Power grid operation situation reasoning method and system based on echo state network

Abstract

The invention belongs to the technical field of power systems, relates to a power grid operation situation reasoning method and system based on an echo state network, and aims to improve the accuracy and efficiency of power grid fault prediction. According to the method, an echo state network model is constructed, and network hyper-parameters are automatically adjusted by using a Bayesian optimization technology, so that the model performance is improved. Meanwhile, a characteristic value obtained through wavelet packet decomposition calculation is input into the model, and prediction of the power grid state and recognition of the fault position are achieved. The system comprises an echo network construction module, a Bayesian hyper-parameter optimization module, a characteristic decomposition calculation module and an output reasoning optimization module. The invention further relates to computer equipment and a computer readable storage medium which are used for implementing the method. According to the technical scheme, through automatic hyper-parameter adjustment and systematic verification, the accuracy of power grid fault detection and the efficiency of power grid operation situation reasoning are improved, rapid power supply recovery of power grid faults is supported, and the method has important practical value and wide application prospects.

Description

Technical Field

[0001] The invention belongs to the technical field of power systems, and in particular relates to a power grid operation status reasoning method based on an echo state network. Background Art

[0002] Echo State Network (ESN) prediction is widely used and has memory capability. The reserve pool of ESN can save information about previous states, which enables it to memorize and model past input data, thereby making better predictions for future inputs. It can effectively process high-dimensional data. The random connections in the reserve pool produce a large number of nonlinear mappings, which helps to efficiently reduce the dimension or extract features of input data. For the hyperparameter value, Bayesian optimization is used to improve ESN to adapt it to the needs of power grid situation reasoning.

[0003] Most of the existing distributed power grids are radial and have many line branches. In actual operation, there will be problems such as ground faults and overloads. These faults are long-term and difficult problems. The maintenance process is cumbersome and time-consuming, difficult to implement, and not conducive to the rapid recovery of the power grid. Therefore, it is necessary to build a fault identification model for the operation status, and use technologies such as echo state network to infer the fault status of the power grid operation to ensure the stable operation of the power grid.

[0004] This patent proposes a method for reasoning about the operation status of a power grid based on a Bayesian optimized echo state network, which is used to reason about the fault status of the power grid. This technical solution constructs an echo state network based on Bayesian optimization and proposes a method for predicting power grid faults; it innovatively designs a fault status reasoning model based on an echo state network, reasoning about its fault identification problem, and supporting rapid power restoration after a power grid fault. Summary of the invention

[0005] In view of the above problems existing in the prior art, the present invention is proposed.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for reasoning about power grid operation status based on echo state network, comprising: constructing an echo state network model;

[0007] Hyperparameters are found through Bayesian optimization, and evaluation indicators of the performance of the echo state network model are established through sampling of acquisition functions based on prior information, using an optimization algorithm with Gaussian process as the probability proxy model;

[0008] The eigenvalues ​​of wavelet packet coefficients are obtained by decomposition calculation and input into the model to obtain the predicted values, and the eigenvalues ​​of output inference are obtained;

[0009] Application based on Bayesian optimization echo state network.

[0010] As a preferred solution of the method for reasoning about power grid operation status based on echo state network described in the present invention, the echo state network model consists of an input layer, a reserve pool, and an output layer.

[0011] As a preferred solution of the method for reasoning about power grid operation status based on echo state network described in the present invention, the hyperparameters found by Bayesian optimization are expressed as:

[0012] y=f(x)+∈

[0013] Among them, x represents the hyperparameter data set, y represents the observed value, and ∈ represents a with a mean of 0 and a variance of 2 noise.

[0014] As a preferred solution of the power grid operation status reasoning method based on echo state network described in the present invention, wherein: the optimization algorithm using Gaussian process as the probability proxy model is expressed as:

[0015] The distribution of a Gaussian process can be represented by the mean function m() and the covariance function k():

[0016] f(x)~GP(m(x), k(x, x'))

[0017]

[0018] f is a function containing hyperparameters, K is the covariance matrix, which is the hyperparameter group data X obtained by the previous t optimizations 1：t ={x 1 , …, x t}Calculated;

[0019] The probability proxy model distribution obtained by the first t times and t+1 times of optimization is as follows:

[0020]

[0021] k t+1 =[k(x t+1 , x1),…,k(x t+1 , xt)] T

[0022] Get the predicted distribution:

[0023]

[0024] As a preferred solution of the method for reasoning about the operation status of a power grid based on an echo state network described in the present invention, the predicted value obtained by inputting the eigenvalue of the wavelet packet coefficient obtained by decomposition calculation into the model includes: the three-phase state values ​​A, B, and C in the output eigenvalues ​​observe whether the three-phase power supply is normal or faulty, the fault type identifies the lateral fault or the longitudinal fault, and the position eigenvalue can identify 4 evenly distributed distances.

[0025] As a preferred solution of the power grid operation status reasoning method based on echo state network described in the present invention, wherein: the application based on Bayesian optimization echo state network includes:

[0026] According to the constructed echo state network model, the dataset is divided into training set, optimization set, validation set, and test set. The state echo network hyperparameter group is initialized, the Bayesian optimization hyperparameter group is used, the optimized dataset and the initialized hyperparameter group are input, the probability improved acquisition function is selected, and the probability proxy model adopts the Gaussian model.

[0027] As a preferred solution of the method for reasoning about power grid operation status based on echo state network described in the present invention, the application based on Bayesian optimization echo state network also includes:

[0028] According to the data set and the proxy model, the distribution p(f * |X 1：t , x t+1 ), maximize the acquisition function to solve the optimal hyperparameter group x for this cycle t+1 , use x t+1 Calculate the objective function value f * , the data (x t+1 , f * ) is added to the optimized data set, t=t+1;

[0029] When the number of iterations exceeds 500, the hyperparameter group optimization is output, otherwise the optimization is performed again;

[0030] Input the optimal hyperparameter group and build the echo state network model;

[0031] Select training set data and update the reserve pool status; train the output weight matrix;

[0032] For the trained model, input verification data, verify the model, and analyze its effectiveness;

[0033] To test the model, the test set is input into the model, the output results are compared with the data set, and the accuracy of detecting power grid fault types and fault distances is tested.

[0034] As a preferred solution of the power grid operation status reasoning system based on echo state network described in the present invention, it includes: an echo network construction module, a Bayesian hyperparameter optimization module, a feature decomposition calculation module, and an output reasoning optimization module.

[0035] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, the steps of any one of the methods for inferring power grid operation status based on an echo state network are implemented.

[0036] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of any one of the methods of reasoning about power grid operation status based on an echo state network are implemented.

[0037] Beneficial effects of the present invention:

[0038] (1) Improve prediction accuracy: Use wavelet packet eigenvalues ​​to predict power grid status and fault locations.

[0039] (2) Optimize model performance: Automatically adjust network hyperparameters through Bayesian optimization.

[0040] (3) Reasonable distribution of data sets: ensuring the comprehensiveness of model training and validation.

[0041] (4) Automated hyperparameter adjustment: reduce manual intervention and improve optimization efficiency.

[0042] (5) Systematic verification: Ensure model reliability through training, verification, and testing.

[0043] (6) Modular design: facilitates system maintenance and upgrades.

[0044] (7) Software and hardware integration: facilitates practical deployment and application. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0046] Figure 1 A schematic diagram of a general echo state network model of an echo state network-based power grid operation status reasoning method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0047] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0048] Example 1

[0049] Reference Figure 1 , which is the first embodiment of the present invention, and provides a method for reasoning about power grid operation status based on echo state network, such as Figure 1 As shown, including:

[0050] In the embodiment of the present application, the echo state network model

[0051] The echo state network (ESN) consists of an input layer, a reservoir, and an output layer, such as Figure 1 shown.

[0052] The input layer shows that u(n) is the input parameter, the number of neurons K. Win is the input weight matrix from the input layer to the reservoir.

[0053] The reserve pool x(n) is the internal neuron state, and the number is N. The reserve pool neurons store historical sequence information, the connection state is random and fixed, and W is the internal weight matrix of the reserve pool.

[0054] The output layer y(n) is the output state, and the number of neurons is L. out is the output weight matrix. back is the connection weight matrix from the output layer to the reserve pool. in , W, W back Generated before training and does not change during training.

[0055] The state update equation of the reserve pool is:

[0056] x(n+1)=f n (W in u(n+1)+Wx(n)+W back y(n)) (1)

[0057] f n is the activation function of the reserve pool neurons, and the Sigmoid function is usually selected.

[0058] The network output equation is:

[0059] y(n+1)=f out (W out(u(n+1),x(n+1))) (2)

[0060] fout is the activation function of the output layer neurons. During training, only W is updated out .

[0061] 5.2 Bayesian Optimization

[0062] Bayesian optimization is an algorithm that effectively finds hyperparameters with good generalization ability. It establishes an evaluation index for the performance of the echo state network model by sampling the acquisition function based on prior information. The patent uses an optimization algorithm with Gaussian process as the probability proxy model. x is the hyperparameter data set, and y is the observation value:

[0063] y=f(x)+∈。∈ is an independent and identically distributed noise with mean 0 and variance a2.

[0064] The distribution of a Gaussian process can be represented by the mean function m() and the covariance function k():

[0065] f(x)~GP(m(x),k(x,x')) (3)

[0066]

[0067] f is a function containing hyperparameters, K is the covariance matrix, which is the hyperparameter group data X obtained by the previous t optimizations 1：t ={x 1 , …, x t}Calculated.

[0068] The probability proxy model distribution obtained by the first t times and t+1 times of optimization is as follows:

[0069]

[0070] k t+1 =[k(x t+1 , x1),…,k(x t+1 , xt)] T (6)

[0071] Get the predicted distribution:

[0072]

[0073] 5.3 Power Grid Fault Situation Reasoning Model

[0074] The eigenvalues ​​of the wavelet packet coefficients obtained by decomposition calculation are input into the above model to obtain the predicted value, and further the eigenvalues ​​of the output inference are obtained.

[0075] The three-phase status values ​​of A, B, and C in the output characteristic values ​​indicate whether the three-phase power supply is normal or faulty. The fault type identifies a horizontal fault or a vertical fault. The position characteristic value can identify four evenly distributed distances.

[0076] Table 1-1 Grid fault identification model

[0077]

[0078] 5.4 Application of Echo State Network Based on Bayesian Optimization

[0079] Step 1: Construct a state echo network according to the principle in 5.1, and divide the data set into training set, optimization set, validation set, and test set according to 5:2:2:1. Initialize the state echo network hyperparameter group and use Bayesian optimization hyperparameter group. Input the optimized data set and the initialized hyperparameter group, select the probability improvement acquisition function, and use the Gaussian model for the probability proxy model. The number of optimization cycles is 500.

[0080] Step 2: Calculate the possible distribution p(f * |X 1：t , x t+1 ), maximize the acquisition function to solve the optimal hyperparameter group xt+1 for this cycle, and use xt+1 to calculate the objective function value f * , this group of data (x t+1 , f * ) is added to the optimized data set, t=t+1;

[0081] Step 3: When the number of iterations in step 2 exceeds 500, the hyperparameter group optimization is output, otherwise repeat step 2;

[0082] Step 4: Input the optimal hyperparameter group from step 3 to build the echo state network model. The input layer is a column vector consisting of 39 eigenvalues, the reserve pool contains 39 neural units, and the output layer contains a column vector of 5 elements;

[0083] Step 5: Select training set data and update the reserve pool state in step 4; train the output weight matrix Wout;

[0084] Step 6: For the model trained in step 4, input the verification data, verify the model, and analyze its effectiveness;

[0085] Step 7: Test the model by inputting the test set into the model, comparing the output results with the data set, and testing its accuracy in detecting power grid fault types and fault distances.

[0086] Example 2

[0087] The second embodiment of the present invention is different from the previous embodiment in that:

[0088] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0089] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0090] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0091] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0092] Example 3

[0093] The third embodiment of the present invention provides a power grid operation status reasoning system based on echo state network, which is characterized by: including an echo network construction module, a Bayesian hyperparameter optimization module, a feature decomposition calculation module, and an output reasoning optimization module.

[0094] Echo network building module, building the echo state network model;

[0095] Bayesian hyperparameter optimization module, which finds hyperparameters through Bayesian optimization, establishes evaluation indicators of echo state network model performance through acquisition function sampling based on prior information, and adopts optimization algorithm with Gaussian process as probability proxy model;

[0096] The characteristic decomposition calculation module uses the decomposition calculation to obtain the characteristic value of the wavelet packet coefficient and inputs it into the model to obtain the predicted value, thereby obtaining the characteristic value of the output inference;

[0097] Output inference optimization module, based on Bayesian optimization echo state network application.

[0098] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A power grid operation status reasoning method based on echo state network, characterized by: include, Construct an echo state network model; Hyperparameters are found through Bayesian optimization, and evaluation indicators of the performance of the echo state network model are established through sampling of acquisition functions based on prior information, using an optimization algorithm with Gaussian process as the probability proxy model; The eigenvalues ​​of wavelet packet coefficients are obtained by decomposition calculation and input into the model to obtain the predicted values, and the eigenvalues ​​of output inference are obtained; Application based on Bayesian optimization echo state network.

2. The method for reasoning about power grid operation status based on echo state network according to claim 1, characterized in that: The echo state network model consists of an input layer, a reserve pool, and an output layer.

3. The method for reasoning about power grid operation status based on echo state network according to claim 2, characterized in that: The hyperparameters found by Bayesian optimization are expressed as, y=f(x)+∈ Among them, x represents the hyperparameter data set, y represents the observed value, and ∈ represents a with a mean of 0 and a variance of 2 noise.

4. The method for reasoning about power grid operation status based on echo state network according to claim 3, characterized in that: The optimization algorithm using Gaussian process as the probability proxy model is expressed as: The distribution of the Gaussian process can be expressed by the mean function m() and the covariance function k() as follows, f(x)~GP(m(x), k(x, x')) f is a function containing hyperparameters, K is the covariance matrix, which is the hyperparameter group data X obtained by the previous t optimizations 1：t ={x1,…,x t }Calculated; The probability proxy model distribution obtained by the first t times and t+1 times of optimization is expressed as: k t+1 =[k(x t+1 ,x1),...,k(x t+1 ,xt)] T The predicted distribution is obtained, expressed as, 5. The method for reasoning about power grid operation status based on echo state network according to claim 4, characterized in that: The predicted value obtained by inputting the eigenvalue of the wavelet packet coefficient obtained by decomposition calculation into the model includes: the three-phase state values ​​A, B, and C in the output eigenvalue observe whether the three-phase power supply is normal or faulty, the fault type identifies the horizontal fault or the longitudinal fault, and the position eigenvalue can identify 4 uniformly distributed distances.

6. The method for reasoning about power grid operation status based on echo state network according to claim 5, characterized in that: The application based on Bayesian optimization echo state network includes: According to the constructed echo state network model, the dataset is divided into training set, optimization set, validation set, and test set. The state echo network hyperparameter group is initialized, the Bayesian optimization hyperparameter group is used, the optimized dataset and the initialized hyperparameter group are input, the probability improved acquisition function is selected, and the probability proxy model adopts the Gaussian model.

7. The method for reasoning about power grid operation status based on echo state network according to claim 6, characterized in that: The application based on Bayesian optimization echo state network also includes: According to the data set and the proxy model, the distribution p(f*IX 1：t , x t+1 ), maximize the acquisition function to solve the optimal hyperparameter group x for this cycle t+1 , use x t+1 Calculate the objective function value f * , the data (x t+1 , f * ) is added to the optimized data set, t=t+1; When the number of iterations exceeds 500, the hyperparameter group optimization is output, otherwise the optimization is performed again; Input the optimal hyperparameter group and build the echo state network model; Select training set data and update the reserve pool status; Training output weight matrix; For the trained model, input verification data, verify the model, and analyze its effectiveness; To test the model, the test set is input into the model, the output results are compared with the data set, and the accuracy of detecting power grid fault types and fault distances is tested.

8. A system based on the method for reasoning power grid operation status based on echo state network according to any one of claims 1 to 7, characterized in that: It includes echo network construction module, Bayesian hyperparameter optimization module, feature decomposition calculation module, and output inference optimization module.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images