CN-121984226-A - Multi-source data fusion-based distribution network user side resource state monitoring method and system

CN121984226ACN 121984226 ACN121984226 ACN 121984226ACN-121984226-A

Abstract

The invention relates to a distribution network user side resource state monitoring method and system based on multi-source data fusion, comprising the following steps of S1, uniformly modeling user side resources according to equipment, measuring points, communication, topology and service to form traceable resource main data and relation mapping, S2, obtaining user side multi-source data, preprocessing the user side multi-source data to obtain preprocessed user side data, S3, constructing uniform feature vectors based on the preprocessed user side data, combining electric quantity, power quality indexes, equipment operation events, environmental factors and user behavior features, S4, obtaining state labels and probability distribution of the equipment based on a state identification model according to the uniform feature vectors, S5, evaluating the health degree of the user side resources according to the state labels and the probability distribution of the equipment to obtain health degree scores of the user side resources. The invention effectively improves the monitoring efficiency and reliability of the resource state of the distribution network user side.

Inventors

LIN HAN
WANG WEI
XU XINJIE
ZHANG YIQI
CHEN ZHIPENG
LIU CAI
WANG CHUANJIANG
CHEN TING
YANG ZHIPENG
XIA YULIAN

Assignees

国网信通亿力科技有限责任公司
国网信息通信产业集团有限公司

Dates

Publication Date: 20260505
Application Date: 20260129

Claims (8)

1. The method for monitoring the resource state of the distribution network user side based on the multi-source data fusion is characterized by comprising the following steps: S1, uniformly modeling a user side resource according to equipment, measuring points, communication, topology and service to form traceable resource main data and relation mapping; s2, acquiring multi-source data of a user side, and preprocessing to obtain preprocessed data of the user side; S3, constructing a unified feature vector based on the preprocessed user side data, and combining the electric quantity, the power quality index, the equipment operation event, the environmental factors and the user behavior characteristics; s4, acquiring state labels and probability distribution of each device based on a state identification model according to the unified feature vector; s5, evaluating the health degree of the user side resources according to the state labels of the devices and probability distribution thereof to obtain the health degree score of the user side resources.
2. The method for monitoring the resource state of the distribution network user side based on multi-source data fusion, which is disclosed by claim 1, is characterized in that the user side resource is uniformly modeled according to equipment, measuring points, communication, topology and service to form traceable resource main data and relation mapping, and specifically comprises the steps of distributing unique resource IDs for each inverter, BMS, charging piles, intelligent electric meters and home gateways of the user side, establishing a measuring point dictionary for each measuring point, defining units, dimensions, sampling periods, precision and physical meanings, and solidifying the corresponding relation between a meter and the same, loops, areas and feeder lines in the topology mapping.
3. The method for monitoring the resource state of the distribution network user side based on multi-source data fusion according to claim 1, wherein the multi-source data of the user side comprises electric quantity, voltage, current and power factor data of an electric meter side, direct current, alternating current power and operation modes of an inverter side, state of charge (SOC), state of charge (SOH), temperature and charge and discharge power of an energy storage side, state of an order gun of a charging pile side, output power, environmental and meteorological data and alarm logs and work order records of a service side.
4. The method for monitoring the resource state of the distribution network user side based on the multi-source data fusion of claim 3 comprises the following steps of realizing unified access to various communication modes and data formats through protocol adaptation and sending the unified access to a data bus in a streaming mode, denoising, limiting, anomaly filtering and characteristic pre-aggregation of original data at an edge node of a user side or a platform side, calibrating time stamps from different devices through a time synchronization mechanism and ensuring that subsequent fusion is carried out on the same time reference.
5. The method for monitoring the resource state of the distribution network user side based on the multi-source data fusion according to claim 1, wherein the method is characterized in that a unified feature vector is constructed based on the preprocessed user side data, and the electric quantity, the power quality index, the equipment operation event, the environmental factors and the user behavior feature are combined, specifically as follows: firstly, determining object granularity and time granularity of a feature structure, wherein the object granularity is selected according to single equipment, and each piece of data is accurately attributed to a corresponding object through the resource main data mapping of S1; Respectively constructing an electric quantity, a power quality index, an equipment operation event, an environmental factor and a user behavior characteristic; The method comprises the steps of completing the five types of feature construction, carrying out unified standardization and assembly on the features to form feature vector output with fixed dimension, wherein the method specifically comprises the steps of normalizing numerical features according to equipment types and dimensions, adopting one-hot coding for category and event features, marking missing or unavailable features with an explicit mask and adding update time, splicing static attributes in S1) as static features into vectors, and finally outputting a unified feature vector on each time window and each object granularity.
6. The method for monitoring the state of the resources on the user side of the distribution network based on the multi-source data fusion according to claim 1, wherein the method for acquiring the state label and the probability distribution of each device based on the state identification model according to the unified feature vector comprises the following steps: firstly, defining a state space and a hierarchy according to the equipment type and the business semantics of S1, defining a trigger condition, a continuous criterion and observable evidence to be bound for the state to avoid the synonym of different equipment names, and forming a configurable state machine constraint; The model reasoning stage receives the feature vector in a sliding time window mode and combines the static attribute of the equipment and the constraint of a state machine to judge the state, adopts a pre-trained supervised learning classification model LightGBM to output the posterior probability of the state of each equipment, introduces the constraint of time smoothing and consistency in the model output, eliminates the jitter of single-window erroneous judgment based on the post-treatment of a state transition matrix, and penalizes unreasonable jump; the output stage generates a structured result for each device, each time window, containing a state label, a full-scale state probability distribution, and confidence and interpretation information.
7. The method for monitoring the state of the resources on the user side of the distribution network based on the multi-source data fusion according to claim 6, wherein the health degree of the resources on the user side is evaluated according to the state labels of the devices and the probability distribution thereof to obtain the health degree score of the resources on the user side, and the method specifically comprises the following steps: Firstly, establishing health index systems and working condition baselines for different equipment types, and dividing the health index systems and working condition baselines into electric performance types, electric energy quality types, heat and ageing related types and event and reliability types; the sub-models corresponding to the working conditions are weighted according to the state probability distribution of the S4, wherein when the state probability concentration is higher than a preset threshold value, the deviation degree of each index is calculated by adopting a base line and the threshold value of the corresponding working conditions and mapped into sub-health scores; And finally outputting the health score and composition explanation thereof to form a result set which can be directly used for operation and maintenance and asset management.
8. The distribution network user side resource state monitoring system based on multi-source data fusion is characterized by comprising a processor, a memory and a computer program stored on the memory, wherein the steps in the distribution network user side resource state monitoring method based on multi-source data fusion are specifically executed when the processor executes the computer program.

Description

Multi-source data fusion-based distribution network user side resource state monitoring method and system Technical Field The invention relates to the field of intelligent monitoring, in particular to a method and a system for monitoring a resource state of a distribution network user side based on multi-source data fusion. Background Along with the promotion of double-carbon targets and the rapid access of distributed energy sources, the scale of user side resources (such as distributed photovoltaics, household/industrial and commercial energy storage, electric vehicle charging piles, controllable loads, intelligent home and electricity consumption acquisition terminals and the like) of the power distribution network is continuously increased, and the resource running state has the characteristics of large quantity, miscellaneous types, scattered access and rapid working condition change. On one hand, the user side resource has the adjusting and supporting capability, and on the other hand, the problems of voltage out-of-limit, reverse tide, harmonic pollution, three-phase unbalance, frequent start and stop and the like can be caused, so that the operation of the distribution network is gradually evolved from the traditional unidirectional power supply and predictable load to a complex scene of 'source network load storage interaction, strong fluctuation and strong uncertainty'. Therefore, there is a need to establish user-side resource oriented condition monitoring and health assessment capabilities to support fine-grained operations, proactive operations, and risk advancement. The key technical difficulty faced by the current-stage user-side resource state monitoring is multi-source heterogeneous and difficult fusion of data. Different manufacturer devices use different communication protocols and data formats, the sampling period is inconsistent with a time stamp system, the same physical quantity can be reported by a plurality of end sides and has deviation, and meanwhile, due to the influence of communication quality, field environment and equipment aging, data are frequently lost, noisy, drifting and abnormal points. On the other hand, the association relationship between the user side resource and the distribution network topology, the metering point location and the business event (such as worksheets, alarms and policy instructions) is often incomplete or difficult to trace back, so that false alarm and false alarm easily occur in a monitoring method which only depends on a single data source or a simple threshold rule, the reasons are difficult to explain, objects are difficult to position, and executable treatment suggestions are difficult to form. Disclosure of Invention In order to solve the problems, the invention aims to provide a method and a system for monitoring the state of a distribution network user side resource based on multi-source data fusion, which effectively improve the efficiency and the reliability of monitoring the state of the user side resource. In order to achieve the above purpose, the present invention adopts the following technical scheme: a method for monitoring the resource state of a distribution network user side based on multi-source data fusion comprises the following steps: S1, uniformly modeling a user side resource according to equipment, measuring points, communication, topology and service to form traceable resource main data and relation mapping; s2, acquiring multi-source data of a user side, and preprocessing to obtain preprocessed data of the user side; S3, constructing a unified feature vector based on the preprocessed user side data, and combining the electric quantity, the power quality index, the equipment operation event, the environmental factors and the user behavior characteristics; s4, acquiring state labels and probability distribution of each device based on a state identification model according to the unified feature vector; s5, evaluating the health degree of the user side resources according to the state labels of the devices and probability distribution thereof to obtain the health degree score of the user side resources. Further, unified modeling is conducted on user side resources according to equipment, measuring points, communication, topology and service to form traceable resource main data and relation mapping, and the method comprises the steps of distributing unique resource IDs for each inverter, BMS, charging piles, intelligent electric meters and home gateways on the user side, building a measuring point dictionary for each measuring point, defining units, dimensions, sampling periods, precision and physical meanings, and solidifying corresponding relations between the meters and the same, loops, areas and feeder lines in the topology mapping. Further, the multi-source data of the user side comprises electric quantity, voltage, current and power factor data of the ammeter side, direct current, alternating current power and operat