Power oscillation data processing method and device, storage medium and electronic equipment

Abstract

The application discloses a power oscillation data processing method and device, a storage medium and an electronic device. The method comprises the following steps: determining a plurality of sites, and determining oscillation data corresponding to the plurality of sites respectively; inputting the oscillation data corresponding to the plurality of sites respectively into a target knowledge graph representation learning model to obtain a target knowledge graph representing the relationship between the plurality of sites; and determining an oscillation propagation path of a target oscillation anomaly included in the oscillation anomaly corresponding to the plurality of sites respectively based on the oscillation data corresponding to the plurality of sites respectively. The application solves the technical problem of lacking the ability of monitoring information analysis and display for wideband oscillation in the related art.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power systems, in particular to a power oscillation data processing method and device, a storage medium and an electronic equipment. BACKGROUND

[0002] In the new power system, the proportion of new energy is high, and a large number of new energy sources such as ultra-high voltage direct current, flexible alternating current transmission system (FACTS), energy storage devices, electric vehicles, etc. emerge in an endless stream, which makes the source, network and load of each link present the trend of power electronicization, and the system is prone to stable oscillation. For example, the wind farm in some areas has 250-350Hz oscillation; the photovoltaic station has 30Hz subsynchronous oscillation; the large-scale wind power / photovoltaic power collection area of the power grid finds 10Hz, 20Hz, 30Hz, 40Hz, 60Hz, 70Hz, 80Hz, 90Hz, 110Hz and other subsynchronous / ultra-synchronous frequency components, which are transmitted in the 5 voltage level power grids of 35kV, 330kV and 750kV in the new energy station; the DC project has 1200Hz harmonic oscillation, 1800Hz high-frequency oscillation, etc. The above makes the oscillation of the power grid gradually extend to a wide frequency band, which seriously threatens the safe operation of the system. With the increase of the scale of new energy in the new power system, the scale of system oscillation may gradually expand, which seriously affects the safety of equipment and threatens the stable operation of the power grid, and becomes an important factor restricting the efficient consumption of new energy.

[0003] The phasor measurement device (PMU) in the related technology can realize low-frequency oscillation monitoring, and can realize 10Hz-40Hz subsynchronous oscillation and 50Hz-90Hz ultrasonic oscillation monitoring after upgrading. Correspondingly, the related technology also brings new problems. A large amount of wide-frequency oscillation data is uploaded to the dispatching master station, the existing master station lacks the ability of analyzing and displaying the wide-frequency oscillation monitoring information, it is difficult to trace the oscillation data, and it is not conducive to fault positioning and emergency handling by the related technical personnel.

[0004] At present, no effective solution has been proposed for the above problems. SUMMARY

[0005] The embodiments of the present application provide a power oscillation data processing method and device, a storage medium and an electronic equipment to at least solve the technical problem that the related technology lacks the ability of analyzing and displaying the wide-frequency oscillation monitoring information.

[0006] According to an aspect of some embodiments of the present application, there is provided a power oscillation data processing method, comprising: determining a plurality of sites, and determining oscillation data corresponding to the plurality of sites respectively, wherein the oscillation data is used to represent power data generated by an electrical signal oscillation anomaly occurring at a corresponding site; inputting the oscillation data corresponding to the plurality of sites respectively into a target knowledge graph representation learning model to obtain a target knowledge graph representing relationships between the plurality of sites, wherein the target knowledge graph comprises entities and relationships, the entities are the oscillation data corresponding to the plurality of sites respectively, and the relationships are connection relationships between the plurality of sites; determining, for a target oscillation anomaly included in the oscillation anomalies corresponding to the plurality of sites respectively, an oscillation propagation path of the target oscillation anomaly based on the oscillation data corresponding to the plurality of sites respectively; and displaying the oscillation propagation path on the target knowledge graph to obtain an oscillation display result of the target oscillation anomaly.

[0007] According to another aspect of some embodiments of the present application, there is provided a power oscillation data processing apparatus, comprising: a first determining module configured to determine a plurality of sites, and determine oscillation data corresponding to the plurality of sites respectively, wherein the oscillation data is used to represent power data generated by an electrical signal oscillation anomaly occurring at a corresponding site; an input module configured to input the oscillation data corresponding to the plurality of sites respectively into a target knowledge graph representation learning model to obtain a target knowledge graph representing relationships between the plurality of sites, wherein the target knowledge graph comprises entities and relationships, the entities are the oscillation data corresponding to the plurality of sites respectively, and the relationships are connection relationships between the plurality of sites; a second determining module configured to determine, for a target oscillation anomaly included in the oscillation anomalies corresponding to the plurality of sites respectively, an oscillation propagation path of the target oscillation anomaly based on the oscillation data corresponding to the plurality of sites respectively; and a result display module configured to display the oscillation propagation path on the target knowledge graph to obtain an oscillation display result of the target oscillation anomaly.

[0008] According to another aspect of some embodiments of the present application, there is provided a non-transitory storage medium storing a plurality of instructions adapted to be loaded and executed by a processor to perform any of the power oscillation data processing methods.

[0009] According to another aspect of some embodiments of the present application, there is provided an electronic device comprising: one or more processors and a memory configured to store one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement any of the power oscillation data processing methods.

[0010] In the embodiment of the present application, the propagation path of the oscillation is displayed on the target knowledge graph. The multiple sites are determined, and the oscillation data corresponding to the multiple sites is determined, wherein the oscillation data is used to represent the power data generated by the abnormal oscillation of the corresponding site. The oscillation data corresponding to the multiple sites is input into a target knowledge graph representation learning model to obtain a target knowledge graph representing the relationship between the multiple sites, wherein the target knowledge graph includes entities and relationships, the entities are the oscillation data corresponding to the multiple sites, and the relationships are the connection relationships between the multiple sites. For the target oscillation anomaly included in the oscillation anomaly corresponding to the multiple sites, the oscillation propagation path of the target oscillation anomaly is determined based on the oscillation data corresponding to the multiple sites. The oscillation propagation path is displayed on the target knowledge graph to obtain the oscillation display result of the target oscillation anomaly. The purpose of determining the oscillation propagation path of the target oscillation anomaly is achieved, the technical effect of displaying the oscillation of the target oscillation anomaly is achieved, and the technical problem of lacking the wideband oscillation monitoring information analysis and display capability in the related art is solved. BRIEF DESCRIPTION OF DRAWINGS

[0011] The accompanying drawings, which are included to provide a further understanding of the present application, constitute a part of this application, illustrate embodiments of the present application and the description thereof, and do not constitute improper limitations on the present application. In the drawings:

[0012] Figure 1 is a flowchart of an optional power oscillation data processing method according to an embodiment of the present application;

[0013] Figure 2 is a module schematic diagram of an optional power oscillation data processing method according to an embodiment of the present application;

[0014] Figure 3 is a structural block diagram of an optional power oscillation data processing device according to an embodiment of the present application. DETAILED DESCRIPTION

[0015] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0016] It should be noted that the terms "first", "second", and the like in the description and in the claims of the present application and above-described accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0017] According to an embodiment of the present application, a method embodiment of power oscillation data processing is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0018] Figure 1 is a flowchart of the power oscillation data processing method according to an embodiment of the present application, as shown in Figure 1 The method comprises the following steps:

[0019] Step S102, determine a plurality of sites, and determine a plurality of sites respectively corresponding to oscillation data, wherein the oscillation data is used to represent the power data generated by the corresponding site when the electrical signal oscillation anomaly occurs.

[0020] It can be understood that the plurality of sites are usually important nodes of power transmission and distribution, which can include power plants, substations, transmission lines, etc. For each site, it is necessary to collect the oscillation data related thereto. The oscillation data usually comes from sensors and monitoring devices installed on the site. These data can include the fluctuation of voltage, current, power, phase and other parameters, which reflect the specific performance of the electrical signal oscillation anomaly.

[0021] As an optional embodiment, the station can be a substation, and the oscillation data can be wide-band oscillation data, wherein the wide-band oscillation data includes wide-band oscillation alarm information and fault recording data. The wide-band oscillation alarm information includes oscillation occurrence time and oscillation alarm type of each electrical interval, and the fault recording data includes oscillation voltage and current measurement values, recording starting time and ending time of each electrical interval. The oscillation alarm type includes oscillation divergence, constant amplitude oscillation and oscillation convergence. The oscillation data of the station can be collected by using a vibration sensor or other devices. The devices are installed at different stations to record and store the oscillation data. Alternatively, the oscillation data can be obtained from a database storing the oscillation data. There are many ways to determine the oscillation data, and the above is only an example. Many other determination methods are not listed here.

[0022] As an optional embodiment, there are many ways to determine the oscillation data corresponding to the multiple stations, for example, the following determination method can be adopted: for one station of the multiple stations, the oscillation voltage signal and the oscillation current signal of the station under the condition of occurrence of corresponding oscillation anomaly are obtained; the discrete signal of the station is generated based on the oscillation voltage signal and the oscillation current signal; the oscillation data corresponding to the station is obtained by signal processing based on the discrete signal of the station; and the oscillation data corresponding to the multiple stations is obtained by using the method of obtaining the oscillation data corresponding to the station.

[0023] In the above optional embodiment, the oscillation voltage signal and the oscillation current signal of the station under the condition of occurrence of oscillation anomaly can be obtained by installing an oscillation monitoring device on the station. The oscillation data can be oscillation data of the same frequency band or oscillation data of different frequency bands. In the case that the oscillation data is in different frequency bands, the frequency range of 0Hz to 2500Hz can be defined by using appropriate low-pass, band-pass and high-pass filters to eliminate the frequency bands not concerned in the original sampling signal, and to eliminate the influence of noise interference and frequency aliasing. The oscillation data corresponding to the station can be quickly and accurately obtained by the above signal processing based on the discrete signal of the station. The running state of the substation equipment can be known through the oscillation data, and potential faults can be discovered and predicted in time.

[0024] Optionally, according to the Nyquist theorem, in order to ensure the integrity of the collected data signal, the sampling frequency fs.max is greater than 2 times the highest frequency fmax in the oscillation voltage signal or the oscillation current signal. Preferably, the maximum oscillation frequency of the wide-band oscillation is 2500Hz, and therefore the sampling frequency must be higher than 5000Hz.

[0025] As an optional embodiment, based on the discrete signal of a station, the signal processing is performed to obtain the oscillation data corresponding to the station, which can be performed in the following manner: based on the discrete signal of a station, the fast Fourier algorithm is used to process to obtain the frequency domain signal of the station; based on the frequency domain signal of a station, the interpolation compensation method is used to process to obtain the oscillation component, the oscillation amplitude and the oscillation initial phase of the station; based on the oscillation component, the oscillation amplitude and the oscillation initial phase of a station, the oscillation data corresponding to the station is determined.

[0026] In the optional embodiment, when the fast Fourier algorithm is used to process the discrete signal to obtain the frequency domain signal, the spectrum leakage may exist, therefore, the spectral line with the maximum amplitude in the oscillation inter-harmonic frequency of the frequency domain signal is obtained, the spectral line corresponds to the calculation result of the oscillation inter-harmonic component at the spectral line frequency. The compensation coefficient is obtained according to the spectral line with the maximum amplitude and the amplitudes of the adjacent two spectral lines, and the interpolation compensation method is used to process the frequency domain signal according to the compensation coefficient, so as to obtain the oscillation frequency, the amplitude and the initial phase of different oscillation types. The obtained oscillation component frequency, the amplitude and the initial phase are taken as the oscillation data corresponding to the station. The oscillation component frequency, the amplitude and the initial phase have a unified format. The unified format refers to the unified spatial naming and the standardized data format, which specifically includes the standardized naming of the busbar port oscillation voltage, the oscillation current, the oscillation frequency and the oscillation type, and is expressed in the structured and semi-structured data. The unified format can facilitate subsequent calculation. Through the above steps, the fast Fourier algorithm and the interpolation compensation method can obtain more accurate oscillation data, and can avoid the spectrum leakage.

[0027] Optionally, for determining the oscillation data corresponding to a station, in order to make the oscillation voltage signal and the oscillation current signal of each substation interval include the fundamental component and the oscillation component when the system oscillation alarm occurs, the discrete signal x[n] corresponding to the nth time window after sampling is expressed as:

[0028]

[0029] In the above formula, n is the time window, i.e. the sampling time, a is the fundamental component of the oscillation voltage signal or the oscillation current signal, a i is the amplitude of the i-th oscillation component of the oscillation voltage signal or the oscillation current signal, f is the oscillation frequency of the fundamental frequency of the oscillation voltage signal or the oscillation current signal, f i is the oscillation frequency of the i-th oscillation component of the oscillation voltage signal or the oscillation current signal, preferably f = 50 Hz, T s is the sampling interval, f s is the sampling frequency, is the initial phase of the fundamental component and the i-th oscillation component, respectively.

[0030] The oscillation data is classified and differentially processed, the fast Fourier algorithm and the interpolation compensation method are applied, and the oscillation frequency, amplitude and initial phase of the oscillation voltage and current measurement data in the 0Hz-2.5Hz low frequency oscillation, 2.5Hz-100Hz sub / super frequency oscillation and 100Hz-300Hz medium-high frequency oscillation range are extracted.

[0031] The discrete signal is converted into a frequency domain signal by using the fast Fourier algorithm, and the frequency domain signal is as follows:

[0032]

[0033] Wherein, X[n] is the frequency domain signal corresponding to the nth time window, r[n] is a window function, preferably a Hanning window, N r is the window length, 10 base wave time windows are used for 0Hz-2.5Hz low frequency oscillation signal and 2.5Hz-100Hz sub / super frequency oscillation signal, and 2 base wave time windows are used for 100Hz-2500Hz medium-high frequency oscillation signal, m is the frequency spectrum line number, m=1, 2, 3,..., N r -1. The spectrum line number with the maximum amplitude near the oscillation inter-harmonic frequency is M, and the spectrum line corresponds to the calculation result of the oscillation inter-harmonic component at the spectrum line frequency.

[0034] Considering that the fast Fourier algorithm may have spectrum leakage, the interpolation compensation method is applied to calculate the oscillation frequency, amplitude and initial phase of different oscillation types. According to the amplitude of the maximum amplitude spectrum line and the amplitudes of the adjacent two spectrum lines, the compensation coefficient is

[0035] The compensation coefficient is applied to determine the oscillation component f i , the oscillation amplitude a i and the oscillation initial phase

[0036]

[0037]

[0038]

[0039] Optionally, the unified format refers to a unified space naming and standardized data format, specifically including standardized naming of substation bus port oscillation voltage, current, oscillation frequency and oscillation type, expressed in structured and semi-structured data.

[0040] ​In step S104, the oscillation data corresponding to the plurality of stations is input into the target knowledge graph representation learning model to obtain a target knowledge graph representing the relationship between the plurality of stations, wherein the target knowledge graph includes entities and relationships, the entities are the oscillation data corresponding to the plurality of stations, and the relationships are the connection relationships between the plurality of stations.

[0041] It can be understood that by learning the characteristics and patterns of the oscillation data, the potential relationships and connections between the stations are extracted and represented as entities and relationships in the knowledge graph. The entities correspond to the oscillation data of each station, and the relationships describe the connection mode and degree between the entities. The model outputs a target knowledge graph representing the relationship between the plurality of stations, which includes the oscillation data of the stations and also shows the correlation and propagation patterns between the data.

[0042] As an optional embodiment, a preset time sequence knowledge graph is obtained, wherein the time sequence knowledge graph includes a plurality of reference entities, reference relationships between the plurality of reference entities, and a time stamp; a training sample set is generated based on the plurality of reference entities, the reference relationships between the plurality of reference entities, and the time stamp; an initial knowledge graph representation learning model and a target loss function are obtained, wherein the target loss function includes entity embedding representation loss and relationship embedding representation loss; the training sample set is used to train the initial knowledge graph representation learning model to obtain a candidate knowledge graph representation learning model, and the target loss function is updated to obtain an updated target loss function; and in a case where the updated target loss function is less than a predetermined loss threshold, the candidate knowledge graph representation learning model is determined as the target knowledge graph representation learning model.

[0043] In the above optional embodiment, the oscillation data in the transformer station is taken as an entity object, the connection relationship between each transformer station and the oscillation propagation path are taken as relationships, and the TransE-TAE model is used for solving and optimization to obtain the target knowledge graph representation learning model. The target knowledge graph representation learning model can learn the time sequence order relationship related to the time stamp in a specific time sequence space. The time stamp is the time when the oscillation occurs. The target loss function can be used to measure the difference between the prediction result of the candidate knowledge graph representation learning model and the true value, and is used to evaluate the performance of the candidate knowledge graph representation learning model. Through the target loss function, the candidate knowledge graph representation learning model can be continuously optimized in the training process, thereby improving the prediction ability and accuracy of the candidate knowledge graph representation learning model. The target loss function can use the mean square error method and the cross-entropy method, and there are many methods for calculating the target loss function, which are not listed here and can be set according to specific needs.

[0044] The TransE-TAE model is a knowledge graph-based representation learning model that combines the TransE model and the idea of an automatic encoder (TAE). Its main function is to learn the vector representation of entities and relationships in a knowledge graph, in order to perform representation learning and reasoning tasks on the knowledge graph. The TransE-TAE model represents entities and relationships as vectors and uses an automatic encoder to learn low-dimensional representations of these vectors, thereby better capturing the semantic relationships between entities and relationships. The TransE-TAE model can use the stochastic gradient descent method to solve the optimal value of the optimization objective function. Therefore, the TransE-TAE model can help us better understand the associations between entities and relationships in the knowledge graph, and can be used for reasoning, link prediction, and other tasks on the knowledge graph.

[0045] Optionally, a preset gallery module can be used to construct a knowledge graph database based on preprocessed data. The above-mentioned gallery module specifically applies a knowledge graph modeling tool or platform to model and abstract the preprocessed data, and establishes a knowledge graph database.

[0046] Through the above processing, the triple samples and relationship samples are generated as a training sample set, which can be used for subsequent training of the initial knowledge graph representation learning model, and then the target knowledge graph representation learning model is obtained. After inputting the above-mentioned training sample set into the initial knowledge graph representation learning model, optimization iteration is performed using a loss function method, and the loss function is set as the following objective function F.

[0047]

[0048] where q(x) is the scoring function of the TransE model, represented as q(ε i ,γ,ε k )=||ε i +γ-ε k ||, the relationship r k and r l share the same head entity e i . is the set of positive samples in the relationship sample, is the set of negative samples in the relationship sample.

[0049] In the training process, the TransE-TAE model uses the stochastic gradient descent method to solve the optimal value of the optimization objective function, and then determines the model iterated under the condition of the minimum loss function as the target knowledge graph representation learning model.

[0050] It should be noted that the time sequence knowledge graph is used to generate samples, and can generate triple samples and relation samples, and the triple samples and the relation samples are taken as a training sample set. The smaller the optimization objective function F is, the higher the accuracy of the conversion into a graph structure is. The formula F is composed of two parts of loss, including a feature embedding loss of the triple sample, and a feature embedding loss of the relation sample.

[0051] As an optional embodiment, based on the plurality of reference entities, the reference relationships between the plurality of reference entities, and the timestamps, the training sample set can be generated by using the following generation method. For one of the plurality of reference entities, based on the timestamps corresponding to the plurality of reference entities respectively, a previous entity before the time sequence of the one reference entity is determined, and a next entity after the time sequence of the one reference entity is determined. Based on the reference relationships between the plurality of reference entities, a first relationship between the previous entity and the one reference entity, and a second relationship between the one reference entity and the next entity are determined. Based on the first relationship and the second relationship, a relation sample corresponding to the one reference entity is generated. Based on the first relationship, the previous entity, and the one reference entity, a triple sample corresponding to the one reference entity is generated. The way of generating the relation sample and the triple sample by using the one reference entity is used to obtain the relation sample and the triple sample corresponding to the plurality of reference entities respectively. Based on the relation sample and the triple sample corresponding to the plurality of reference entities respectively, the training sample set is obtained.

[0052] In the above optional embodiment, the relation sample and the triple sample are two common data forms in machine learning, which are usually used to represent the relationship or connection between entities. The relation sample represents the relationship between entities, while the triple sample is a special form of the relation sample, usually consisting of subject, predicate and object. The relation sample is usually extracted from existing corpus or knowledge graph, and the entities and their relationships can be extracted from the text by natural language processing technology, and then the information is stored as a relation sample. The triple sample is a further abstracted data form based on the relation sample, usually consisting of subject, predicate and object, and is used to represent the relationship between entities. The triple sample can be obtained by abstracting and labeling the relation sample, usually by manual labeling or using natural language processing technology for automatic extraction. The relation sample can include positive samples and negative samples, and the negative samples can be obtained based on the positive samples. Through the relation sample and the triple sample, a rich training sample set is generated, which helps the model to better understand the relationship between entities and improves the accuracy and generalization ability of the model.

[0053] Optionally, based on the TransE-TAE model representation learning method, Δt represents an existing set of four-tuple instances in a given time sequence knowledge graph, and there are any two four-tuple instances (e i ,r k,e j ,t rk )∈Δt,(e i ,r l ,e m ,t rl )∈Δt, i.e., e i As a reference entity, e j For the previous entity, e m For the next entity, r k As the first relation, r l For the second relation, t rk For the timestamp of the first relation, t rl This is the timestamp of the second relation. Based on this, we can build upon the first relation r... k Second relation r l This generates a corresponding temporal sequence relation pair. <r k ,r l >

[0054] If t rk <t rl , use y + =(r k ,r l ) represents a positive sample in a temporal sequence relationship, denoted by y. - =(r k ,r l ) -1 This represents a negative sample constructed for the positive samples. The time-series scoring function is defined as: f( <r k ,r l >)=||r k Tr l ||. Here, T is the evolution matrix characterizing the temporal sequence relationship. T projects rk and rl onto a specific temporal vector space, indicating that the smaller the distance between them, the earlier rk occurs before rl, and vice versa. In the model, following the temporal sequence relationship chain, the score f-value for positive samples is sufficiently small, and the score f-value for negative samples is sufficiently large. The above y + y - All of them belong to relational samples.

[0055] Optionally, based on the time-series knowledge graph, ternary samples can also be established for e. i Let ej be a reference entity and ej be the previous entity. The first relation r between the two entities is... k It can generate a positive ternary sample x. + (e i ,r k ,e j), in this way, multiple positive triple samples can be generated for different entities as positive triple Δ. For positive triple sample x + (e i ,r k ,e j ) against x + Triple head, tail entity replacement, can get negative triple sample x - =(e' i ,r k ,e' j )∈Δ', where Δ' represents negative triple.

[0056] In step S106, for the target oscillation anomaly included in the oscillation anomalies corresponding to the multiple sites respectively, the oscillation propagation path of the target oscillation anomaly is determined based on the oscillation data corresponding to the multiple sites respectively.

[0057] It can be understood that in the oscillation data of multiple sites, specific target oscillation anomalies that need to be paid attention to are identified. These anomalies may be caused by specific reasons, or have special propagation characteristics. By analyzing the representation of the target oscillation anomaly in the knowledge graph, combining the oscillation data of the sites and the relationship between them, the propagation path of the anomaly is determined to reveal the propagation mode and path of the anomaly between different sites.

[0058] As an optional embodiment, based on the oscillation data corresponding to the multiple sites respectively, the oscillation propagation path of the target oscillation anomaly can be determined in the following manner: in the oscillation data corresponding to the multiple sites respectively, the target oscillation data matched with the target oscillation anomaly is determined, and the target site corresponding to the target oscillation data is determined; the target frequency range corresponding to the target oscillation data is determined; in the case that the target frequency range is a predetermined first oscillation frequency range, the port transient energy of the target site is determined; based on the port transient energy, the oscillation propagation path is determined.

[0059] In the optional embodiment described above, the port transient energy refers to the energy that appears temporarily on the port (input / output interface of the system or device). This energy is usually caused by the connection or disconnection of the port, the transmission or interference of the signal, etc. By determining the target oscillation data that matches the target oscillation anomaly and the corresponding target site, the source of the oscillation anomaly can be quickly locked. By determining the target frequency range corresponding to the target oscillation data, the abnormal frequency range can be more accurately located, thereby better analyzing and handling the abnormal situation. By measuring the port transient energy of the target site, the oscillation propagation path can be accurately and quickly determined, so that corresponding measures can be taken for processing and prevention. Through the above steps, the oscillation anomaly can be located and handled more quickly and accurately, improving the efficiency and accuracy of responding to abnormal situations, and helping to ensure the normal operation of the device and system. At the same time, by analyzing the oscillation propagation path, similar abnormal situations that may occur in the future can also be prevented, improving the stability and reliability of the system. Therefore, the method of determining the oscillation propagation path based on the port transient energy can effectively improve the efficiency and safety of the device and system.

[0060] As an optional embodiment, in the case where the target frequency range is a predetermined second oscillation frequency range, the port equivalent impedance of the target site is determined, wherein the first oscillation frequency range is smaller than the second oscillation frequency range; and the oscillation propagation path is determined based on the port equivalent impedance.

[0061] In the optional embodiment described above, in the case where the target frequency range is a predetermined second oscillation frequency range, the port equivalent impedance of the target site is determined, which refers to calculating the port equivalent impedance of the target site in a specific frequency range, i.e. the equivalent circuit parameters of the target site in this frequency range. The measurement and calculation can be performed by electromagnetic simulation software or network analysis instruments and other tools. By analyzing the equivalent circuit parameters of the target site, the propagation path and characteristics of the signal in the site can be determined. Based on the oscillation propagation path, the system oscillation monitoring and analysis results can be intuitively seen, which is beneficial for the operation personnel to quickly make oscillation suppression decisions and more effectively suppress system oscillation and restore stable operation of the system.

[0062] Step S108, the oscillation propagation path is displayed on the target knowledge graph to obtain the oscillation display result of the target oscillation anomaly.

[0063] It can be understood that the determined oscillation propagation path is visually displayed on the target knowledge graph, which can be optionally realized by highlighting the related entities and relationships, adding annotations or labels, etc. Through visual display, the propagation and influence range of the target oscillation anomaly in the power system can be intuitively understood. This is beneficial for quickly locating the abnormal source, evaluating the risk and taking corresponding measures.

[0064] As an optional embodiment, the oscillation propagation path can be displayed on the visualization interface, and the entities, attributes and relationships in the target knowledge graph, and the network structure of the target knowledge graph can be displayed. The target knowledge graph database and the calculated oscillation energy or impedance can also be displayed, and the broadband oscillation data in the transformer substation can also be comprehensively displayed. According to the actual scene requirements, the oscillation propagation path can be displayed by using a mesh graph, a tree graph and a timeline display graph. The mesh graph refers to representing the association between different entities by nodes and edges, and can be used to display the oscillation propagation path between transformer substations at a single oscillation frequency. The tree graph refers to classifying and layering entities according to certain rules to construct a tree structure, and clearly displaying the connection relationship between entities, and can be used to display the oscillation between transformer substations at different oscillation frequencies. The timeline display graph refers to displaying the changes of the entire knowledge graph in time by taking time as the axis and displaying the entire knowledge graph in time order, and can be used to display a single oscillation propagation path in time flow. According to the specific situation, the appropriate display method is selected, and the oscillation propagation path can be more clearly and accurately displayed, which is beneficial to quickly making oscillation suppression decisions and more effectively suppressing system oscillation and restoring stable operation of the system.

[0065] Through the above step S102, a plurality of sites are determined, and a plurality of sites respectively corresponding oscillation data are determined, wherein the oscillation data is used to represent power data generated by the corresponding site occurring telecommunication signal oscillation anomaly; step S104, the oscillation data corresponding to the plurality of sites respectively is input into the target knowledge graph representation learning model, and a target knowledge graph representing the relationship between the plurality of sites is obtained, wherein the target knowledge graph includes entities and relationships, the entities are the oscillation data corresponding to the plurality of sites respectively, and the relationships are the connection relationships between the plurality of sites; step S106, for a target oscillation anomaly included in the oscillation anomaly corresponding to the plurality of sites respectively, the oscillation propagation path of the target oscillation anomaly is determined based on the oscillation data corresponding to the plurality of sites respectively; step S108, the oscillation propagation path is displayed on the target knowledge graph, and the oscillation display result of the target oscillation anomaly is obtained. The purpose of determining the oscillation propagation path of the target oscillation anomaly is achieved, the technical effect of the oscillation display of the target oscillation anomaly is realized, and the technical problem of lacking broadband oscillation monitoring information analysis and display capability in the related art is solved.

[0066] Based on the above embodiment and optional embodiment, the present application provides an optional implementation. According to the Nyquist theorem, in order to ensure the integrity of the collected data signal, the sampling frequency fs.max must be greater than 2 times the highest frequency fmax in the signal. The maximum oscillation frequency of the broadband oscillation is selected as 2500Hz, so the sampling frequency must be higher than 5000Hz.

[0067] When the system oscillation alarm occurs, the oscillation voltage signal and the oscillation current signal of each substation interval include a fundamental component and an oscillation component, and the discrete signal x[n] corresponding to the nth time window after sampling is expressed as:

[0068]

[0069] In the above formula, n is the time window, that is, the sampling time, a is the fundamental component of the oscillation voltage signal or the oscillation current signal, a i is the amplitude of the i-th oscillation component of the oscillation voltage signal or the oscillation current signal, f is the oscillation frequency of the fundamental frequency of the oscillation voltage signal or the oscillation current signal, f i is the oscillation frequency of the i-th oscillation component of the oscillation voltage signal or the oscillation current signal, preferably f = 50 Hz, T s is the sampling interval, f s is the sampling frequency, is the initial phase of the fundamental component and the i-th oscillation component, respectively.

[0070] The oscillation data is classified and differentially processed, the fast Fourier algorithm and the interpolation compensation method are applied, and the oscillation frequency, amplitude and initial phase of the oscillation voltage and current measurement data in the 0Hz-2.5Hz low frequency oscillation, 2.5Hz-100Hz sub / super frequency oscillation and 100Hz-300Hz medium-high frequency oscillation range are extracted.

[0071] The fast Fourier algorithm is applied to convert the above discrete signal into a frequency domain signal as follows:

[0072]

[0073] Where X[n] is the frequency domain signal corresponding to the nth time window, r[n] is the window function, preferably the Hanning window, N r is the window length, 10 fundamental wave time windows are used for 0Hz-2.5Hz low frequency oscillation signals and 2.5Hz-100Hz sub / super frequency oscillation signals, and 2 fundamental wave time windows are used for 100Hz-2500Hz medium-high frequency oscillation signals, m is the frequency spectrum line number, m = 1, 2, 3,..., N r -1. The spectrum line number with the maximum amplitude near the oscillation harmonic frequency is M, and the spectrum line corresponds to the calculation result of the oscillation harmonic component at the spectrum line frequency.

[0074] Considering the possibility of spectrum leakage of the fast Fourier algorithm, the interpolation compensation method is applied to calculate the oscillation frequency, oscillation amplitude and oscillation initial phase of different oscillation types. The compensation coefficient is obtained according to the maximum amplitude of the spectrum line and the amplitudes of the adjacent two spectrum lines as

[0075] The oscillation components f of different oscillation types are determined respectively by using compensation coefficients i , the oscillation amplitude a i , and the oscillation initial phase are as follows:

[0076]

[0077]

[0078]

[0079] These data are used for analysis and calculation, and one-step standardization processing can be performed to form a unified format for different substation data. The unified format refers to a unified spatial naming and standardized data format, specifically including standardized naming of substation busbar port oscillation voltage, current, oscillation frequency, and oscillation type, expressed in structured and semi-structured data.

[0080] The knowledge graph database is used to store entities, attributes, and relationships of knowledge graph data. It should be noted that the representation learning method based on the time sequence knowledge graph is used to learn the time sequence order relationship related to the fact in a specific time sequence space combined with the fact occurrence time. The substation end oscillation data are taken as entity objects, the connection relationship of each plant and the oscillation propagation path are taken as relationships, and the TransE-TAE model is used for solving and optimization to obtain the knowledge graph database.

[0081] The time sequence knowledge graph can be represented as G t ={ε,γ,τ}. Wherein, ε and γ represent the set of entities and relationships respectively, and τ represents the set of timestamps related to the fact. Each element in Gt is represented in the form of a quadruple (e s , r, e o , t r ), where e s , e0∈ε, r∈γ represent two entities and the relationship therebetween, e s and e0 are two entities having a relationship, r is the relationship therebetween, and t r is the timestamp of the event occurrence (such as abnormality occurrence).

[0082] In order to train the target knowledge graph representation learning model, the relationship samples and triple samples are generated based on the above time sequence knowledge graph. The time sequence knowledge graph is used to generate samples, and the triple samples and relationship samples are generated. The triple samples and relationship samples are taken as the training sample set. The smaller the optimization objective function F is, the higher the accuracy of the graph structure is. The formula F is composed of two loss parts, including the feature embedding loss of the triple sample, and the feature embedding loss of the relationship sample.

[0083] Based on the TransE-TAE model representation learning method, where Δt represents the set of existing quadruple instances in a given temporal knowledge graph, any two quadruple instances (e) exist on the temporal relation chain of the same head entity. i ,r k ,e j ,t rk )∈Δt,(e i ,r l ,e m ,t rl )∈Δt, i.e., e i As a reference entity, e j For the previous entity, e m For the next entity, r k As the first relation, r l For the second relation, t rk For the timestamp of the first relation, t rl This is the timestamp of the second relation. Based on this, we can build upon the first relation r... k Second relation r l This generates a corresponding temporal sequence relation pair. <r k ,r l >

[0084] If t rk <t rl , use y + =(r k ,r l ) represents a positive sample in a temporal sequence relationship, denoted by y. - =(r k ,r l ) -1 This represents a negative sample constructed for the positive samples. The time-series scoring function is defined as: f(< r k,r l >)=||r k Tr l ||. Where T is the evolution matrix characterizing the temporal sequence relationship, through which r... k and r l Projecting onto a specific temporal vector space indicates that the smaller the distance between them, the better r k Prior to r l If it occurs, then r is likely to occur. k Lagging behind r l This occurs. The model follows a temporal chain of relationships, where the score f-value for positive samples is sufficiently small, and the score f-value for negative samples is sufficiently large. The above y + y - All of them belong to relational samples.

[0085] Based on the time sequence knowledge graph, a triple sample can also be established, and for e i is a reference entity, e j is the previous entity, and the first relationship r k between the above two entities can generate a positive example triple sample x + (e i , r k , e j ), and in this way, multiple positive example triple samples can be generated for different entities as positive example triples Δ.

[0086] For the positive example triple sample x + (e i , r k , e j ), the head and tail entities of x+triple are replaced to obtain a negative example triple sample x - =(e' i , r k , e' j )∈Δ', where Δ' represents a negative example triple. The above head entity is an entity in a group of entities that precedes in time sequence, and the tail entity is an entity in a group of entities that follows in time sequence.

[0087] Through the above processing, triple samples and relationship samples are generated as a training sample set, which can be used for subsequent training of an initial knowledge graph representation learning model to obtain a target knowledge graph representation learning model. After inputting the above training sample set into the initial knowledge graph representation learning model, optimization iteration is performed using a loss function, and the loss function is set as the following objective function F.

[0088]

[0089] where q(x) is a scoring function of the TransE model, represented as q(ε i , γ, ε k )=||ε i +γ-ε k ||, the relationship r k and rl share the same head entity e i . is a set of positive samples in the relationship sample, is a set of negative samples in the relationship sample.

[0090] In the training process, the TransE-TAE model uses a stochastic gradient descent method to solve the optimal value of the optimization objective function, and then determines the model iterated under the condition of the minimum loss function as the target knowledge graph representation learning model.

[0091] After the target knowledge graph representation learning model is generated through the above processing, the target knowledge graph representation learning model is processed on the oscillation data, and a target knowledge graph representing the relationship between multiple stations can be generated.

[0092] In order to determine the path of the oscillation anomaly in the target knowledge graph, the preprocessed oscillation data is de-duplicated and cleaned to remove invalid oscillation monitoring data and duplicate data. The similarity analysis method is used to compare the two plant station port oscillation data at both ends of the same line and the wideband oscillation data of different lines at the same plant station port to remove duplicate and incorrect monitoring data. Through the above processing, the oscillation data of the target oscillation anomaly can be determined.

[0093] The above target oscillation anomaly can be in a variety of oscillation frequency ranges, such as low-frequency oscillation frequency range, sub / super-frequency oscillation frequency range, and medium / high-frequency oscillation frequency range, etc. The port energy method or impedance method is used to calculate the oscillation energy or impedance of each substation port for classification calculation.

[0094] For the case of low-frequency oscillation of the target oscillation anomaly, the transient energy method is used to calculate the transient energy absorbed by each electrical interval port as follows:

[0095]

[0096] Where Im() is the imaginary part calculation; I ij is the current phasor injected by line j in plant station (i.e. station) i, with the line pointing to the bus direction as the positive direction; U Ai is the highest level bus voltage phasor of the plant station, with phase A as the basis for calculating transient energy; P ij , Q ij are the active power and reactive power of line j flowing to plant station i, which can be calculated according to the measured voltage and current; U i , are the amplitude and initial phase of the A-phase voltage of the highest level bus of the plant station.

[0097] The transient energy of the plant station port is calculated using the oscillation data in the low-frequency oscillation frequency range. According to the second Lyapunov stability principle, for a free dynamic system, if the total energy W (W > 0) of the system changes at a rate is always negative, the total energy of the system will continuously decrease and eventually reach a minimum value, i.e. the equilibrium state, and the system is stable. By analyzing the accumulation and dissipation trend of the transient energy of each plant station port, the energy change trend can be determined: when the transient energy of a certain plant station port is always positive, the transient energy of the plant station port continuously increases, showing an absorption characteristic, indicating that the line continuously absorbs energy at the plant station port, and the oscillation is transmitted from the line to the plant station, thereby determining that the oscillation source is upstream of the line; on the contrary, when the transient energy of a certain substation port is always negative, the transient energy of the substation port continuously decreases, showing a release characteristic, indicating that the line continuously releases energy at the substation port, and the oscillation is transmitted from the substation to the line, thereby determining that the oscillation source is downstream of the line. ​ When the constant is negative, the transient energy of the station port is continuously reduced, showing a sending characteristic, indicating that the station port continuously sends energy, and the oscillation is transmitted from the station to the line, so as to determine that the oscillation source is on the other line connected to the bus of the station. Therefore, the propagation path of the low-frequency oscillation between the stations can be determined by analyzing the sending and absorbing characteristics of the transient energy of the port.

[0098] For the case that the target oscillation anomaly is sub / super-frequency oscillation and medium / high-frequency oscillation, the equivalent impedance of the port is calculated by using the impedance analysis method as follows:

[0099]

[0100]

[0101] wherein r i , x i are the equivalent resistance and reactance of the station port respectively; I ij , are the amplitude and initial phase of the oscillation current flowing from the line j to the bus i of the station respectively; the port impedance is calculated based on the A-phase as the basic phase. According to the oscillation impedance characteristics of the system, if the system port resistance r<0 or reactance x<0, it indicates that there is an oscillation source in the system. When the equivalent resistance r i >0 or the equivalent reactance x i >0 of a certain station port, it indicates that the oscillation source exists on the upstream of the line, and the oscillation is transmitted from the line to the station; on the contrary, when the equivalent resistance r i <0 and the equivalent reactance x i <0 of a certain station port, it indicates that the oscillation source is not on the line, and the oscillation is transmitted from the other lines connected to the bus of the station to the line. Therefore, the propagation path of the sub / super-frequency oscillation and the medium / high-frequency oscillation between the stations can be determined by the resistance and inductance (capacitance) characteristics of the port impedance.

[0102] Through various ways of display, the user (dispatching and operation personnel) can more intuitively use the broadband oscillation atlas data to efficiently analyze the propagation path of the oscillation and the oscillation source, so as to quickly specify the oscillation control measures. Further, the visual interface display module supports the user interaction function, including the query, navigation and positioning functions. At the same time, the user configuration page display layout, style and function and the like are supported.

[0103] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a group of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0104] Figure 2It is a module schematic diagram of the power oscillation data processing method provided according to the embodiment of the present application, and the specific modules include a data module, a library module, a platform module and a display module.

[0105] The data module is used for accessing substation end broadband oscillation data, and pre-processing, classifying and differentiating the accessed data to form a unified data format. The data module includes a data acquisition module and a data preprocessing module. The data acquisition module is used for acquiring substation broadband oscillation data. The substation broadband oscillation data includes broadband oscillation alarm information and fault recording data. The broadband oscillation alarm information includes oscillation occurrence time and oscillation alarm type of each electrical interval. The fault recording data includes oscillation voltage and current measurement values, recording start time and end time of each electrical interval. The data preprocessing module is used for oscillation data classification and differentiation processing. The fast Fourier algorithm and the interpolation compensation method are applied to extract oscillation frequency, amplitude and initial phase of oscillation voltage and current measurement data in the range of 0Hz-2.5Hz low frequency, 2.5Hz-100Hz sub / super frequency and 100Hz-300Hz medium / high frequency. The data of different substations is formed into a unified format. The unified format refers to a unified spatial naming and standardized data format.

[0106] The library module is used for constructing a knowledge graph database based on the pre-processed data. The library module specifically applies a knowledge graph modeling tool or platform to model and abstract the pre-processed data, and establishes a knowledge graph database. The knowledge graph database is used for storing entities, attributes and relationships of knowledge graph data.

[0107] The platform module is used for providing hardware support for calculation and storage of the library module. The platform module includes a data verification module and a data calculation module. The data verification module is used for removing invalid oscillation monitoring data and repeated data by de-duplication and cleaning of the pre-processed oscillation data. The calculation module is used for classifying and calculating the verified data, and applies a port energy method or an impedance method to calculate oscillation energy or impedance of each substation port.

[0108] The display module is used for displaying knowledge graph data and providing an external interface for calling knowledge graph data. The display module includes a visual interface display module and an external data interface module. The visual interface display module is used for displaying entities, attributes and relationships of knowledge graph data, and network structure of the knowledge graph data. Further, the visual interface display module can adopt a mesh graph, a tree graph and a timeline display graph according to requirements. The visual interface display module supports user interaction functions, including query, navigation and positioning functions. The external data interface module is used for providing a programming interface to provide a data interface for external applications, and integrating broadband oscillation graph with external application systems for integrated analysis.

[0109] The power oscillation data processing system has the beneficial technical effects as follows: the broadband oscillation data from different station ends can be normalized, multi-level and multi-aspect data calculation and integration of the oscillation data can be realized, and the broadband oscillation path and the system oscillation source end can be more clearly displayed. The broadband oscillation data can be visually displayed, which provides intuitive system oscillation monitoring and analysis results for dispatching operation personnel, helps the operation personnel to quickly make oscillation suppression decisions, and more effectively suppresses system oscillation and restores system stable operation.

[0110] In the embodiment, an electric power oscillation data processing device is also provided, which is used to realize the above-mentioned embodiments and preferred embodiments, and will not be described again. As used below, the term "module" "device" can be a combination of software and / or hardware that realizes a predetermined function. Although the device described in the following embodiments is preferably realized in software, hardware or a combination of software and hardware is also possible and is contemplated.

[0111] According to the embodiment of the present application, a device embodiment for implementing the electric power oscillation data processing method is also provided, Figure 3 is a schematic diagram of an electric power oscillation data processing device according to an embodiment of the present application, as Figure 3 shown, the electric power oscillation data processing device comprises a first determination module 302, an input module 304, a first determination module 306, a result display module 308, which will be described below.

[0112] The first determination module 302 is configured to determine a plurality of sites and determine a plurality of sites respectively corresponding to oscillation data, wherein the oscillation data is used to represent the power data generated by the abnormal oscillation of the corresponding site.

[0113] The input module 304 is connected with the first determination module 302, and is configured to input the oscillation data corresponding to the plurality of sites into a target knowledge graph representation learning model respectively, to obtain a target knowledge graph representing the relationship between the plurality of sites, wherein the target knowledge graph comprises entities and relationships, the entities are the oscillation data corresponding to the plurality of sites respectively, and the relationships are the connection relationships between the plurality of sites.

[0114] The first determination module 306 is connected with the input module 304, and is configured to determine an oscillation propagation path of a target oscillation anomaly included in the oscillation anomaly corresponding to the plurality of sites respectively based on the oscillation data corresponding to the plurality of sites respectively.

[0115] The result display module 308 is connected with the first determination module 306, and is configured to display the oscillation propagation path on the target knowledge graph to obtain an oscillation display result of the target oscillation anomaly.

[0116] This invention provides a power oscillation data processing device. A first determining module identifies multiple stations and their corresponding oscillation data, where the oscillation data represents the power data generated by an abnormal electrical signal oscillation at the corresponding station. An input module inputs the oscillation data from the multiple stations into a target knowledge graph representation learning model to obtain a target knowledge graph representing the relationships between the multiple stations. The target knowledge graph includes entities and relationships; entities are the oscillation data corresponding to the multiple stations, and relationships are the connections between the multiple stations. A second determining module determines the oscillation propagation path of the target oscillation anomaly, based on the oscillation data corresponding to the multiple stations. A result display module displays the oscillation propagation path on the target knowledge graph, obtaining the oscillation display result of the target oscillation anomaly. This achieves the goal of determining the oscillation propagation path of the target oscillation anomaly and realizes the technical effect of displaying the oscillation of the target oscillation anomaly, thereby solving the technical problem of lacking the ability to analyze and display broadband oscillation monitoring information in related technologies.

[0117] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0118] It should be noted that the first determining module 302, input module 304, first determining module 306, and result display module 308 mentioned above correspond to steps S102 to S108 in the embodiments. The instances and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments. It should be noted that the above modules, as part of the device, can run in a computer terminal.

[0119] It should be noted that the optional or preferred implementation methods of this embodiment can be found in the relevant descriptions in the embodiments, and will not be repeated here.

[0120] The aforementioned power oscillation data processing device may further include a processor and a memory. The first determination module 302, the input module 304, the first determination module 306, the result display module 308, etc., are all stored in the memory as program units, and the processor executes the aforementioned program units stored in the memory to realize the corresponding functions.

[0121] The processor comprises a core, and the core retrieves corresponding program units in the memory. The core can be one or more. The memory can include a non-permanent memory in a computer readable medium, a random access memory (RAM), and / or a non-volatile memory such as a read-only memory (ROM) or a flash memory (flash RAM), and the memory comprises at least one memory chip.

[0122] The embodiment of the present application provides a non-volatile storage medium, which stores a program, and the program is executed by a processor to realize the power oscillation data processing method.

[0123] The embodiment of the present application provides an electronic device, which comprises a processor, a memory and a program stored in the memory and executable on the processor. When the processor executes the program, the following steps are implemented: determining a plurality of sites, and determining oscillation data corresponding to the plurality of sites respectively, wherein the oscillation data is used to represent power data generated by an electric signal oscillation anomaly of a corresponding site; inputting the oscillation data corresponding to the plurality of sites respectively into a target knowledge graph representation learning model to obtain a target knowledge graph representing relationships between the plurality of sites, wherein the target knowledge graph comprises entities and relationships, the entities are the oscillation data corresponding to the plurality of sites respectively, and the relationships are connection relationships between the plurality of sites; determining an oscillation propagation path of a target oscillation anomaly included in the oscillation anomalies corresponding to the plurality of sites respectively based on the oscillation data corresponding to the plurality of sites respectively; and displaying the oscillation propagation path on the target knowledge graph to obtain an oscillation display result of the target oscillation anomaly. The device in the present application can be a server, a PC or the like.

[0124] The present application also provides a computer program product, which, when executed on a data processing device, is adapted to execute a program initialized with the following method steps: determining a plurality of sites, and determining oscillation data corresponding to the plurality of sites respectively, wherein the oscillation data is used to represent power data generated by an electric signal oscillation anomaly of a corresponding site; inputting the oscillation data corresponding to the plurality of sites respectively into a target knowledge graph representation learning model to obtain a target knowledge graph representing relationships between the plurality of sites, wherein the target knowledge graph comprises entities and relationships, the entities are the oscillation data corresponding to the plurality of sites respectively, and the relationships are connection relationships between the plurality of sites; determining an oscillation propagation path of a target oscillation anomaly included in the oscillation anomalies corresponding to the plurality of sites respectively based on the oscillation data corresponding to the plurality of sites respectively; and displaying the oscillation propagation path on the target knowledge graph to obtain an oscillation display result of the target oscillation anomaly.

[0125] Those skilled in the art will appreciate that embodiments of the application can be devised for a method, a system, or a computer program product. Accordingly, the present application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer readable storage media (including, but not limited to, disk memory, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0126] The present application is described in reference to the flowchart illustrations and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams.

[0127] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams.

[0128] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams.

[0129] In one typical configuration, the computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0130] The memory can include non-persistent memory and / or volatile memory, such as random access memory (RAM) and / or cache memory. The memory can also include non-volatile memory, such as read-only memory (ROM), electrically programmable read-only memory (EPROM), electrically erasable read-only memory (EEPROM), flash memory, or a combination of non-volatile memories in different forms. The memory is an example of computer readable storage media.

[0131] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer-readable media does not include transitory media such as modulated data signals and carriers.

[0132] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusions, so that a process, method, article or apparatus that includes a list of elements does not only include those elements, but also includes other elements not explicitly listed, or other elements inherent to such a process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0133] Those skilled in the art will appreciate that embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0134] The above merely illustrates the embodiments of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application shall be included in the scope of claims of the present application.

Claims

1. A power oscillation data processing method characterized by, The method comprises the following steps: determining a plurality of sites, and determining oscillation data corresponding to the plurality of sites respectively, wherein the oscillation data is used to represent power data generated by an oscillation anomaly of a corresponding site; inputting the oscillation data corresponding to the plurality of sites respectively into a target knowledge graph representation learning model to obtain a target knowledge graph representing the relationship between the plurality of sites, wherein the target knowledge graph comprises entities and relationships, the entities are the oscillation data corresponding to the plurality of sites respectively, and the relationships are the connection relationships between the plurality of sites; determining an oscillation propagation path of a target oscillation anomaly included in the oscillation anomalies corresponding to the plurality of sites respectively based on the oscillation data corresponding to the plurality of sites respectively; displaying the oscillation propagation path on the target knowledge graph to obtain an oscillation display result of the target oscillation anomaly; wherein the determining a plurality of sites, and determining oscillation data corresponding to the plurality of sites respectively, comprises: obtaining discrete signals of the plurality of sites in the case of corresponding oscillation anomalies, and processing the discrete signals by using a fast Fourier algorithm and an interpolation compensation method to obtain the oscillation data corresponding to the plurality of sites respectively; wherein the determining an oscillation propagation path of a target oscillation anomaly based on the oscillation data corresponding to the plurality of sites respectively comprises: determining target oscillation data matched with the target oscillation anomaly and a target site corresponding to the target oscillation data in the oscillation data corresponding to the plurality of sites respectively; determining a target frequency range corresponding to the target oscillation data; in the case that the target frequency range is a predetermined first oscillation frequency range, determining a port transient energy of the target site; determining the oscillation propagation path based on the port transient energy; in the case that the target frequency range is a predetermined second oscillation frequency range, determining a port equivalent impedance of the target site, wherein the first oscillation frequency range is smaller than the second oscillation frequency range; determining the oscillation propagation path based on the port equivalent impedance; The target knowledge graph representation learning model is obtained by training an initial knowledge graph representation learning model based on a training sample set, and the training sample set is obtained based on the following manner: a preset time sequence knowledge graph is obtained, wherein the time sequence knowledge graph includes a plurality of reference entities, reference relationships between the plurality of reference entities, and timestamps; for one of the plurality of reference entities, based on the timestamps corresponding to the plurality of reference entities respectively, a previous entity before the one reference entity in time sequence is determined, and a next entity after the one reference entity in time sequence is determined; based on the reference relationships between the plurality of reference entities, a first relationship between the previous entity and the one reference entity and a second relationship between the one reference entity and the next entity are determined; based on the first relationship and the second relationship, a relationship sample corresponding to the one reference entity is generated; based on the first relationship, the previous entity, and the one reference entity, a triple sample corresponding to the one reference entity is generated; the one reference entity is used to generate the relationship sample and the triple sample to obtain the relationship sample and the triple sample corresponding to the plurality of reference entities respectively; and the training sample set is obtained based on the relationship sample and the triple sample corresponding to the plurality of reference entities respectively.

2. The method of claim 1, wherein, The determination of the oscillation data corresponding to the plurality of stations comprises: For one of the plurality of stations, an oscillation voltage signal and an oscillation current signal of the one station in a case where the corresponding oscillation anomaly occurs are obtained; Based on the oscillation voltage signal and the oscillation current signal, a discrete signal of the one station is generated; Based on the discrete signal of the one station, signal processing is performed to obtain oscillation data corresponding to the one station; The oscillation data corresponding to the plurality of stations is obtained in the same manner as the oscillation data corresponding to the one station.

3. The method of claim 2, wherein, The signal processing based on the discrete signal of the one station to obtain the oscillation data corresponding to the one station comprises: Based on the discrete signal of the one station, a fast Fourier algorithm is used for processing to obtain a frequency domain signal of the one station; Based on the frequency domain signal of the one station, an interpolation compensation method is used for processing to obtain an oscillation component, an oscillation amplitude, and an oscillation initial phase of the one station; Based on the oscillation component, the oscillation amplitude, and the oscillation initial phase of the one station, the oscillation data corresponding to the one station is determined.

4. The method of claim 1, wherein, The method further comprises: An initial knowledge graph representation learning model and a target loss function are obtained, wherein the target loss function includes entity embedding representation loss and relationship embedding representation loss; The initial knowledge graph representation learning model is trained based on the training sample set to obtain a candidate knowledge graph representation learning model, and the target loss function is updated to obtain an updated target loss function; In a case where the updated target loss function is less than a predetermined loss threshold, the candidate knowledge graph representation learning model is determined as the target knowledge graph representation learning model.

5. A power oscillation data processing device characterized by comprising: Comprises: The first determining module is configured to determine a plurality of sites and determine oscillation data corresponding to the plurality of sites respectively, wherein the oscillation data is used to represent power data generated by an oscillation anomaly of a corresponding site; The input module is configured to input the oscillation data corresponding to the plurality of sites respectively into a target knowledge graph representation learning model to obtain a target knowledge graph representing relationships between the plurality of sites, wherein the target knowledge graph includes entities and relationships, the entities are the oscillation data corresponding to the plurality of sites respectively, and the relationships are connection relationships between the plurality of sites; The second determining module is configured to determine, for a target oscillation anomaly included in the oscillation anomalies corresponding to the plurality of sites respectively, an oscillation propagation path of the target oscillation anomaly based on the oscillation data corresponding to the plurality of sites respectively; The result display module is configured to display the oscillation propagation path on the target knowledge graph to obtain an oscillation display result of the target oscillation anomaly; The first determining module is further configured to obtain discrete signals of the plurality of sites in a case where the corresponding oscillation anomalies occur, and process the discrete signals by using a fast Fourier algorithm and an interpolation compensation method to obtain the oscillation data corresponding to the plurality of sites respectively; The second determining module is further configured to determine, in the oscillation data corresponding to the plurality of sites respectively, target oscillation data matched with the target oscillation anomaly, and determine a target site corresponding to the target oscillation data; determine a target frequency range corresponding to the target oscillation data; in a case where the target frequency range is a predetermined first oscillation frequency range, determine a port transient energy of the target site; determine the oscillation propagation path based on the port transient energy; in a case where the target frequency range is a predetermined second oscillation frequency range, determine a port equivalent impedance of the target site, wherein the first oscillation frequency range is smaller than the second oscillation frequency range; and determine the oscillation propagation path based on the port equivalent impedance. The target knowledge graph representation learning model is obtained by training an initial knowledge graph representation learning model based on a training sample set. The training sample set is obtained based on the following manner: a preset time sequence knowledge graph is obtained, wherein the time sequence knowledge graph includes a plurality of reference entities, reference relationships and timestamps between the plurality of reference entities; for one reference entity in the plurality of reference entities, based on the timestamps corresponding to the plurality of reference entities respectively, a previous entity before the one reference entity in time sequence is determined, and a next entity after the one reference entity in time sequence is determined; based on the reference relationships between the plurality of reference entities, a first relationship between the previous entity and the one reference entity and a second relationship between the one reference entity and the next entity are determined; based on the first relationship and the second relationship, a relationship sample corresponding to the one reference entity is generated; based on the first relationship, the previous entity and the one reference entity, a triple sample corresponding to the one reference entity is generated; the one reference entity is used to generate the relationship sample and the triple sample to obtain the relationship sample and the triple sample corresponding to the plurality of reference entities respectively; and the training sample set is obtained based on the relationship sample and the triple sample corresponding to the plurality of reference entities respectively.

6. A non-volatile storage medium, comprising: The non-volatile storage medium stores a plurality of instructions, the instructions being adapted to be loaded and executed by the processor to implement the power oscillation data processing method of any one of claims 1 to 4.

7. An electronic device, comprising: Comprise: One or more processors and a memory, the memory being configured to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the power oscillation data processing method of any one of claims 1 to 4.

Images