Transformer substation secondary equipment modeling method and equipment based on knowledge graph, and medium

Abstract

The invention discloses a transformer substation secondary equipment modeling method and equipment based on a knowledge graph and a medium, and the method comprises the steps: carrying out the first classification of a basic database of transformer substation secondary equipment for the types of the secondary equipment, obtaining a first classification database, and carrying out the data collection of each type of secondary equipment in the first classification database, performing processing according to the feature information of the secondary equipment to obtain a feature matrix and an equipment type matrix of all the secondary equipment of the transformer substation, and for the feature matrix and the equipment type matrix of all the secondary equipment of the transformer substation, calculating the correlation degree of all the secondary equipment of the transformer substation through the trained correlation degree calculation network to obtain the correlation degree of all the secondary equipment of the transformer substation. According to the substation secondary equipment modeling method, the secondary equipment is classified firstly, the features are extracted from the classified data set, the correlation degree is calculated for each feature, and the modeling accuracy is improved while the working efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical secondary equipment management, and in particular to a knowledge graph-based substation secondary equipment modeling method, equipment and medium. Background Art

[0002] Power secondary equipment is an auxiliary equipment for monitoring, measuring, controlling, regulating and protecting the primary equipment of the power system, and is not directly connected to the main circuit of electric energy. With the development of knowledge graph technology, more and more people are paying attention to the fact that knowledge graph can be applied to the power system. CN116187175A discloses a neural network identification method for substations based on the knowledge graph of the electrical industry, which makes the graphic elements and data models correspond one to one, introduces the power drawing standard, and constructs a knowledge graph for the design engineering field of substations for complex users. On the basis of the knowledge graph, an intelligent identification method for electrical system diagram parameters based on a recursive neural network learning algorithm and a method for generating a distribution station equipment plan layout diagram based on graph reasoning are used. Through the definition of relevant rules by graph reasoning, various types of power design drawings are automatically generated, realizing the automatic understanding of electrical system drawings and architectural plan drawings, and realizing the automatic generation of substation design schemes based on the automatic acquisition of design parameters. However, in actual scenarios, the cost of a complete knowledge graph model is relatively high. Due to the complex types of secondary equipment, there is currently no standardized data division in the construction of the knowledge graph. Direct processing operations have low work efficiency and cannot be performed on a single feature attribute, reducing the accuracy of modeling. Summary of the invention

[0003] Technical purpose: In view of the defects in the prior art, the present invention discloses a substation secondary equipment modeling method, equipment and medium based on knowledge graph. The secondary equipment is first classified, and features are extracted from the classified data set. The correlation degree is calculated for each feature to obtain entities, attributes and relationships. Then, a knowledge graph is constructed according to the entities, attributes and relationships, thereby improving work efficiency and modeling accuracy.

[0004] Technical solution: In order to achieve the above technical objectives, the present invention adopts the following technical solution.

[0005] A method for modeling secondary equipment of a substation based on a knowledge graph comprises the following steps:

[0006] Step S1, performing a first classification on the basic database of substation secondary equipment according to the types of secondary equipment to obtain a first classification database;

[0007] Step S2: Process the data set of each type of secondary equipment in the first classification database according to the characteristic information of the secondary equipment to obtain a characteristic matrix and an equipment type matrix of all secondary equipment in the substation;

[0008] Step S3, for the feature matrix and device type matrix of all secondary devices in the substation, the correlation degree of all secondary devices in the substation is calculated by the trained correlation degree calculation network;

[0009] Step S4: construct a knowledge graph based on the entities, attributes and relationships obtained in step S3 to complete the substation secondary equipment modeling.

[0010] Preferably, in step S1, the secondary devices are divided into T types according to the device type, and a first classification database DATA is constructed, wherein DATA={data 1 , data 2 , ..., data t , ..., data T}, 1≤t≤T, data t represents the data set of the t-th type of secondary equipment, 1≤n≤N, N is the total number of secondary devices of type t; It is the nth secondary equipment set in the tth type of secondary equipment.

[0011] Preferably, in step S2, The feature information of the secondary device is included, and the feature information includes ID feature attribute, static feature attribute, dynamic feature attribute, warning feature attribute and control feature attribute; the ID feature attribute is the feature used by the secondary device to characterize its own identity, the static feature attribute is the feature that does not change with time in the secondary device, the dynamic feature attribute is the feature that changes with time in the secondary device, the warning feature attribute is whether the secondary device can feedback warning information, and the control feature attribute is whether the secondary device can control the signal or loop;

[0012] Encode the characteristic information of all secondary equipment in the substation to obtain the characteristic matrix FM of the secondary equipment e =[fm e0 ,fm e1 ,…,fm er ,…,fm eR ], where 1≤r≤R, R is the number of feature information, fm e0 is the type of the e-th secondary device, fm er For the encoding of the rth feature information of the eth secondary device, a device type matrix TYPE=[fm 10 ,fm 20 , ..., fm e0 , ..., fm E0 ], the characteristic information matrix of the e-th secondary device is FM′ e =[fm e1 ,…,fm er ,…,fmeR ]; where 1≤c≤E, E is the total number of all secondary equipment in the substation.

[0013] Preferably, the association calculation network in step S3 includes a first network and a second network, the first network includes an ID feature attribute association network, which is used to output entities in the knowledge graph; the second network includes a static feature attribute association network, a dynamic feature attribute association network, a warning feature attribute association network and a control feature attribute association network, which are used to output attributes and relationships in the knowledge graph.

[0014] Preferably, the ID feature attribute association network includes:

[0015] Input unit, input and output are device type matrix TYPE, TYPE = [fm 10 ,fm 20 , ..., fm e0 , ..., fn E0 ]; dimension is 1,

[0016] Hidden unit one connected to the input unit and hidden unit three, the input of hidden unit one is the device type matrix TYPE, the ID feature attribute in the feature information and the linear conversion output OUT′3 of hidden unit three; the weight between hidden unit one and the input unit is 1≤i≤I, I is the dimension of hidden unit one, and the activation function of hidden unit one is f 1 (x); the output of hidden unit 1 is

[0017] Hidden unit 2 is connected to hidden unit 1 and input unit. The input of hidden unit 2 is device type matrix TYPE and output OUT1 of hidden unit 1. The weight between hidden unit 2 and hidden unit 1 and input unit is The dimension of hidden unit 2 is E, and the activation function of hidden unit 2 is f 2 (x), the output of hidden unit 2 is

[0018] Hidden unit three is connected to hidden unit two. The input of hidden unit three is the output OUT2 of hidden unit two. The weight between hidden unit three and hidden unit two is The activation function of hidden unit three is f 3 (x), the dimension of hidden unit three is I, and the output of hidden unit three is The output threshold interval of the preset hidden unit three [δ min , δ max ],δ min , δ maxare the minimum output threshold and the maximum output threshold respectively; if the output OUT3 of the hidden unit three is within the output threshold interval, the output OUT3 of the hidden unit three is input to the output unit; if the output OUT3 of the hidden unit three is not within the output threshold interval, the output OUT3 of the hidden unit three is linearly transformed to obtain the linear transformation output OUT′3 of the hidden unit three. The linear transformation output OUT′3 of hidden unit three is returned to hidden unit one for processing;

[0019] The output unit, the input of the output unit is the output OUT3 of the hidden unit three, and the weight between the output unit and the hidden unit three is The dimension of the output unit is E, and the output is the correlation between the e-th secondary device and other secondary devices. The association between each secondary device based on the ID feature attribute is obtained as the entity in the knowledge graph.

[0020] Preferably, the warning feature attribute association network includes:

[0021] The first input unit, the input and output are both the characteristic information matrix of the e-th secondary device, that is, FM′ e =[fm e1 ,…,fm er ,…,fm eR ], 1≤r≤R, R is the number of feature information, and the dimension of the first input unit is 1;

[0022] The first hidden unit connected to the first input unit, the input of the first hidden unit is the characteristic information matrix of the e-th secondary device, and the weight between the first hidden unit and the first input unit is The dimension of the first hidden unit is 1, and the activation function of the first hidden unit is g 1 (x); the output of the first hidden unit is λ1 is the internal parameter of the first hidden unit, and 0<λ1<1. By adjusting the value of the internal parameter λ1 of the first hidden unit, the output of the first hidden unit is within the first warning feature attribute threshold interval [0, 1], and the output of the first hidden unit is input to the first output unit;

[0023] The first output unit connected to the first hidden unit, the input of the first output unit is the output of the first hidden unit, and the output of the first output unit is Wherein, θ is the preset second warning characteristic attribute threshold; the output OUT of the first output unit 2 It is the warning characteristic attribute of the secondary equipment, and the output result is 0 or 1;

[0024] The control feature attribute association network includes:

[0025] The second input unit, the input and output are both the characteristic information matrix of the e-th secondary device, that is, FM′ e =[fm e1 ,…,fm er ,…,fm eR ], 1≤r≤R, R is the number of feature information, and the dimension of the second input unit is 1;

[0026] The second hidden unit connected to the second input unit, the input of the second hidden unit is the characteristic information matrix of the e-th secondary device, and the weight between the second hidden unit and the second input unit is The dimension of the second hidden unit is 1, and the activation function of the second hidden unit is g 2 (x); the output of the second hidden unit is λ2 is the internal parameter of the second hidden unit, and 0<λ2<1. By adjusting the value of the internal parameter λ2 of the second hidden unit, the output of the second hidden unit is within the first control feature attribute threshold interval [0, 1], and the output of the second hidden unit is input to the second output unit;

[0027] The second output unit connected to the second hidden unit has the second hidden unit as its input; the output of the second output unit is Wherein, n is the preset second control characteristic attribute threshold; the output OUT of the second output unit 4 It is the control characteristic attribute of the secondary device, and the output result is 0 or 1;

[0028] The static feature attribute association network includes:

[0029] The third input unit connected to the first output unit and the second output unit has the characteristic information matrix of the e-th secondary device and the output OUT of the first output unit as input and output. 2 , the output OUT of the second output unit 4 , also written as IN 3 , the dimension of the second input unit is R+2; where IN 3 It is expressed as:

[0030]

[0031] in, is the jth value in the output of the third input unit;

[0032] The third hidden unit connected to the third input unit, the input of the third hidden unit is the output of the third input unit and the output OUT4 of the knowledge graph model association calculation network; the weight between the third hidden unit and the third input unit is 1≤j≤R+2, R is the number of feature information, 1≤e≤E, E is the total number of all secondary devices in the substation; the activation function of the third hidden unit is g 3 (x); the output of the third hidden unit is

[0033] A fourth hidden unit connected to the third hidden unit, the input of the fourth hidden unit is the output OUT of the third hidden unit 5 ; The weight between the fourth hidden unit and the third hidden unit is 1≤h≤H, H is the dimension of the fourth hidden unit; the activation function of the fourth hidden unit is g 4 (x); the output of the fourth hidden unit is

[0034] The third output unit connected to the fourth hidden unit, the input of the third output unit is the output OUT of the fourth hidden unit 6 , the weight between the fourth hidden unit and the third output unit is The activation function of the third output unit is g 5 (x); the dimension S of the third output unit is also the number of static feature attributes; the output of the third output unit is It is also the static characteristic property of the secondary equipment;

[0035] The dynamic feature attribute association network includes:

[0036] The fourth input unit connected to the first output unit, the second output unit, and the third output unit has the characteristic information matrix of the e-th secondary device and the output OUT of the first output unit as input and output. 2 , the output OUT of the second output unit 4 , the output OUT of the third output unit 7 , also written as IN 4 , the dimension of the fourth input unit is S+R+2; where IN 4 It is expressed as:

[0037]

[0038] in, is the vth value in the output of the third input unit; 1≤v≤S+R+2, R is the number of feature information, S is the number of static feature attributes;

[0039] The fifth hidden unit connected to the fourth input unit, the input of the fifth hidden unit is the output of the fourth input unit and the output OUT4 of the knowledge graph model association calculation network; the dimension of the fifth hidden unit is E; the weight between the fifth hidden unit and the fourth input unit is 1≤e≤E, E is the total number of all secondary devices in the substation; the activation function of the fifth hidden unit is g 6 (x); the output of the fifth hidden unit is

[0040] The sixth hidden unit connected to the fifth hidden unit, the input of the sixth hidden unit is the output OUT of the fifth hidden unit 8 ; The weight between the sixth hidden unit and the fifth hidden unit is 1≤p≤P, P is the dimension of the sixth hidden unit; the activation function of the sixth hidden unit is g 7 (x); the output of the sixth hidden unit is

[0041] A fourth output unit connected to the sixth hidden unit, the input of the fourth output unit being the output OUT of the sixth hidden unit 9 , the weight between the sixth hidden unit and the fourth output unit is The activation function of the fourth output unit is g 8 (x); the dimension D of the fourth output unit is also the number of dynamic feature attributes; the output of the fourth output unit is It is also the static characteristic property of the secondary equipment;

[0042] The output of the second network is the correlation between the secondary devices of the substation, relate=[OUT 2 ,OUT 4 ,OUT 7 ,OUT 10 ]; used as attributes and relations in knowledge graphs.

[0043] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, any of the above-mentioned methods for modeling substation secondary equipment based on a knowledge graph is implemented.

[0044] A computer-readable storage medium stores computer-executable instructions, wherein the computer-executable instructions are used to execute any of the above-mentioned methods for modeling substation secondary equipment based on knowledge graphs.

[0045] Beneficial effect: The present invention first classifies the knowledge graph model database to facilitate rapid identification of substation secondary equipment for knowledge graph query, and then calculates the correlation of all substation secondary equipment, and then calculates the correlation between the substation secondary equipment and the knowledge graph model, and finally, according to the calculated overlap rate, determines the knowledge graph corresponding to the highest overlap rate as the knowledge graph corresponding to the substation secondary equipment, selects the knowledge graph with the highest match from the knowledge graph model database, completes the substation secondary equipment modeling, realizes the reuse of knowledge graphs in the knowledge graph model database, greatly improves work efficiency, and reduces unnecessary work costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is a flow chart of the method of the present invention;

[0047] Figure 2 is a schematic diagram of a first network structure of the present invention;

[0048] Figure 3 This is a schematic diagram of the second network structure of the present invention. DETAILED DESCRIPTION

[0049] The present invention is further described and explained below in conjunction with the accompanying drawings.

[0050] As attached Figure 1 As shown, a substation secondary equipment modeling method based on knowledge graph includes the following steps:

[0051] Step S1, performing a first classification on the basic database of substation secondary equipment according to the types of secondary equipment to obtain a first classification database;

[0052] The secondary equipment is divided into T types according to the equipment type, and a first classification database INTA is constructed. In some embodiments of the present invention, the secondary equipment can be divided into seven categories, namely T=7, including: electrical parameter measurement type, insulation monitoring type, signal control type, relay protection type, power supply type, high-frequency wave blocking type, and other comprehensive automation equipment types; the secondary equipment of the electrical parameter measurement type is mainly used to monitor and measure the electrical parameters such as current, voltage, power, etc. of the circuit, such as various measuring meters, etc. The secondary equipment of the insulation monitoring type is mainly used to monitor the insulation condition of AC and DC power grids; the secondary equipment of the signal control type is mainly used to realize the closing and tripping of the circuit breaker in the distribution device according to the operating circuit in a manual or automatic manner; the secondary equipment of the relay protection type is mainly used to realize the closing and tripping of the circuit breaker in the distribution device according to the operating circuit; the secondary equipment of the electrical parameter measurement ... The secondary equipment of the primary equipment is mainly used to monitor the operating status of the primary equipment, quickly respond to abnormalities and accidents, and perform protection and control through circuit breakers, such as relays, automatic devices, etc.; the secondary equipment of the power supply type is mainly used to provide other secondary equipment with connected and stable quality DC power, such as battery packs, DC generators, silicon rectifiers, emergency lighting power, backup power supplies, etc.; the secondary equipment of the high-frequency wave blocking type is mainly used to prevent high-frequency current from leaking to the substation or branch line, reduce high-frequency energy loss, and form a high-frequency communication channel for power lines with coupling capacitors, combined filters, high-frequency cables, high-frequency communication machines, etc.; the secondary equipment of other comprehensive automation equipment is automation equipment other than the above six types of secondary equipment, such as standby automatic transfer devices, etc.

[0053] Where DATA = {data 1 , data 2 , ..., data t , ..., data T}, 1≤t≤T, data t represents the data set of the t-th type of secondary equipment, 1≤n≤N, N is the total number of secondary devices of type t; is the nth secondary equipment set in the tth type of secondary equipment;

[0054] Step S2: Process the data set of each type of secondary equipment in the first classification database according to the characteristic information of the secondary equipment to obtain a characteristic matrix and an equipment type matrix of all secondary equipment in the substation;

[0055] The feature information of the secondary equipment is included in it, and the feature information includes ID feature attributes, static feature attributes, dynamic feature attributes, warning feature attributes and control feature attributes; among them, the ID feature attribute is the feature used by the secondary equipment to characterize its own identity, including name, model, manufacturer, etc.; the static feature attribute is the feature that does not change with time in the secondary equipment, and the number of static feature attributes is S, such as the voltage value of the backup power supply, etc.; the dynamic feature attribute is the feature that changes with time in the secondary equipment, and the number of dynamic feature attributes is D, such as the real-time voltage value and current value of the circuit, etc.; the warning feature attribute is whether the secondary equipment can feedback warning information, if it can, the warning feature attribute value is 1, if not, the warning feature attribute value is 0; the warning information includes fault reporting, insulation status reporting, etc.; the control feature attribute is whether the secondary equipment can control the signal or circuit, for example, whether it can control the closing and tripping of the circuit breaker, whether to turn on the backup power supply, etc., then the control feature attribute value is 1, if not, then the control feature attribute value is 0;

[0056] Encode the characteristic information of all secondary equipment in the substation to obtain the characteristic matrix FM of the secondary equipment e =[fm e0 ,fm e1 ,…,fm er ,…,fm eR ], where 1≤r≤R, R is the number of feature information, fm e0 is the type of the e-th secondary device, fm er For the encoding of the rth feature information of the eth secondary device, a device type matrix TYPE=[fm 10 ,fm 20 , ..., fm e0 , ..., fm E0 ], the characteristic information matrix of the e-th secondary device is FM′ e =[fm e1 ,…,fm er ,…,fm eR ]; where 1≤e≤E, E is the total number of all secondary equipment in the substation;

[0057] Step S3, for the feature matrix and device type matrix of all secondary devices in the substation, the correlation degree of all secondary devices in the substation is calculated by the trained correlation degree calculation network;

[0058] The association calculation network includes a first network and a second network, the first network includes an ID feature attribute association network, which is used to output entities in the knowledge graph; the second network includes a static feature attribute association network, a dynamic feature attribute association network, a warning feature attribute association network and a control feature attribute association network, which are used to output attributes and relationships in the knowledge graph;

[0059] The ID feature attribute association network includes:

[0060] Input unit, input and output are device type matrix TYPE, TYPE = [fm 10 ,fm 20 , ..., fm e0 , ..., fn E0 ]; dimension is 1,

[0061] Hidden unit 1 connected to input unit and hidden unit 3, the input of hidden unit 1 is device type matrix TYPE, ID feature attribute in feature information and linear conversion output OUT3 of hidden unit 3; the weight between hidden unit 1 and input unit is 1≤i≤I, I is the dimension of hidden unit one, and the activation function of hidden unit one is f 1 (x); the output of hidden unit 1 is

[0062] Hidden unit 2 is connected to hidden unit 1 and input unit. The input of hidden unit 2 is device type matrix TYPE and output OUT1 of hidden unit 1. The weight between hidden unit 2 and hidden unit 1 and input unit is The dimension of hidden unit 2 is E, and the activation function of hidden unit 2 is f 2 (x), the output of hidden unit 2 is

[0063] Hidden unit three is connected to hidden unit two. The input of hidden unit three is the output OUT2 of hidden unit two. The weight between hidden unit three and hidden unit two is The activation function of hidden unit three is f 3 (x), the dimension of hidden unit three is I, and the output of hidden unit three is The output threshold interval of the preset hidden unit three [δ min , δ max ],δ min , δ max are the minimum output threshold and the maximum output threshold respectively; if the output OUT3 of the hidden unit three is within the output threshold interval, the output OUT3 of the hidden unit three is input to the output unit; if the output OUT3 of the hidden unit three is not within the output threshold interval, the output OUT3 of the hidden unit three is linearly transformed to obtain the linear transformation output OUT′3 of the hidden unit three. The linear transformation output OUT′3 of hidden unit three is returned to hidden unit one for processing;

[0064] The output unit, the input of the output unit is the output OUT3 of the hidden unit three, and the weight between the output unit and the hidden unit three is The dimension of the output unit is E, and the output is the correlation between the e-th secondary device and other secondary devices. Obtain the association between each secondary device based on the ID feature attribute as an entity in the knowledge graph;

[0065] The warning feature attribute correlation network includes:

[0066] The first input unit, the input and output are both the characteristic information matrix of the e-th secondary device, that is, FM′ e =[fm e1 ,…,fm er ,…,fm eR ], 1≤r≤R, R is the number of feature information, and the dimension of the first input unit is 1;

[0067] The first hidden unit connected to the first input unit, the input of the first hidden unit is the characteristic information matrix of the e-th secondary device, and the weight between the first hidden unit and the first input unit is The dimension of the first hidden unit is 1, and the activation function of the first hidden unit is g 1 (x); the output of the first hidden unit is λ1 is the internal parameter of the first hidden unit, and 0<λ1<1. By adjusting the value of the internal parameter λ1 of the first hidden unit, the output of the first hidden unit is within the first warning feature attribute threshold interval [0, 1], and the output of the first hidden unit is input to the first output unit;

[0068] The first output unit connected to the first hidden unit, the input of the first output unit is the output of the first hidden unit, and the output of the first output unit is Wherein, θ is the preset second warning characteristic attribute threshold; the output OUT of the first output unit 2 It is the warning characteristic attribute of the secondary equipment, and the output result is 0 or 1;

[0069] The control feature attribute association network includes:

[0070] The second input unit, the input and output are both the characteristic information matrix of the e-th secondary device, that is, FM′ e =[fm e1 ,…,fm er ,…,fm eR ], 1≤r≤R, R is the number of feature information, and the dimension of the second input unit is 1;

[0071] The second hidden unit connected to the second input unit, the input of the second hidden unit is the characteristic information matrix of the e-th secondary device, and the weight between the second hidden unit and the second input unit is The dimension of the second hidden unit is 1, and the activation function of the second hidden unit is g 2 (x); the output of the second hidden unit is λ2 is the internal parameter of the second hidden unit, and 0<λ2<1. By adjusting the value of the internal parameter λ2 of the second hidden unit, the output of the second hidden unit is within the first control feature attribute threshold interval [0, 1], and the output of the second hidden unit is input to the second output unit;

[0072] The second output unit connected to the second hidden unit has the second hidden unit as its input; the output of the second output unit is Wherein, n is the preset second control characteristic attribute threshold; the output OUT of the second output unit 4 It is the control characteristic attribute of the secondary device, and the output result is 0 or 1;

[0073] The static feature attribute association network includes:

[0074] The third input unit connected to the first output unit and the second output unit has the characteristic information matrix of the e-th secondary device and the output OUT of the first output unit as input and output. 2 , the output OUT of the second output unit 4 , also written as IN 3 , the dimension of the second input unit is R+2; where IN 3 It is expressed as:

[0075]

[0076] in, is the jth value in the output of the third input unit;

[0077] The third hidden unit connected to the third input unit, the input of the third hidden unit is the output of the third input unit and the output OUT4 of the knowledge graph model association calculation network; the weight between the third hidden unit and the third input unit is 1≤j≤R+2, R is the number of feature information, 1≤e≤E, E is the total number of all secondary devices in the substation; the activation function of the third hidden unit is g 3 (x); the output of the third hidden unit is

[0078] A fourth hidden unit connected to the third hidden unit, the input of the fourth hidden unit is the output OUT of the third hidden unit 5; The weight between the fourth hidden unit and the third hidden unit is 1≤h≤H, H is the dimension of the fourth hidden unit; the activation function of the fourth hidden unit is g 4 (x); the output of the fourth hidden unit is

[0079] The third output unit connected to the fourth hidden unit, the input of the third output unit is the output OUT of the fourth hidden unit 6 , the weight between the fourth hidden unit and the third output unit is The activation function of the third output unit is g 5 (x); the dimension S of the third output unit is also the number of static feature attributes; the output of the third output unit is It is also the static characteristic property of the secondary equipment;

[0080] The dynamic feature attribute association network includes:

[0081] The fourth input unit connected to the first output unit, the second output unit, and the third output unit has the characteristic information matrix of the e-th secondary device and the output OUT of the first output unit as input and output. 2 , the output OUT of the second output unit 4 , the output OUT of the third output unit 7 , also written as IN 4 , the dimension of the fourth input unit is S+R+2; where IN 4 It is expressed as:

[0082]

[0083] in, is the vth value in the output of the third input unit; 1≤v≤S+R+2, R is the number of feature information, S is the number of static feature attributes;

[0084] The fifth hidden unit connected to the fourth input unit, the input of the fifth hidden unit is the output of the fourth input unit and the output OUT4 of the knowledge graph model association calculation network; the dimension of the fifth hidden unit is E; the weight between the fifth hidden unit and the fourth input unit is 1≤e≤E, E is the total number of all secondary devices in the substation; the activation function of the fifth hidden unit is g 6 (x); the output of the fifth hidden unit is

[0085] The sixth hidden unit connected to the fifth hidden unit, the input of the sixth hidden unit is the output OUT of the fifth hidden unit 8 ; The weight between the sixth hidden unit and the fifth hidden unit is 1≤p≤P, P is the dimension of the sixth hidden unit; the activation function of the sixth hidden unit is g 7 (x); the output of the sixth hidden unit is

[0086] A fourth output unit connected to the sixth hidden unit, the input of the fourth output unit being the output OUT of the sixth hidden unit 9 , the weight between the sixth hidden unit and the fourth output unit is The activation function of the fourth output unit is g 8 (x); the dimension D of the fourth output unit is also the number of dynamic feature attributes; the output of the fourth output unit is It is also the static characteristic property of the secondary equipment;

[0087] The output of the second network is the correlation between the secondary devices of the substation, relate=[OUT 2 ,OUT 4 ,OUT 7 ,OUT 10 ]; used as attributes and relations in knowledge graphs;

[0088] The first network and the second network are trained; samples are constructed for the first network and the second network and trained to obtain a trained correlation calculation network. Network training is a conventional technical means in this field and will not be described in detail here.

[0089] Step S4: construct a knowledge graph based on the entities, attributes and relationships obtained in step S3 to complete the substation secondary equipment modeling.

[0090] The present invention first classifies the secondary equipment, extracts features from the classified data set, calculates the correlation for each feature, obtains entities, attributes and relationships, and then constructs a knowledge graph based on the entities, attributes and relationships, thereby improving work efficiency and modeling accuracy.

[0091] The present invention also discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, any of the above-mentioned methods for modeling substation secondary equipment based on knowledge graph is implemented. The memory may be various types of memory, such as random access memory, read-only memory, flash memory, etc. The processor may be various types of processors, such as a central processing unit, a microprocessor, a digital signal processor, or an image processor, etc.

[0092] A computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to execute any of the above-mentioned substation secondary equipment modeling methods based on knowledge graphs. The storage medium includes: U disk, mobile hard disk, ROM, RAM, disk or CD-ROM and other media that can store program codes.

[0093] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A knowledge graph-based substation secondary equipment modeling method, characterized in that: The following steps are involved: Step S1, performing a first classification on the basic database of substation secondary equipment according to the types of secondary equipment to obtain a first classification database; Step S2: Process the data set of each type of secondary equipment in the first classification database according to the characteristic information of the secondary equipment to obtain a characteristic matrix and an equipment type matrix of all secondary equipment in the substation; Step S3, for the feature matrix and device type matrix of all secondary devices in the substation, the correlation degree of all secondary devices in the substation is calculated by the trained correlation degree calculation network; Step S4: construct a knowledge graph based on the entities, attributes and relationships obtained in step S3 to complete the substation secondary equipment modeling.

2. According to claim 1, a knowledge graph-based substation secondary equipment modeling method is characterized by: In step S1, the secondary devices are divided into T types according to the device type, and a first classification database DATA is constructed, where DATA={data 1 , data 2 , ..., data t , ..., data T }, 1≤t≤T, data t represents the data set of the t-th type of secondary equipment, 1≤n≤N, N is the total number of secondary devices of type t; It is the nth secondary equipment set in the tth type of secondary equipment.

3. According to a knowledge graph-based substation secondary equipment modeling method according to claim 2, it is characterized by: In step S2, The feature information of the secondary device is included, and the feature information includes ID feature attribute, static feature attribute, dynamic feature attribute, warning feature attribute and control feature attribute; the ID feature attribute is the feature used by the secondary device to characterize its own identity, the static feature attribute is the feature that does not change with time in the secondary device, the dynamic feature attribute is the feature that changes with time in the secondary device, the warning feature attribute is whether the secondary device can feedback warning information, and the control feature attribute is whether the secondary device can control the signal or loop; Encode the characteristic information of all secondary equipment in the substation to obtain the characteristic matrix FM of the secondary equipment e =[fm e0 ,fm e1 ,…,fm er ,…,fm eR ], where 1≤r≤R, R is the number of feature information, fm e0 is the type of the e-th secondary device, fm er For the encoding of the rth feature information of the eth secondary device, a device type matrix TYPE=[fm 10 ,fm 20 , ..., fm e0 , ..., fm E0 ], the characteristic information matrix of the e-th secondary device is FM′ e =[fm e1 ,…,fm er ,…,fm eR ]; where 1≤e≤E, E is the total number of all secondary equipment in the substation.

4. According to claim 3, a knowledge graph-based substation secondary equipment modeling method is characterized by: In step S3, the association calculation network includes a first network and a second network. The first network includes an ID feature attribute association network, which is used to output entities in the knowledge graph; the second network includes a static feature attribute association network, a dynamic feature attribute association network, a warning feature attribute association network and a control feature attribute association network, which are used to output attributes and relationships in the knowledge graph.

5. According to claim 4, a knowledge graph-based substation secondary equipment modeling method is characterized by: ID feature attribute association network includes: Input unit, input and output are device type matrix TYPE, TYPE = [fm 10 ,fm 20 , ..., fm e0 , ..., fm E0 ]; dimension is 1, Hidden unit 1 connected to input unit and hidden unit 3, the input of hidden unit 1 is device type matrix TYPE, ID feature attribute in feature information and linear conversion output OUT3 of hidden unit 3; the weight between hidden unit 1 and input unit is I is the dimension of hidden unit one, and the activation function of hidden unit one is f 1 (x); the output of hidden unit 1 is Hidden unit 2 is connected to hidden unit 1 and input unit. The input of hidden unit 2 is device type matrix TYPE and output OUT1 of hidden unit 1. The weight between hidden unit 2 and hidden unit 1 and input unit is The dimension of hidden unit 2 is E, and the activation function of hidden unit 2 is f 2 (x), the output of hidden unit 2 is Hidden unit three is connected to hidden unit two. The input of hidden unit three is the output OUT2 of hidden unit two. The weight between hidden unit three and hidden unit two is The activation function of hidden unit three is f 3 (x), the dimension of hidden unit three is I, and the output of hidden unit three is The output threshold interval of the preset hidden unit three [δ min , δ max ],δ min , δ max are the minimum output threshold and the maximum output threshold respectively; if the output OUT3 of the hidden unit three is within the output threshold interval, the output OUT3 of the hidden unit three is input to the output unit; if the output OUT3 of the hidden unit three is not within the output threshold interval, the output OUT3 of the hidden unit three is linearly transformed to obtain the linear transformation output OUT′3 of the hidden unit three. The linear transformation output OUT′3 of hidden unit three is returned to hidden unit one for processing; The output unit, the input of the output unit is the output OUT3 of the hidden unit three, and the weight between the output unit and the hidden unit three is The dimension of the output unit is E, and the output is the correlation between the e-th secondary device and other secondary devices. The association between each secondary device based on the ID feature attribute is obtained as the entity in the knowledge graph.

6. According to claim 4, a knowledge graph-based substation secondary equipment modeling method is characterized by: The warning feature attribute correlation network includes: The first input unit, the input and output are both the characteristic information matrix of the e-th secondary device, that is, FM′ e =[fm e1 ,…,fm er ,…,fm eR ], 1≤r≤R, R is the number of feature information, and the dimension of the first input unit is 1; The first hidden unit connected to the first input unit, the input of the first hidden unit is the characteristic information matrix of the e-th secondary device, and the weight between the first hidden unit and the first input unit is The dimension of the first hidden unit is 1, and the activation function of the first hidden unit is g 1 (x); the output of the first hidden unit is λ1 is the internal parameter of the first hidden unit, and 0<λ1<1. By adjusting the value of the internal parameter λ1 of the first hidden unit, the output of the first hidden unit is within the first warning feature attribute threshold interval [0, 1], and the output of the first hidden unit is input to the first output unit; The first output unit connected to the first hidden unit, the input of the first output unit is the output of the first hidden unit, and the output of the first output unit is Wherein, θ is the preset second warning characteristic attribute threshold; the output OUT of the first output unit 2 It is the warning characteristic attribute of the secondary equipment, and the output result is 0 or 1; The control feature attribute association network includes: The second input unit, the input and output are both the characteristic information matrix of the e-th secondary device, that is, FM′ e =[fm e1 ,…,fm er ,…,fm eR ], 1≤r≤R, R is the number of feature information, and the dimension of the second input unit is 1; The second hidden unit connected to the second input unit, the input of the second hidden unit is the characteristic information matrix of the e-th secondary device, and the weight between the second hidden unit and the second input unit is The dimension of the second hidden unit is 1, and the activation function of the second hidden unit is g 2 (x); the output of the second hidden unit is λ2 is the internal parameter of the second hidden unit, and 0<λ2<1. By adjusting the value of the internal parameter λ2 of the second hidden unit, the output of the second hidden unit is within the first control feature attribute threshold interval [0, 1], and the output of the second hidden unit is input to the second output unit; The second output unit connected to the second hidden unit has the second hidden unit as its input; the output of the second output unit is Wherein, n is the preset second control characteristic attribute threshold; the output OUT of the second output unit 4 It is the control characteristic attribute of the secondary device, and the output result is 0 or 1; The static feature attribute association network includes: The third input unit connected to the first output unit and the second output unit has the characteristic information matrix of the e-th secondary device and the output OUT of the first output unit as input and output. 2 , the output OUT of the second output unit 4 , also written as IN 3 , the dimension of the second input unit is R+2; where IN 3 It is expressed as: in, is the jth value in the output of the third input unit; The third hidden unit connected to the third input unit, the input of the third hidden unit is the output of the third input unit and the output OUT4 of the knowledge graph model association calculation network; the weight between the third hidden unit and the third input unit is R is the number of feature information, 1≤e≤E, E is the total number of all secondary devices in the substation; the activation function of the third hidden unit is g 3 (x); the output of the third hidden unit is A fourth hidden unit connected to the third hidden unit, the input of the fourth hidden unit is the output OUT of the third hidden unit 5 ; The weight between the fourth hidden unit and the third hidden unit is H is the dimension of the fourth hidden unit; the activation function of the fourth hidden unit is g 4 (x); the output of the fourth hidden unit is The third output unit connected to the fourth hidden unit, the input of the third output unit is the output OUT of the fourth hidden unit 6 , the weight between the fourth hidden unit and the third output unit is The activation function of the third output unit is g 5 (x); the dimension S of the third output unit is also the number of static feature attributes; the output of the third output unit is It is also the static characteristic property of the secondary equipment; The dynamic feature attribute association network includes: The fourth input unit connected to the first output unit, the second output unit, and the third output unit has the characteristic information matrix of the e-th secondary device and the output OUT of the first output unit as input and output. 2 , the output OUT of the second output unit 4 , the output OUT of the third output unit 7 , also written as IN 4 , the dimension of the fourth input unit is S+R+2; where IN 4 It is expressed as: in, is the vth value in the output of the third input unit; 1≤v≤S+R+2, R is the number of feature information, S is the number of static feature attributes; The fifth hidden unit connected to the fourth input unit, the input of the fifth hidden unit is the output of the fourth input unit and the output OUT4 of the knowledge graph model association calculation network; the dimension of the fifth hidden unit is E; the weight between the fifth hidden unit and the fourth input unit is E is the total number of all secondary devices in the substation; the activation function of the fifth hidden unit is g 6 (x); the output of the fifth hidden unit is The sixth hidden unit connected to the fifth hidden unit, the input of the sixth hidden unit is the output OUT of the fifth hidden unit 8 ; The weight between the sixth hidden unit and the fifth hidden unit is P is the dimension of the sixth hidden unit; the activation function of the sixth hidden unit is g 7 (x); the output of the sixth hidden unit is A fourth output unit connected to the sixth hidden unit, the input of the fourth output unit being the output OUT of the sixth hidden unit 9 , the weight between the sixth hidden unit and the fourth output unit is The activation function of the fourth output unit is g 8 (x); the dimension D of the fourth output unit is also the number of dynamic feature attributes; the output of the fourth output unit is It is also the static characteristic property of the secondary equipment; The output of the second network is the correlation between the secondary devices of the substation, relate=[OUT 2 ,OUT 4 ,OUT 7 ,OUT 10 ]; used as attributes and relations in knowledge graphs.

7. An electronic device, characterized in that: It includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, a knowledge graph-based modeling method for substation secondary equipment described in any one of claims 1 to 6 is implemented.

8. A computer-readable storage medium, characterized in that: Computer executable instructions are stored, and the computer executable instructions are used to execute a knowledge graph-based substation secondary equipment modeling method as described in any one of claims 1-6.

Images