compensation coefficient range that optimizes the calibration effect in each interval is selected. In the low drift risk interval, the feature fit is highest and there is no overcalibration when the compensation coefficient is 0.1 to 0.3. In the medium drift risk interval, the compensation coefficient is 0.4 to 0.7, which can effectively balance the calibration effect and data authenticity. In the high drift risk interval, the compensation coefficient is 0.8 to 1.0, which can correct feature drift deviation to the greatest extent. The selected coefficient range is associated with the corresponding stability index interval to form an initial mapping relationship. The initial mapping relationship is then verified using an independent validation dataset. The validation dataset contains 2000 sets of data from different working conditions that were not used in the training. If the calibration effect evaluation index of a certain interval is lower than the preset threshold, the compensation coefficient range of that interval is adjusted, and the verification is repeated until the calibration effect of all intervals meets the requirements. Finally, the mapping relationship that has been verified and optimized through multiple rounds of experiments is solidified and stored to form a preset mapping relationship. Based on the partition stability index calculated in step 312, the preset mapping relationship is queried to obtain the dynamic compensation coefficient corresponding to each partition.

[0067] Step 314 involves receiving the dynamic compensation coefficients and organizing them into partition compensation vectors according to the partition configuration strategy. Specifically, this includes: receiving the dynamic compensation coefficients corresponding to each partition; expanding the dynamic compensation coefficients according to the partition identifier and feature dimension information in the partition configuration strategy to form a 20-dimensional partition compensation vector consistent with the feature vector dimension; each dimension value of the compensation vector is the dynamic compensation coefficient corresponding to that partition, ensuring that the compensation coefficients can be evenly applied to each dimension of the feature vector; sorting and indexing the partition compensation vectors according to the partition identifier, establishing the association between the compensation vectors and the partition identifier and partition stability index, forming a structured set of partition compensation vectors; simultaneously validating the compensation vectors to ensure that the compensation coefficient values ​​for each dimension are within a preset range, removing abnormal compensation vectors that exceed the range and regenerating them, standardizing the storage and retrieval methods of the compensation vectors, and improving the execution efficiency of subsequent feature calibration.

[0068] Step 315: Based on the partition configuration strategy, establish the correspondence between each feature vector in the basic feature set and the partition compensation vector. Specifically, this includes: based on the partition boundary range and feature data mapping rules in the partition configuration strategy, traversing each feature vector in the basic feature set, calculating the distance between the centroid coordinates of the feature vector and the centroid position of each partition; assigning the feature vector to the nearest partition, determining the partition identifier of each feature vector; querying the partition compensation vector set through the partition identifier to obtain the partition compensation vector corresponding to the feature vector; establishing a correspondence table between feature vectors and partition compensation vectors, which contains information such as the unique identifier of the feature vector, the partition identifier, the compensation vector index, the corresponding time, and stability indicators; this correspondence table supports fast querying and dynamic updating. When the partition configuration strategy is adjusted or a new feature vector is added, the correspondence is updated in real time to ensure that each feature vector can be associated with the corresponding partition compensation vector, avoiding matching deviations during the calibration process.

[0069] Step 316: Based on the correspondence between each feature vector and the partition compensation vector, perform a dot product operation on each feature vector and its corresponding partition compensation vector to complete the weighted calibration of the feature values, resulting in a weighted calibrated feature vector. Specifically, this includes: obtaining the partition compensation vector corresponding to each feature vector; performing a dot product operation on each dimension value of the feature vector and the compensation coefficient of the corresponding dimension of the partition compensation vector; monitoring the numerical range in real time during the operation to ensure the result falls within the standardized interval of 0 to 1; truncating the result at the interval boundary value if it exceeds the range; adjusting the weighted values ​​of each dimension of the feature vector through the dot product operation, strengthening the calibration for feature dimensions with drastic fluctuations and high drift risk, and weakening the calibration effect for stable feature dimensions with low drift risk, correcting the deviation data of the feature vector, and strengthening the signal strength of effective features; after completing the weighted calibration of the feature values, outputting the weighted calibrated feature vector. This vector, while maintaining the original feature information, improves the adaptability to real-time operating conditions and reduces the impact of feature drift.

[0070] Step 317: Integrate all weighted and calibrated feature vectors to obtain an optimized feature set. This includes: collecting all weighted and calibrated feature vectors, integrating and sorting them according to the chronological order of the original basic feature set and the device source identifier to ensure that the integrated feature set maintains data continuity and correlation; performing consistency verification on the integrated feature vectors, checking whether the number of dimensions and numerical range of each feature vector meet the preset standards, removing abnormal vectors that do not meet the standards and recording the reasons for the abnormalities, and simultaneously initiating the recalibration mechanism for abnormal vectors; classifying and organizing the verified feature vectors according to feature type, determining the partition identifier, calibration time, corresponding compensation coefficient, and stability index information for each feature vector; and finally forming an optimized feature set that is complete in information, accurate in representation, and highly adaptable.

[0071] In this embodiment of the invention, the fusion architecture of pruned decision trees and deep forests adapts to the computing power limitations of HarmonyOS edge computing, enabling lightweight model deployment and reducing memory consumption and runtime. It rapidly extracts and initially maps features from standardized data streams, efficiently transforming them into feature forms that the model can process, providing basic feature support for subsequent classification, detection, and analysis. The hierarchical discriminative structure of the pruned decision tree simplifies classification logic and improves the efficiency of load pattern classification. It divides different types of load operation modes, forming structured load state feature vectors that clearly characterize the load operation status of the power distribution system. The multi-granularity scanning mechanism of the deep forest expands the feature perception range and comprehensively captures features related to abnormal patterns. By combining load status feature vectors with primary feature mapping, abnormal states of the power distribution system are identified from multiple dimensions, generating abnormal indicator vectors and presenting complete information related to abnormalities. The regression analysis unit focuses on the output characteristics of new energy sources, integrates load status and abnormal indicator-related features, and systematically analyzes the potential for new energy absorption. The generated absorption potential feature vector quantifies parameters related to new energy absorption capacity, providing key basis for subsequent management and control decisions. Feature vectorization standardizes the form of different types of features and unifies feature expression standards. The weighted fusion strategy integrates three core features: load status, abnormal indicators, and absorption potential, highlighting the influence weight of key features to form a basic feature set with a unified structure and comprehensive information.

[0072] This system defines feature dimension coordinate axes to clarify the quantification standards and value ranges of each feature, establishing a structured multi-dimensional feature space. It provides a unified spatial framework for feature data distribution analysis and clustering, standardizing the logical basis for subsequent feature processing. The system maps the basic feature set to the multi-dimensional feature space, intuitively presenting the distribution patterns and clustering trends of feature data. It clearly demonstrates the relationships between features, providing a visual and analyzable data distribution foundation for subsequent clustering analysis. Density-based spatial clustering methods accurately identify high-density core regions of feature data distribution, accurately capturing feature clustering patterns. It generates a set of partition identifiers to achieve effective partitioning of the feature space, providing a partitioning basis for subsequent partition management and compensation coefficient generation. It clarifies the core location and boundary range of each partition, defining the feature coverage area of ​​the partition. It clarifies the spatial topological relationships between partitions, determining the degree of correlation and mutual influence between different partitions, providing spatial parameter support for the formulation of partition configuration strategies. The system integrates key information such as partition centroids, boundary ranges, and spatial topological relationships to form a logically complete and parameter-clear partition configuration strategy. It provides a structured execution basis for the generation of dynamic compensation coefficients and feature calibration, standardizing the subsequent feature optimization process.

[0073] Extracting feature data subsets from each partition to focus on the characteristic features within each partition; calculating the mean and variance through statistical analysis to quantify the central tendency and dispersion of feature data within each partition, providing data support for stability assessment; the ratio of variance to mean objectively reflects the fluctuation range of feature values ​​within a partition, quantifying the partition's stability level; forming intuitive partition stability indicators to provide a quantitative basis for generating dynamic compensation coefficients, ensuring the targeted nature of the compensation strategy; converting partition stability indicators into corresponding dynamic compensation coefficients based on a preset mapping relationship to achieve dynamic adaptation of compensation coefficients; ensuring precise matching of compensation coefficients with the fluctuation state of partition features, providing adjustment parameters for feature calibration; and distributing the dynamic compensation coefficients according to partitions. The configuration strategy is organized into partitioned compensation vectors, standardizing the structural form of the compensation coefficients; this facilitates accurate matching between the compensation coefficients and the corresponding partitioned feature vectors, improving the execution efficiency of subsequent calibration operations; it establishes the mapping relationship between each feature vector in the basic feature set and the partitioned compensation vectors, determining the corresponding compensation rule for each feature vector; it achieves weighted fusion of feature vectors and compensation vectors through dot product operations, adjusting the numerical distribution of feature values; it optimizes the representation effect of feature vectors on the operating status of the power distribution system, strengthens the signal strength of effective features, and weakens the influence of interference factors; it integrates all weighted calibrated feature vectors to form an optimized feature set that is complete in information, accurate in representation, and highly adaptable, improving the adaptability of the feature set to real-time operating conditions.

[0074] In a preferred embodiment of the present invention, step 400 involves performing feature parsing and state mapping on the optimized feature set to obtain an operating state vector; performing multi-objective optimization calculations based on the operating state vector to obtain a preliminary control parameter set; and performing instruction encoding and strategy fusion on the preliminary control parameter set to obtain power distribution management instructions, including:

[0075] Step 401 involves analyzing the feature dimensions of the optimized feature set, extracting key operational features, and establishing a feature weight matrix. Specifically, this includes: analyzing each feature dimension in the optimized feature set, categorizing features according to their physical meaning into power supply status features, load characteristic features, renewable energy output features, equipment operation features, and anomaly correlation features; extracting key operational features that play a decisive role in the operating status of the power distribution system from each category, including voltage amplitude, effective current value, active power, reactive power, actual renewable energy output, load fluctuation coefficient, and equipment health indicators; calculating the weight values ​​of each key feature using a combination of mutual information and analytic hierarchy process (AHP), where mutual information quantifies the correlation between features and system operating status, and AHP adjusts weight allocation based on the priority requirements of power distribution management; arranging the weight values ​​of each key feature in order of feature dimensions to construct a feature weight matrix with dimensions consistent with the optimized feature set, where each element corresponds to the weight coefficient of a single feature, ensuring that the weight matrix reflects the degree of influence of different features on the operating status assessment, and providing a quantitative basis for subsequent weighted aggregation.

[0076] Step 402 involves weighted aggregation of the optimized feature set based on the feature weight matrix to obtain the operating state vector. Specifically, this includes: element-wise multiplication of each feature vector in the optimized feature set with the feature weight matrix, i.e., multiplying the value of each feature dimension with the corresponding weight coefficient in the weight matrix to enhance the representational role of high-weight key features; performing dimensional aggregation on the multiplied feature vectors, calculating the weighted mean of each feature category, and then combining the mean values ​​of each category in a preset order to form a fixed-dimensional aggregated vector; normalizing the aggregated vector, mapping the values ​​of each element in the vector to a preset interval to eliminate the influence of differences in the numerical ranges of different feature dimensions; through the above weighted aggregation and normalization operations, an operating state vector that comprehensively reflects the overall operating status of the power distribution system is obtained. This vector balances the contribution of key features with data consistency, providing concise and complete input data for multi-objective optimization calculations.

[0077] Step 403: Based on the operating state vector, construct a multi-objective optimization function with power supply reliability, economic operating cost, and renewable energy absorption rate as core objectives, and set weight coefficients for each objective according to operational requirements. Specifically, this includes: constructing a multi-objective optimization function based on the core feature parameters in the operating state vector. The function transforms the abstract control requirements of power supply reliability, economic operating cost, and renewable energy absorption rate into quantifiable and solvable mathematical logic, coordinates the balance of interests among different objectives, avoids system operational imbalance caused by single-objective optimization, and provides a unified objective-oriented framework for solving subsequent control parameters. The function has three core optimization objectives: power supply reliability, economic operating cost, and renewable energy absorption rate. The power supply reliability objective is quantified by parameters such as power outage duration, voltage qualification rate, and equipment failure rate. The economic operating cost objective is quantified by parameters such as electricity purchase cost, equipment operation and maintenance cost, and network loss cost. The energy absorption rate target is quantitatively represented by the ratio of actual renewable energy absorption to theoretical maximum output. The quantitative values ​​of each target reflect the actual achievement degree of each target under the current operating conditions, providing a quantitative basis for subsequent weighted calculations and iterative solutions, and making the optimization process measurable and comparable. The weight coefficients of each target are set according to the real-time operating requirements of the power distribution system. When the power consumption period is high, the weight coefficient of the power supply reliability target is increased; when the renewable energy output is high, the weight coefficient of the renewable energy absorption rate target is increased; and when there is a cost control requirement, the weight coefficient of the economic operating cost target is increased. The value range of the weight coefficients is set between 0 and 1, and the sum of the weight coefficients of the three targets is 1. The combination of these quantitative values ​​and weight coefficients can guide the optimization process to tilt towards the preset priority, ensuring that the optimization function is solved according to the preset priority, so that the final control parameters can adapt to the core requirements of the real-time operating scenario.

[0078] Step 404 involves iteratively solving the weighted multi-objective optimization function using an adaptive decision-making algorithm to obtain a set of candidate control parameters. Specifically, this includes: employing an adaptive decision-making algorithm adapted to the computing power of the HarmonyOS edge computing system. This algorithm integrates the parabolic tangent equation algorithm optimization iteration step size adjustment mechanism, dynamically adjusting the iteration step size and convergence threshold to adapt to the dynamic changes in the operating state vector, avoiding low solution efficiency or insufficient accuracy due to fixed parameters; and initializing the solution space boundary of the algorithm. This boundary is pre-set based on power grid safety operation specifications, equipment rated parameters, and management requirements, limiting the candidate control parameters. The range of values ​​includes core parameter intervals such as voltage regulation range, power output threshold, and equipment action time limit; the weighted multi-objective optimization function is used as the objective function of the algorithm, and the characteristic parameters corresponding to the operating state vector are input, including power supply reliability related parameters, economic cost parameters, and new energy consumption related parameters, to start the iterative solution process; during the iteration process, the objective function values ​​corresponding to the three consecutive iterations are extracted after every three consecutive iterations, and a local parabolic fitting model of the objective function is constructed based on these three sets of function values. The tangent slope of the fitting model at the current iteration point is calculated using the tangent equation algorithm of the parabola. The tangent slope reflects the rate of change of the objective function in the current iteration stage. The iteration step size is dynamically adjusted based on the tangent slope. If the absolute value of the tangent slope is greater than a preset slope threshold, it indicates that the objective function is changing drastically, and the iteration step size is reduced by a preset proportion to avoid overshooting the optimal solution during iteration. If the absolute value of the tangent slope is less than or equal to the preset slope threshold, it indicates that the objective function is changing gradually, and the iteration step size is increased by a preset proportion to accelerate the approach to the convergence direction. Simultaneously, the convergence threshold is dynamically adjusted based on the trend of the tangent slope change. When the slope gradually decreases and tends to stabilize, the convergence threshold is appropriately lowered to improve the solution accuracy. When the slope fluctuates significantly, the convergence threshold is maintained to ensure solution stability. The changing trend of the objective function is monitored in real time, and the iteration is stopped when the difference between the objective functions of two adjacent iterations is less than the adjusted convergence threshold. All non-dominated solutions generated during the iteration process are collected to form a candidate control parameter set, which covers optimization schemes under different objective combinations, providing a rich selection space for subsequent screening. At the same time, the lightweight design of the algorithm, combined with the efficient computational characteristics of the parabolic tangent equation, ensures that the solution process is completed at the edge, avoiding the delay caused by data uploading to the cloud, and adapting to the memory capacity and computing power limitations of edge terminals.

[0079] Step 405 involves verifying the constraints of the candidate control parameter set and screening feasible control parameters that meet the boundaries of safe grid operation. Specifically, this includes: pre-setting a set of constraints for safe grid operation, including voltage constraints, current constraints, power constraints, equipment capacity constraints, and frequency constraints, with each constraint having a set boundary threshold based on power system operation procedures and equipment technical parameters; verifying each control parameter in the candidate control parameter set one by one by substituting the control parameter into the corresponding constraint condition to determine whether the corresponding operating indicator is within the constraint boundary threshold range; specific verification includes whether the node voltage corresponding to the control parameter is within the allowable deviation range of the rated voltage, whether the line current does not exceed the rated current carrying capacity, whether the equipment output power does not exceed the rated capacity, and whether the system frequency is maintained within the standard frequency range; screening out all control parameters that meet all constraints, eliminating non-compliant parameters that exceed the constraint boundaries, forming a set of feasible control parameters to ensure that the subsequently generated control strategy meets the requirements of safe grid operation and avoids safety risks.

[0080] Step 406: Based on the weight coefficients of each objective, a weighted comprehensive evaluation of the feasible control parameters is performed to obtain a preliminary control parameter set. Specifically, this includes: establishing a weighted comprehensive evaluation model based on the weight coefficients of each objective set in step 403; calculating the individual evaluation score for each parameter in the feasible control parameter set under the three objectives of power supply reliability, economic operating cost, and new energy absorption rate, with the individual score obtained after standardization of the objective function value corresponding to the parameter; weighting and summing the three individual evaluation scores according to the weight coefficients to obtain the comprehensive evaluation total score for each feasible control parameter; sorting all feasible control parameters from highest to lowest comprehensive evaluation score, selecting a preset number of control parameters with the highest ranking, or selecting control parameters with a total score higher than a preset evaluation threshold; performing consistency verification on the selected control parameters to ensure no logical conflicts between parameters, and integrating them to form a preliminary control parameter set. This parameter set takes into account both multi-objective optimization needs and practical feasibility, providing a parameter basis for subsequent instruction generation.

[0081] Step 407 involves performing instruction encoding conversion on the preliminary control parameter set, mapping the preliminary control parameters into standardized equipment control instructions. Specifically, this includes: retrieving instruction encoding standards compatible with various power distribution terminals. These standards are based on the HarmonyOS system's device communication protocol and cover the instruction format requirements of terminals such as substation monitoring and control devices, photovoltaic inverters, charging pile controllers, and smart meters; encoding and converting each control parameter in the preliminary control parameter set, mapping the parameter's numerical information into standardized instruction fields, including instruction type identifier, parameter value field, execution time limit field, and terminal address field; determining the data type, byte length, and arrangement order of each field to ensure that the encoded instructions meet the terminal's reception and parsing requirements; adding an instruction verification field during the encoding process, which is calculated based on the instruction text using a preset hash algorithm and used for integrity verification during subsequent instruction transmission; through the above encoding conversion, the preliminary control parameters are mapped into standardized equipment control instructions with a unified structure and standardized format, realizing the conversion of control parameters into terminal-executable instructions and improving the compatibility of instructions with different terminals.

[0082] Step 408: Based on the pre-set power distribution network topology, strategy fusion is performed on the equipment control commands to obtain coordinated control power distribution management commands. Specifically, the pre-set power distribution network topology is completed in the following ways: First, a comprehensive collection of basic data of the power distribution system is performed, including the installation location coordinates of all power distribution equipment, electrical connection relationships (such as the connection method between busbars and lines, and the series and parallel connection relationships between equipment), line parameters (including line length, conductor cross-section, resistance and reactance values), equipment rated capacity, equipment model, voltage level, and geographical area division, etc. The collection method combines on-site survey data, equipment factory ledger import, and SC (System-Specific Data Collection). The ADA system synchronizes data in real time, ensuring data coverage of all elements of the power distribution network. The collected basic data undergoes structured processing, categorized by equipment type (e.g., transformers, switchgear, energy storage devices, new energy grid-connected equipment), voltage level (e.g., 10kV, 0.4kV), and geographical region. Unique equipment identifiers and relationship indexes are established, determining power transmission paths and constraints between equipment, forming a standardized topology data model. This topology data model is stored in the HarmonyOS distributed database, with data access interfaces configured to support fast querying and retrieval. A dynamic update mechanism is also established to respond when equipment additions occur in the power distribution network. When deleting, modifying lines, or adjusting parameters, change information is entered through the equipment management terminal, and topology data is updated in real time to ensure that the preset data is consistent with the actual power distribution network status. The preset power distribution network topology data is retrieved to determine the collaborative operation constraints between devices. Standardized equipment control commands are analyzed one by one, and the network topology is used to determine whether there are mutual influences between the equipment operations corresponding to each command, including line load distribution conflicts, equipment action timing conflicts, and voltage and current coordination conflicts. For equipment control commands with correlations, a collaborative fusion strategy is formulated to adjust the execution timing, parameter values, or action priorities of the commands. When there is a conflict between the tie switch switching command and the energy storage charging / discharging command, the charging / discharging power or switch switching time is adjusted according to the line load in the topology. The merged command is subjected to global consistency verification. Based on the topology, the system operation status after the command is executed is simulated to verify whether key indicators such as voltage, current, and power are within the safe threshold range, ensuring that the operation status of the power distribution system meets the requirements of safety, stability, and efficiency after all commands are executed. All coordinated commands are integrated to form a globally coordinated power distribution control command. This command can avoid logical conflicts between multiple device operations and ensure the coordination and stability of the overall operation of the power distribution system.

[0083] In this embodiment of the invention, the feature dimensions of the optimized feature set are analyzed to focus on the core operating indicators of the power distribution system and efficiently extract key operating features. A feature weight matrix is ​​established to quantify the influence of different features, highlighting the representational role of core features on the operating state and providing a basis for subsequent weighted aggregation, thus improving the pertinence and effectiveness of state mapping. Weighted aggregation of the optimized feature set is performed based on the feature weight matrix, strengthening the contribution of high-weight key features and weakening the interference of secondary features. A well-structured and information-condensed operating state vector is formed, comprehensively representing the overall operating status of the power distribution system and providing a simple and efficient input foundation for multi-objective optimization calculations. A multi-objective optimization function covering power supply reliability, economic operating cost, and renewable energy absorption rate is constructed, comprehensively covering the core needs of power distribution management. The weight coefficients of each objective are flexibly set according to the actual needs of different operating scenarios, matching the optimization direction with the management priority and providing objective guidance for subsequent solutions. The adaptive decision algorithm can dynamically adjust the iterative solution strategy to adapt to the dynamic changes of the operating state vector. The solution space is efficiently traversed to generate diverse candidate control parameter sets, fully exploring different combinations of objectives. The optimization potential is explored; based on the boundary conditions of voltage, current, and power for safe grid operation, the candidate control parameter set is verified one by one; parameters exceeding the safety threshold are eliminated to ensure that the selected feasible control parameters comply with the grid operation specifications and avoid safety risks caused by control actions; the feasible control parameters are comprehensively evaluated in combination with the preset weight coefficients of each objective to balance the interests among multiple objectives; the optimal parameters that take into account power supply reliability, economy, and new energy consumption are selected to form a logically unified and highly adaptable preliminary control parameter set; the preliminary control parameter set is converted into standardized equipment control instructions, and the encoding format and data structure of the instructions are unified; the instructions are adapted to the instruction reception and execution requirements of different types of distribution terminals to reduce the adaptation cost in the instruction transmission process and improve the compatibility and accuracy of instruction execution; based on the preset distribution network topology, the electrical connection relationship and coordination constraints between each device are determined; the standardized equipment control instructions are strategically integrated to avoid logical conflicts between multiple device control instructions; a globally coordinated distribution control instruction is formed to strengthen the systematicness and coordination of control actions and ensure the overall stable operation of the distribution system.

[0084] In a preferred embodiment of the present invention, step 500, which involves sending power distribution control instructions to the corresponding terminals and employing a timing synchronization and hash verification mechanism to ensure the sequential consistency and integrity of instruction transmission, includes:

[0085] Step 501: Receive the power distribution control instructions and sort and encapsulate them according to a preset instruction execution sequence to obtain an instruction execution queue. Specifically, this includes: receiving the power distribution control instructions generated in step 408, and retrieving a preset instruction execution sequence rule base. The instruction execution sequence is preset in the following way: First, comprehensively collect core basic data of the power distribution system, including the power distribution network topology, equipment types and technical parameters, equipment collaborative operation constraints, historical control task execution records, and safe operation procedure requirements. Data collection covers all power distribution terminals and control scenarios to ensure the comprehensiveness and consistency of the sequence setting. Adaptability; secondly, the power distribution control instructions are classified and sorted according to their functions and scope of impact into core safety instructions, important operation instructions, routine control instructions, and auxiliary monitoring instructions. Among them, core safety instructions include instructions that directly affect the safety and stability of the power grid, such as fault clearing, overload protection, and emergency voltage regulation; important operation instructions include instructions that affect the system's operating status, such as tie switch switching, energy storage system charging and discharging control, and new energy output regulation; routine control instructions include daily operation instructions such as load distribution fine-tuning and equipment parameter optimization; and auxiliary monitoring instructions include auxiliary instructions such as data acquisition and status feedback.

[0086] Next, based on the principle of safety first, instruction priority rules were formulated. Core safety instructions were set as the highest priority, important operation instructions as the second highest priority, routine control instructions as the medium priority, and auxiliary monitoring instructions as the lowest priority. Simultaneously, considering the power distribution network topology and equipment relationships, the execution logic order of instructions within the same priority was determined. For example, for equipment operation instructions on the same line, the logical order was ordered as disconnecting before closing and power supply side before load side. Then, based on equipment response characteristics, operation time consumption, and system stability requirements, the minimum execution time interval between different types of instructions was determined. This interval was based on equipment factory technical parameters, historical operating data statistics, and simulation experiments. For example, the execution interval for core safety instructions was no less than 50ms, and the interval for important operation instructions was no less than 100ms, ensuring sufficient response time for equipment to avoid action conflicts. Subsequently, the rationality of the timing sequence was verified through multi-scenario simulation tests. A simulation environment was built to simulate the instruction execution process under different operating conditions, detecting whether there were logical conflicts, action interference, or safety risks in the timing arrangement. Based on the test results, the priority order and time interval parameters were adjusted and optimized, iterating repeatedly until the timing scheme met all safety and operational requirements.

[0087] Finally, the optimized instruction priority rules, execution logic order, and time interval requirements are structurally integrated and stored to form an instruction execution timing rule library. A dynamic update mechanism is also established to promptly revise the timing rules and update the rule library when changes occur in the power distribution network structure, equipment configuration, or operational requirements. The rule library is pre-defined based on the power distribution network topology, equipment collaborative operation constraints, and management task priorities, determining the execution order logic and time interval requirements for different types of instructions. All power distribution management instructions are sorted according to the timing priority in the rule library, prioritizing core instructions related to power grid safety and stability, and then arranging auxiliary instructions according to equipment relationships and operational logic. The sorted instructions are encapsulated, with each encapsulation unit containing an instruction identifier, target terminal address, execution parameters, execution time limit, and associated instruction number. The encapsulated instructions are combined in the sorted order to form an instruction execution queue, with each instruction unit in the queue following a preset timing sequence to ensure no logical conflicts during multi-terminal instruction execution, laying an orderly foundation for subsequent timing synchronization transmission.

[0088] Step 502: A timestamp and sequence number are appended to each instruction in the instruction execution queue to form a time-stamped instruction stream. The time-stamped instruction stream is calculated using the SHA-256 hash algorithm to generate a digital fingerprint of the instruction data as the first verification code. Simultaneously, the CRC32 cyclic redundancy check algorithm is used to calculate the instruction data to generate a second verification code. Based on the first and second verification codes, an instruction data packet is obtained, specifically including: A high-precision timestamp is appended to each instruction in the instruction execution queue based on the global time synchronization mechanism of the HarmonyOS distributed soft bus. The precision of the timestamp is consistent with the time sequence identifier in step 104, ensuring uniform time dimensions throughout the process. Each instruction is assigned a continuously increasing sequence number according to the sorting order of the instruction execution queue. The sequence number contains... Instruction category identifiers and sequence numbers determine the transmission and execution order of instructions. All instructions with attached timestamps and sequence numbers are integrated in queue order to form a time-stamped instruction stream. The complete data of this instruction stream is calculated using the SHA-256 hash algorithm to generate a fixed-length digital fingerprint as the first layer of verification to prevent data tampering. At the same time, the CRC32 cyclic redundancy check algorithm is used to independently calculate the core data fields of each instruction to generate a second layer of verification to quickly detect data transmission errors. The time-stamped instruction stream, the first layer of verification, and the second layer of verification are integrated according to a preset data structure. The first layer of verification is stored in the header of the instruction data packet, and the second layer of verification is bound to the corresponding instruction data for storage, forming a structurally complete instruction data packet and constructing a dual data integrity protection mechanism.

[0089] Step 503: Establish a communication connection with the target terminal through the HarmonyOS distributed network, and send instruction data packets sequentially according to the order of the instruction execution queue. Specifically, this includes: using the device discovery mechanism of the HarmonyOS distributed network, locating and identifying the corresponding power distribution terminal based on the target terminal address in the instruction data packet; sending a connection request to the target terminal, the request carrying the device identity and encryption certificate of the initiating end; establishing a secure communication link after the target terminal is authenticated; synchronously detecting parameters such as network bandwidth, transmission latency, and signal strength during link establishment, and dynamically adjusting the transmission rate according to the parameter detection results to ensure adaptation to the communication conditions of different terminals; extracting instruction data packets sequentially from the queue according to the sorting order of the instruction execution queue, and sending them to the target terminal through the established secure communication link; using streaming transmission during the sending process, continuously pushing the sorted instruction data packets, and recording the sending time, sequence number, and target terminal information of each data packet to ensure that the instruction transmission order is completely consistent with the queue order, avoiding execution logic errors caused by out-of-order transmission.

[0090] Step 504: During the transmission of the instruction data packets, obtain the transmission status feedback of each instruction data packet. Specifically, this includes: during the transmission of the instruction data packets, enabling the transmission status feedback mechanism of the HarmonyOS distributed network, setting the feedback period to a preset time interval, which is calibrated based on the instruction transmission rate and terminal response characteristics; after each instruction data packet is sent, receive the transmission status feedback signal returned by the target terminal in real time. The feedback signal includes the data packet sequence number, reception status code, reception time, and link status parameters; the reception status code determines whether the data packet was successfully received, partially lost, or completely lost, and the link status parameters reflect the bandwidth, latency, and bit error rate of the current communication link; the received transmission status feedback signal is parsed in real time, key information is extracted and structured and stored in sequence number order to form a transmission status log, dynamically grasp the transmission progress, reception status, and link quality of each instruction data packet, and provide data support for subsequent anomaly handling.

[0091] Step 505: Based on the transmission status feedback, identify the instruction data packets that failed to transmit and initiate an automatic retransmission mechanism. Simultaneously, identify the instruction data packets that successfully transmitted and use the checksum returned by the receiving end to verify the integrity of the successfully transmitted instruction data packets, obtaining a data integrity confirmation signal. Specifically, this includes: analyzing the feedback information in the transmission status log one by one, identifying instruction data packets with a received status code indicating transmission failure, and recording their sequence number, target terminal, and reason for failure; initiating the automatic retransmission mechanism, retransmitting the failed data packets according to a preset retransmission limit, with the initial retransmission interval being the base delay, and subsequent retransmission intervals increasing exponentially. If the number of retransmissions exceeds a certain limit... If the transmission fails even after reaching the upper limit, an alarm mechanism is triggered and fault information is recorded. Simultaneously, a command data packet with a received status code indicating successful transmission is identified. A set of checksums is received from the target terminal, containing the SHA-256 hash value and CRC32 checksum calculated by the terminal based on the received data. The returned checksums are compared with the first and second checksums from the sender. If both checksums match, the command data is deemed complete and unaltered, and a data integrity confirmation signal is generated. If either checksum does not match, the data is deemed abnormal, the data packet is marked as a transmission failure, and included in the retransmission queue, ensuring the authenticity and integrity of the command data transmitted to the terminal.

[0092] Step 506 involves obtaining a successful transmission confirmation signal based on the data integrity confirmation signal. Specifically, this includes: collecting all data integrity confirmation signals, each containing the sequence number of the corresponding instruction data packet, the target terminal identifier, and the verification comparison result; filtering the confirmation signals, retaining only valid confirmation signals with consistent double verification comparisons; integrating the valid confirmation signals according to the sequence number of the instruction execution queue to form a structured set of successful transmission confirmation signals; this set clearly identifies all successfully transmitted and complete instruction data packets, including key information such as instruction identifier, sequence number, target terminal, and transmission completion time; and standardizing the set of successful transmission confirmation signals according to a preset format to ensure the signal format is compatible with the receiving requirements of subsequent execution feedback stages, thus building a clear bridge for instruction execution status tracking.

[0093] Step 507: Based on the successful transmission confirmation signal, receive confirmation feedback from the terminal indicating completion of execution, update the instruction execution status, and complete the closed-loop management of the entire process of power distribution control instruction issuance. Specifically, this includes: sending an execution status query request to the target terminal based on the successful transmission confirmation signal set; receiving real-time confirmation feedback from the terminal indicating completion of instruction execution, with feedback information including instruction sequence number, execution result status code, execution completion time, and a snapshot of equipment operating parameters; parsing the execution result status code to distinguish between three states: successful instruction execution, partial execution, and execution failure; recording the failure reason and abnormal equipment parameters if the execution fails; updating the instruction execution status database based on the feedback information, where each record contains full lifecycle information of the instruction, including generation time, transmission status, execution status, completion time, and abnormal records; archiving successfully executed instructions; initiating a fault investigation process for failed instructions; and generating processing suggestions based on the failure reason. Through status updates, log archiving, and fault handling, the entire closed-loop management of power distribution control instructions from issuance, transmission, execution to result feedback is completed, ensuring traceability and controllability of control actions and guaranteeing the integrity and reliability of the power distribution system control process.

[0094] In this embodiment of the invention, power distribution control instructions are sorted and encapsulated according to a preset instruction execution sequence to determine the logical order of instruction execution and avoid conflicts in instruction execution across multiple terminals. A well-structured instruction execution queue is formed, providing an orderly foundation for subsequent synchronous transmission and improving the systematic and collaborative nature of instruction issuance. Timestamps and sequence numbers provide unique timing identifiers for the instruction stream, ensuring the sequential correlation between instruction transmission and execution. A dual verification code mechanism forms complementary verification, strengthening the integrity protection of instruction data. Digital fingerprints and cyclic redundancy check results are integrated into the instruction data packet, constructing a multi-layered data security barrier and reducing the risk of tampering or damage during instruction transmission. The HarmonyOS distributed network adapts to the communication needs of multiple terminals, enabling rapid connection to target terminals. Instructions are sent sequentially according to the instruction execution queue order, ensuring consistency between instruction transmission and the preset execution sequence, improving the accuracy and timeliness of instruction issuance. Real-time feedback on the transmission status of instruction data packets is obtained to dynamically monitor transmission progress and anomalies. The system is designed to: promptly detect transmission interruptions and delays, providing real-time data support for subsequent targeted processing; quickly compensate for failed transmissions of command data using an automatic retransmission mechanism, reducing the impact of data loss on the control process; establish a closed loop between the receiving end's checksum verification and the sending end's dual checksum verification to confirm the integrity of command data; ensure that successfully transmitted command data is tamper-proof and complete, providing reliable assurance for accurate terminal execution; determine the core basis for successful command data transmission and integrate integrity verification results to form a standardized confirmation signal; build a clear bridge for subsequent execution feedback, strengthening the correlation between the transmission and execution processes and improving the overall efficiency of the process; receive feedback from the receiving terminal upon completion of execution, achieving closed-loop confirmation of command issuance and execution results; update command execution status in real time to fully grasp the implementation of control actions; and complete closed-loop management of the entire process, ensuring that the entire chain of power distribution control commands from issuance to execution is traceable and controllable, improving the integrity and controllability of the control process.

[0095] like Figure 2 As shown, embodiments of the present invention also provide a power distribution panoramic intelligent management and control decision data mining and processing system based on HarmonyOS, including:

[0096] The data acquisition module is used to access and acquire multi-source heterogeneous raw data from power distribution terminals via HarmonyOS distributed soft bus; it performs protocol parsing and unified encapsulation on the multi-source heterogeneous raw data to obtain standardized data packets; and it performs encryption and device authentication on the standardized data packets to obtain a secure encrypted data stream.

[0097] The standardization module is used to decrypt the secure encrypted data stream to obtain the decrypted data stream; perform multi-source protocol parsing on the decrypted data stream to obtain preliminary structured data; perform data cleaning and spatiotemporal alignment on the preliminary structured data to obtain a spatiotemporally consistent data sequence; and perform fragmentation and reassembly processing on the spatiotemporally consistent data sequence to obtain the standardized data stream.

[0098] The optimization module is used to input standardized data streams into a pre-trained lightweight data mining model, and perform load classification prediction, abnormal state detection, and new energy absorption capacity analysis at the HarmonyOS edge to obtain a basic feature set. Based on the basic feature set, a multi-dimensional feature space is constructed, and the multi-dimensional feature space is partitioned to obtain a partition configuration strategy. Dynamic compensation coefficients are generated according to the partition configuration strategy, and the basic feature set is calibrated in real time using the dynamic compensation coefficients to obtain an optimized feature set.

[0099] The fusion module is used to perform feature parsing and state mapping on the optimized feature set to obtain the operating state vector; perform multi-objective optimization calculation based on the operating state vector to obtain the preliminary control parameter set; and perform instruction encoding and strategy fusion on the preliminary control parameter set to obtain the power distribution management and control instructions.

[0100] The execution module is used to send power distribution control instructions to the corresponding terminals, and uses a timing synchronization and hash verification mechanism to ensure the consistency and integrity of the instruction transmission order.

[0101] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for data mining and processing of power distribution panoramic intelligent management and control decision-making based on HarmonyOS, characterized in that, The method includes: Step 100: Access and collect multi-source heterogeneous raw data from the power distribution terminal via HarmonyOS distributed soft bus; perform protocol parsing and unified encapsulation on the multi-source heterogeneous raw data to obtain standardized data packets; perform encryption and device authentication on the standardized data packets to obtain a secure encrypted data stream; Step 200: Decrypt the secure encrypted data stream to obtain a decrypted data stream; perform multi-source protocol parsing on the decrypted data stream to obtain preliminary structured data; perform data cleaning and spatiotemporal alignment on the preliminary structured data to obtain a spatiotemporally consistent data sequence; perform fragmentation and reassembly processing on the spatiotemporally consistent data sequence to obtain a standardized data stream. Step 300: Input the standardized data stream into the pre-trained lightweight data mining model, perform load classification prediction, abnormal state detection, and new energy absorption capacity analysis at the HarmonyOS edge to obtain a basic feature set; construct a multi-dimensional feature space based on the basic feature set, and partition the multi-dimensional feature space to obtain a partition configuration strategy; generate dynamic compensation coefficients according to the partition configuration strategy, and use the dynamic compensation coefficients to calibrate the basic feature set in real time to obtain an optimized feature set; Step 400: Perform feature parsing and state mapping on the optimized feature set to obtain the operating state vector; perform multi-objective optimization calculation based on the operating state vector to obtain the preliminary control parameter set; perform instruction encoding and strategy fusion on the preliminary control parameter set to obtain the power distribution management and control instructions. Step 500: The power distribution control command is sent to the corresponding terminal, and a timing synchronization and hash verification mechanism is used to ensure the consistency and integrity of the command transmission order.

2. The method for data mining and processing of power distribution panoramic intelligent management and control decision-making based on HarmonyOS according to claim 1, characterized in that, Step 100 includes: Collect multi-source heterogeneous raw data through HarmonyOS distributed soft bus; Identify multi-source heterogeneous raw data and distinguish it from IEC61850 protocol data, MQTT telemetry data and custom JSON format data; Multi-protocol parsing was performed on IEC61850 protocol data, MQTT telemetry data, and custom JSON format data to identify and extract valid fields. The valid fields are semantically reconstructed and formatted, and then encapsulated into a standardized data package with a unified time sequence identifier and a data source identifier. The standardized data packets are subjected to redundant compression and length alignment to obtain standardized data packets.

3. The method for data mining and processing of power distribution panoramic intelligent management and control decision-making based on HarmonyOS according to claim 2, characterized in that, Step 200 includes: The secure encrypted data stream is decrypted and its integrity is verified to obtain the decrypted data stream; The decrypted data stream is parsed using a multi-source protocol to identify the data source and format, resulting in preliminary structured data. Outlier detection and null value removal are performed on the preliminary structured data to obtain a valid data set; timestamp calibration and device spatial topology matching are performed on the valid data set to obtain a spatiotemporally consistent data sequence. Transmission characteristic analysis is performed on the spatiotemporally consistent data sequence to determine the fragmentation configuration parameters; the data sequence is dynamically fragmented based on the fragmentation configuration parameters to obtain fragmented data units; redundancy check flags are added to the fragmented data units and sequential encoding is performed to obtain data units to be reassembled. The data units to be reassembled are reassembled and their integrity verified to obtain a standardized data stream.

4. The method for data mining and processing of power distribution panoramic intelligent management and control decision-making based on HarmonyOS according to claim 3, characterized in that, Step 300 includes: The standardized data stream is input into a lightweight data mining model that integrates a pre-trained pruning decision tree and a deep forest to obtain a primary feature map. Based on the primary feature mapping, load pattern classification is performed through the hierarchical discriminative structure of the pruned decision tree to obtain the load state feature vector; The load state feature vector and the primary feature map are input together into the multi-granularity scanning mechanism of the deep forest to perform abnormal pattern recognition and obtain an abnormal index vector. Based on the load state feature vector and abnormal index vector, the characteristics of new energy output are analyzed by the regression analysis unit of a lightweight data mining model to obtain the absorption potential feature vector. The load status feature vector, abnormal index vector, and absorption potential feature vector are processed into feature vectorization to obtain vectorized features; the vectorized features are then weighted and fused to obtain a basic feature set.

5. The method for data mining and processing of power distribution panoramic intelligent management and control decision-making based on HarmonyOS according to claim 4, characterized in that, Step 300 also includes: Based on the aforementioned basic feature set, coordinate axes for each feature dimension are defined, and a multi-dimensional feature space is established. The basic feature set is mapped to the multi-dimensional feature space to form a feature data point distribution; A density-based spatial clustering method is used to perform cluster analysis on the distribution of the feature data points, identify high-density regions of the data distribution, and generate a set of partition identifiers; Calculate the centroid position and boundary range of each partition in the partition identifier set to determine the spatial topological relationship between the partitions; The centroid location, boundary range, and spatial topology of the partitions are integrated to obtain the partition configuration strategy.

6. The method for data mining and processing of power distribution panoramic intelligent management and control decision-making based on HarmonyOS according to claim 5, characterized in that, Step 300 also includes: Based on the partitioning configuration strategy, extract the feature data subsets corresponding to each partition; perform statistical analysis on the feature data subsets of each partition, and calculate the mean and variance of the feature values ​​within the partition respectively. Based on the mean and variance of the feature values ​​within a partition, the ratio of variance to mean is calculated as a partition stability index. Based on the partition stability index, a dynamic compensation coefficient is obtained through a preset mapping relationship; Receive the dynamic compensation coefficients and organize them into a partition compensation vector according to the partition configuration strategy; Based on the partitioning configuration strategy, establish the correspondence between each feature vector in the basic feature set and the partitioning compensation vector; Based on the correspondence between each feature vector and the partition compensation vector, each feature vector is multiplied by its corresponding partition compensation vector to complete the weighted calibration of the feature values ​​and obtain the weighted calibrated feature vector. Integrate all weighted and calibrated feature vectors to obtain an optimized feature set.

7. The method for data mining and processing of power distribution panoramic intelligent management and control decision-making based on HarmonyOS according to claim 6, characterized in that, Step 400 includes: The optimized feature set is analyzed for feature dimensions to extract key operational features and establish a feature weight matrix; The optimized feature set is weighted and aggregated based on the feature weight matrix to obtain the running state vector; Based on the aforementioned operating state vector, a multi-objective optimization function is constructed with power supply reliability, economic operating cost, and renewable energy absorption rate as the core objectives, and the weight coefficients of each objective are set according to the operating requirements. The weighted multi-objective optimization function is solved iteratively using an adaptive decision-making algorithm to obtain a set of candidate control parameters. The candidate control parameter set is subjected to constraint verification, and feasible control parameters that meet the power grid safe operation boundary are selected. Based on the weight coefficients of each objective, the feasible control parameters are evaluated by weighted comprehensive evaluation to obtain a preliminary set of control parameters; The preliminary control parameter set is converted into instruction encoding, mapping the preliminary control parameters into standardized equipment control instructions; Based on the pre-defined power distribution network topology, the equipment control commands are fused into strategies to obtain coordinated power distribution management commands.

8. The method for data mining and processing of power distribution panoramic intelligent management and control decision-making based on HarmonyOS according to claim 7, characterized in that, Step 500 includes: The power distribution control command is received, and the power distribution control command is sorted and encapsulated according to the preset command execution sequence to obtain the command execution queue; Each instruction in the instruction execution queue is appended with a timestamp and a sequence number to form a time-stamped instruction stream; the time-stamped instruction stream is calculated using the SHA-256 hash algorithm to generate a digital fingerprint of the instruction data as the first verification code; simultaneously, the instruction data is calculated using the CRC32 cyclic redundancy check algorithm to generate a second verification code; based on the first and second verification codes, the instruction data packet is obtained. A communication connection with the target terminal is established through the HarmonyOS distributed network, and instruction data packets are sent sequentially according to the order of the instruction execution queue. During the transmission of the instruction data packets, the transmission status feedback of each instruction data packet is obtained; Based on the transmission status feedback, the instruction data packets that failed to be transmitted are identified and an automatic retransmission mechanism is initiated. At the same time, the instruction data packets that were successfully transmitted are identified, and the integrity of the successfully transmitted instruction data packets is verified using the check code fed back by the receiving end, thereby obtaining a data integrity confirmation signal. A confirmation signal for successful transmission is obtained based on the data integrity confirmation signal; Based on the confirmation signal of successful transmission, the receiving terminal confirms the completion of execution, updates the instruction execution status, and completes the closed-loop management of the entire process of issuing power distribution control instructions.

9. A power distribution panoramic intelligent management and control decision data mining and processing system based on HarmonyOS, wherein the system implements the method as described in any one of claims 1 to 8, characterized in that, include: The acquisition module is used to access and acquire multi-source heterogeneous raw data from the power distribution terminal via the HarmonyOS distributed soft bus; it performs protocol parsing and unified encapsulation on the multi-source heterogeneous raw data to obtain standardized data packets. Encryption and device authentication are performed on standardized data packets to obtain a secure encrypted data stream; A standardized module is used to decrypt secure encrypted data streams to obtain decrypted data streams; Multi-source protocol parsing is performed on the decrypted data stream to obtain preliminary structured data; The preliminary structured data is cleaned and spatiotemporally aligned to obtain a spatiotemporally consistent data sequence; the spatiotemporally consistent data sequence is then fragmented and reassembled to obtain a standardized data stream. The optimization module is used to input standardized data streams into a pre-trained lightweight data mining model, and perform load classification prediction, abnormal state detection, and new energy consumption capacity analysis at the HarmonyOS edge to obtain a basic feature set. A multi-dimensional feature space is constructed based on the basic feature set, and the multi-dimensional feature space is partitioned to obtain the partitioning configuration strategy; Dynamic compensation coefficients are generated based on the partitioning configuration strategy, and the basic feature set is calibrated in real time using the dynamic compensation coefficients to obtain an optimized feature set. The fusion module is used to perform feature parsing and state mapping on the optimized feature set to obtain the operating state vector; perform multi-objective optimization calculation based on the operating state vector to obtain the preliminary control parameter set; and perform instruction encoding and strategy fusion on the preliminary control parameter set to obtain the power distribution management and control instructions. The execution module is used to send power distribution control instructions to the corresponding terminals, and uses a timing synchronization and hash verification mechanism to ensure the consistency and integrity of the instruction transmission order.

10. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 8.

Images