CN-122001029-A - Adjusting capability estimation method for flexible adjustable resources of power distribution area

CN122001029ACN 122001029 ACN122001029 ACN 122001029ACN-122001029-A

Abstract

The invention discloses an adjustment capability estimation method for flexible adjustable resources of a power distribution area, which belongs to the technical field of power distribution network monitoring and intelligent scheduling and comprises the steps of collecting operation data and environment data, synchronizing to obtain a multi-source measurement sequence, constructing a three-layer flexible resource map of the power distribution area, updating edge weights, establishing a physical mechanism model for nodes of flexible adjustable resource equipment according to the multi-source measurement sequence to generate physical prior characteristics, inputting the three-layer flexible resource map of the power distribution area into a physical mechanism constraint map neural network, obtaining node characterization vectors based on the physical prior characteristics, identifying flexible adjustable resource types, installed capacity, adjustment capacity estimation and operation states in each node, obtaining flexible adjustable resource identification results and generating a flexible adjustable resource adjustment capability curve of the power distribution area. The invention solves the problems that the prior art is difficult to accurately identify and evaluate the flexible adjustable resources of the power distribution area and lacks the constraint of a physical mechanism.

Inventors

LU XIAOXING
XIAO XIAOLONG
XIE WENQIANG
YU WENBIN
GUO ZIRAN
Lv Shukang
WANG YUN

Assignees

国网江苏省电力有限公司电力科学研究院
南京信息工程大学
江苏省电力试验研究院有限公司

Dates

Publication Date: 20260508
Application Date: 20260410

Claims (10)

1. A method for estimating the conditioning capacity of flexible and adjustable resources for a power distribution area, comprising: collecting operation data and environment data taking a distribution area as a unit, and performing time synchronization to obtain a multi-source measurement sequence; constructing a three-layer flexible resource map of the transformer area based on the distribution transformer node, the branch node, the user node and the flexible adjustable resource equipment node, and dynamically updating the edge weight of the flexible resource map of the three-layer area based on the time correlation of the power mutation event and the voltage fluctuation event; Performing feature extraction on the multi-source measurement sequence based on a time scale to obtain time sequence features, and establishing a physical mechanism model aiming at flexible adjustable resource equipment nodes to generate physical priori features; inputting the three-layer flexible resource map with updated edge weights into a physical mechanism constraint map neural network, and obtaining node characterization vectors based on physical priori features; Constructing a multi-task learning head based on the node characterization vector to identify the flexible adjustable resource category, the installed capacity, the adjustment capacity estimation and the running state in each node, so as to obtain a flexible adjustable resource identification result; Under the condition that physical mechanism constraint and measurement consistency are met, estimating the adjustable capacity, the adjustable capacity and the sustainable time of each node in a preset time window based on the flexible adjustable resource identification result, and generating a flexible adjustable resource adjustment capacity curve of the platform region.
2. The method for estimating capacity of flexible and adjustable resources for a power distribution area according to claim 1, further comprising storing operation data, environment data, multi-source measurement sequences, three-layer area flexible resource diagrams, physical prior characteristics, node characterization vectors, flexible and adjustable resource identification results, and an area flexible and adjustable resource capacity curve in an area sample library, and introducing an online incremental learning mechanism to update the area sample library.
3. The method for estimating capacity for flexibly adjustable resources of a power distribution block according to claim 1, wherein collecting operation data and environmental data in units of the power distribution block comprises: collecting, within a target power distribution area, area-level operation data, branch-level operation data, user-level operation data and equipment-level operation data which take the power distribution area as a unit, and environment data comprising meteorological elements and time features, wherein the meteorological elements comprise illumination intensity and air temperature, and the time features comprise day types; the transformer area level operation data comprise three-phase voltage data, three-phase current data, active power data, reactive power data and harmonic data of the distribution transformer; the branch stage operation data comprise branch current data, branch voltage data and event record data; the user-level operation data comprise electricity consumption data and load curve data; The equipment-level operation data comprise at least one of output power data and state data of a photovoltaic inverter, an energy storage converter or an electric vehicle charging pile.
4. The method for estimating capacity of flexible and adjustable resources of a power distribution substation according to claim 3, wherein constructing a three-layer flexible resource map based on a distribution transformer node, a branch node, a user node and a flexible and adjustable resource equipment node, and dynamically updating edge weights of the three-layer flexible resource map based on time correlation of power abrupt events and voltage fluctuation events, comprises: taking a configuration transformer node, a branch node, a user node and a flexible adjustable resource equipment node as graph nodes; Establishing an initial edge according to the physical connection relation of the distribution network among the graph nodes to form a three-layer flexible resource graph of the platform area; Calculating event correlation among graph nodes based on time synchronicity and change amplitude values of total power of a platform area, branch power, user power and flexible resource power at occurrence time of a load abrupt change event and a voltage fluctuation event; and updating the corresponding edge weight of the flexible resource map of the three layers of areas based on the event correlation among the map nodes.
5. The method for estimating capacity of flexible and adjustable resources for power distribution areas according to claim 4, wherein calculating the event correlation between graph nodes based on the time synchronicity and the variation amplitude of the total power of the areas, the branch power, the user power and the flexible resource power at the occurrence time of the load abrupt event and the voltage fluctuation event comprises: when detecting a power mutation event of the user-level power at the current moment, calculating the total power of a foreground region and a background region at the current moment and the response amplitude and time delay of the power of each branch, and calculating the event correlation degree between the current user and each branch according to the response amplitude and the time delay; when detecting a voltage drop event or a voltage fluctuation event of a transformer area, analyzing the time sequence and the amplitude of the voltage change of each node, and calculating the event correlation degree between the current user and each branch according to the time sequence and the amplitude; and the event correlation between the user and each branch is adopted to represent the event correlation between the nodes of the graph.
6. The method for estimating capacity of flexible and adjustable resources for a power distribution substation according to claim 4, wherein the flexible and adjustable resource equipment nodes comprise at least one of distributed photovoltaics, user-side energy storage devices, electric car charging piles and adjustable loads, and the physical mechanism model comprises at least one of a photovoltaic output model, an energy storage charging and discharging model, an electric car charging model and an adjustable load operation model corresponding to the flexible and adjustable resource equipment nodes: The photovoltaic output model is used for representing the relation between the photovoltaic output power and the illumination intensity, the temperature and the installed capacity; the energy storage charging and discharging model is used for representing the relation between energy storage charging and discharging power and state of charge, converter capacity and scheduling instructions; the electric automobile charging model is used for representing a typical charging curve of the change of charging power with time; The adjustable load operation model is used for representing an adjustable power interval and minimum continuous operation time and minimum downtime constraint.
7. The method for estimating capacity for flexible adjustable resources for a power distribution substation according to claim 6, wherein the time scale comprises a first time scale component, a second time scale component and a third time scale component, and wherein the time scale dividing method is as follows: setting a sampling interval delta t; setting a first preset time threshold T1 and a second preset time threshold T2, defining a time scale, and setting 0< T1< T2; Decomposing the multi-source measurement sequence into signal components with different frequencies based on a sampling interval delta t by adopting at least one method of wavelet decomposition, empirical mode decomposition or frequency band filtering; calculating the average period of each signal component; defining a signal component with the average period not greater than T1 as a first time scale component for characterizing high-frequency fluctuation characteristics from seconds to minutes; defining a signal component with an average period greater than T1 and not greater than T2 as a second time scale component for characterizing the intermediate frequency adjustment characteristics of the minute to hour scale; The signal component with an average period greater than T2 is defined as a third time scale component for characterizing low frequency trend characteristics at and above the hour level.
8. The method for estimating capacity for flexible adjustable resources of a power distribution substation according to claim 7, wherein the extracting the characteristics of the multi-source measurement sequence based on the time scale to obtain the time sequence characteristics comprises: each time scale component time domain feature and/or frequency domain feature is extracted, wherein the time domain feature comprises at least one of mean value, variance, skewness, kurtosis and zero crossing rate, and the frequency domain feature comprises at least one of spectrum energy, main frequency amplitude and spectrum quality.
9. The method for estimating capacity of flexible and adjustable resources of power distribution area according to claim 8, wherein inputting the three-layer flexible resource map with updated edge weights into the physical mechanism constraint map neural network, obtaining node characterization vectors based on physical prior features, comprises: Inputting the three-layer flexible resource map with updated edge weights into a physical mechanism constraint map neural network, and extracting time sequence characteristics and event correlation between map nodes; In the process of training the physical mechanism constraint graph neural network, a physical mechanism constraint loss function is constructed based on physical priori characteristics so as to restrict the photovoltaic power generation capacity not to exceed the rated capacity of the photovoltaic power generation capacity, the charge and discharge power of the constraint energy storage meets the state of charge constraint and the power constraint, the charge power of the constraint electric automobile accords with a typical charge power curve, the power change of the constraint adjustable load meets the operation constraint, and the node representation vector is output.
10. The method for estimating capacity of flexible and adjustable resources for a power distribution area according to claim 9, wherein the multi-task learning head comprises a classification subtask for identifying flexible and adjustable resource types, a regression subtask for estimating installed capacity and adjustment capacity, and a state discrimination subtask for identifying operation states, wherein each subtask shares the node characterization vector and performs joint training.

Description

Adjusting capability estimation method for flexible adjustable resources of power distribution area Technical Field The invention relates to a power distribution area flexible adjustable resource adjustment capability estimation method, and belongs to the technical field of power distribution network monitoring and intelligent scheduling. Background With the massive access of distributed photovoltaics, user side energy storage, electric automobile charging piles and various adjustable loads in a power distribution area, the area side resource structure and the running state are increasingly complex. The method is an important basis for realizing safe and stable operation of the power distribution network, flexible scheduling and collaborative optimization of source network load storage. In the prior art, related researches on flexible resources mainly comprise the following categories (1) a flexible resource importance evaluation method of a planning level, which takes flexible resources such as energy storage, electric vehicles and the like as objects in a power system planning or strategy evaluation level, analyzes the influence degree of different flexible resources on a planning target by utilizing methods such as entropy weight, gray correlation degree and the like to obtain importance sequencing, but the methods are mostly based on a planning model and statistical parameters, do not identify the flexible resources node by node aiming at a specific power distribution area, and lack fine granularity evaluation on real-time adjustment capability; the method comprises the steps of (2) carrying out information modeling and access on distributed power supplies, energy storage and loads in a power distribution network by building a power grid service center, realizing elastic analysis and elastic resource map construction, wherein the scheme is focused on data convergence and service flow integration, flexible resource information mainly depends on equipment account and service system configuration, lacks automatic identification and correction mechanisms based on actual operation measurement data, is difficult to find unrendered or parameter-changed flexible resources, (3) carrying out statistical analysis or clustering regression only by using a total active power, reactive power and voltage curve of a center, estimating capacity or output of resources such as photovoltaic and load, and the like, and does not explicitly introduce physical topology and resource operation mechanisms of the power distribution network due to underutilization of branches and household data, and when the types of resources in the center are complex and the data quality are unbalanced, (4) restricting application of a graph neural network in the power distribution network to physical mechanism constraint graph in the power distribution network, wherein the application of the power distribution network is used for topology identification of the power distribution network in recent years, voltage prediction and other problems are modeled by representing electrical nodes and connection relations thereof as graph structures, but related researches focus on the inference and control of topological structures or voltage states, and a system framework facing the identification and adjustment capability assessment of flexible adjustable resources of a platform region is not formed. In view of the above, the prior art has the following defects that (1) a unified identification and adjustment capability estimation method for flexible adjustable resources such as photovoltaic, energy storage, electric vehicle charging piles and adjustable loads is lacked by taking a power distribution area as a basic unit, (2) area-level, branch-level, household-level and equipment-level multi-source heterogeneous measurement data are not fully fused, robust identification is difficult to be carried out under the conditions of different sampling frequencies, data loss and different quality, 3) a flexible resource identification algorithm is mostly driven by pure data, and the result is lacked in engineering credibility and cross-scene mobility due to the lack of physical mechanism constraints such as photovoltaic output, energy storage SOC constraint, electric vehicle charging curves and adjustable load operation boundaries, and (4) a systematic area sample library and an online increment learning mechanism are lacked, so that a model is difficult to continuously update along with the change of an area resource structure and user behaviors. Therefore, the accurate identification and adjustment capability assessment of the flexible adjustable resources of the power distribution area are difficult to realize in the prior art, and the physical mechanism constraint is generally lacking, so that a special method and a system for the flexible adjustable resources of the power distribution area are necessary to be provided, and the accurate identification and adju