CN116316699A

CN116316699A - Large power grid frequency security situation prediction method, device and storage medium

Info

Publication number: CN116316699A
Application number: CN202310340864.7A
Authority: CN
Inventors: 梁纪峰; 戎士洋; 范辉; 李先妹; 张蕊; 王蕾报; 孟凡超
Original assignee: State Grid Corp of China SGCC; Electric Power Research Institute of State Grid Hebei Electric Power Co Ltd; State Grid Hebei Energy Technology Service Co Ltd
Current assignee: State Grid Corp of China SGCC; Electric Power Research Institute of State Grid Hebei Electric Power Co Ltd; State Grid Hebei Energy Technology Service Co Ltd
Priority date: 2023-03-31
Filing date: 2023-03-31
Publication date: 2023-06-23

Abstract

The application provides a large power grid frequency security situation prediction method, a device and a storage medium. The method comprises the following steps: acquiring generator parameters, wherein the generator parameters comprise mechanical power, electromagnetic power, damping coefficient and initial angular frequency; based on a dynamic prediction model, carrying out feature extraction on generator parameters to obtain a first frequency characteristic, and constructing the dynamic prediction model based on a particle swarm algorithm and support vector regression; performing feature extraction on generator parameters based on a steady-state prediction model to obtain a second frequency characteristic, wherein the steady-state prediction model is constructed based on a lightweight gradient elevator and a production countermeasure network; based on a fusion prediction model, fusing the first frequency characteristic and the second frequency characteristic to obtain a target frequency characteristic, and constructing the fusion prediction model based on a neural network and fuzzy logic; and obtaining a prediction result according to the target frequency characteristic. The method and the device can improve the prediction accuracy of the frequency dynamic response.

Description

Large power grid frequency security situation prediction method, device and storage medium

Technical Field

The application relates to the technical field of power system prediction, in particular to a method and device for predicting a large power grid frequency security situation and a storage medium.

Background

Along with the continuous increase of the uncertainty of the system, the change of the operation mode is more and more severe, the function requirement of real-time online evaluation is also enhanced, and from the evaluation object, network voltage, frequency, phase and the like are all important targets for online evaluation of the system.

In terms of system frequency situation prediction, the following maturing methods have been proposed since the 20 s of the last century: time domain simulation based on causal theory, average system frequency model (average system frequency, ASF), system frequency response model (system frequency response, SFR), etc. The time domain simulation method calculates the frequency dynamics after disturbance by a strategy of gradual integration, and the calculation result is accurate; the average system frequency model (ASF) and the system frequency response model (SFR) both assume that the system frequency is consistent in each region, but the frequency of a large power grid has complex time and space distribution characteristics, and a large error can be generated when dynamic frequency analysis is performed, so that a great number of problems exist when the equivalent model is applied to dynamic frequency analysis after disturbance of a modern large-scale power grid.

The existing two technical systems of online real-time evaluation and frequency security situation prediction of the power grid are relatively fractured, the technical requirements of future system frequency evolution prediction are difficult to meet in the traditional online evaluation mode of obtaining specific indexes through PMU data processing, and the traditional online evaluation mode of determining models and simulating time by time is difficult to be directly applied to online evaluation of the power grid. Therefore, it is necessary to develop a real-time predictive study of the security situation of a large system frequency taking into account multiple uncertainty factors.

Disclosure of Invention

The application provides a large power grid frequency security situation prediction method, a device and a storage medium, which are used for solving the problem that in the prior art, the prediction accuracy of frequency dynamic response is not high under the condition that a system is disturbed.

In a first aspect, the present application provides a method for predicting a frequency security situation of a large power grid, including:

acquiring generator parameters, wherein the generator parameters comprise mechanical power, electromagnetic power, damping coefficient and initial angular frequency;

performing feature extraction on the generator parameters based on a dynamic prediction model to obtain a first frequency characteristic, wherein the dynamic prediction model is constructed based on a particle swarm algorithm and support vector regression;

performing feature extraction on the generator parameters based on a steady-state prediction model to obtain a second frequency characteristic, wherein the steady-state prediction model is constructed based on a lightweight gradient elevator and a production countermeasure network;

based on a fusion prediction model, fusing the first frequency characteristic and the second frequency characteristic to obtain a target frequency characteristic, wherein the fusion prediction model is constructed based on a self-adaptive neural fuzzy system;

and obtaining a prediction result according to the target frequency characteristic.

In a second aspect, the present application provides a large grid frequency security situation prediction apparatus, including:

The acquisition module is used for acquiring generator parameters, wherein the generator parameters comprise mechanical power, electromagnetic power, damping coefficient and initial angular frequency;

the first prediction module is used for extracting characteristics of the generator parameters based on a dynamic prediction model to obtain a first frequency characteristic, and the dynamic prediction model is constructed based on a particle swarm algorithm and support vector regression;

the second prediction model is used for extracting characteristics of the generator parameters based on a steady-state prediction model to obtain a second frequency characteristic, and the steady-state prediction model is constructed based on a lightweight gradient elevator and a production countermeasure network;

the fusion module is used for fusing the first frequency characteristic and the second frequency characteristic based on a fusion prediction model to obtain a target frequency characteristic, and the fusion prediction model is constructed based on a self-adaptive neural fuzzy system;

and the judging module is used for obtaining a prediction result according to the target frequency characteristic.

In a third aspect, the present application provides a terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the method according to the first aspect or any one of the possible implementations of the first aspect when the computer program is executed.

In a fourth aspect, the present application provides a computer readable storage medium storing a computer program which, when executed by a processor, implements the steps of the method as described above in the first aspect or any one of the possible implementations of the first aspect.

The utility model provides a big electric wire netting frequency security situation prediction method, device and storage medium, this application obtains first frequency characteristic and second frequency characteristic respectively through dynamic prediction model and steady state prediction model, and rethread fuses prediction model to first frequency characteristic and second frequency characteristic and fuses and obtain target frequency characteristic, judges big electric wire netting frequency security situation according to target frequency characteristic, because target frequency characteristic becomes more accurate after the fusion to the prediction accuracy of frequency dynamic response has been improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are required for the embodiments or the description of the prior art will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a flowchart of an implementation of a large power grid frequency security situation prediction method provided in an embodiment of the present application;

FIG. 2 is a training flow diagram of a dynamic predictive model provided by an embodiment of the present application;

FIG. 3 is a schematic diagram of a production countermeasure network provided by an embodiment of the present application;

FIG. 4 is a training flow diagram of a steady state predictive model provided by an embodiment of the present application;

FIG. 5 is a schematic structural diagram of a fusion prediction model according to an embodiment of the present disclosure;

fig. 6 is a schematic structural diagram of a large power grid frequency security situation prediction device provided in an embodiment of the present application;

fig. 7 is a schematic diagram of a terminal provided in an embodiment of the present application.

Detailed Description

In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system configurations, techniques, etc. in order to provide a thorough understanding of the embodiments of the present application. It will be apparent, however, to one skilled in the art that the present application may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.

For the purpose of making the objects, technical solutions and advantages of the present application more apparent, the following description will be made with reference to the accompanying drawings by way of specific embodiments.

Fig. 1 is a flowchart of implementation of a method for predicting a frequency security situation of a large power grid according to an embodiment of the present application, which is described in detail below:

in step 101, generator parameters are obtained, including mechanical power, electromagnetic power, damping coefficient, and initial angular frequency.

When the power system is disturbed, the operation frequency of each synchronous generator in the system can be influenced under the action of unbalanced power. Therefore, in the embodiment of the application, the obtained generator parameters include mechanical power, electromagnetic power, damping coefficient and initial angular frequency, and are used for calculating and judging whether the frequency in the power system is safe or not, so that the stability of the safety situation of the frequency of the large power grid is ensured.

In one possible implementation, after step 101, the method may further include:

obtaining the active power variation of the generator according to the difference between the mechanical power and the electromagnetic power, and preprocessing the active power variation of the generator to obtain the preprocessed active power variation of the generator;

performing fixed-order, identification and conversion on the preprocessed active power variation of the generator based on an identification model to obtain a frequency variation, wherein the identification model is constructed based on a red pool information quantity criterion and a system identification method;

And calculating an inertial time constant for the frequency variation by adopting a rocking equation, and determining the inertial center frequency of the system according to the inertial time constant.

In the embodiment of the present application, the mechanical power P obtained in step 101 is set _m And electromagnetic power P _e The difference is taken to obtain the active power variation delta P of the generator, namely delta P=P _m -P _e . Then carrying out trend removal and low-pass filtering noise reduction pretreatment on the generator active power variation delta P to obtain pretreated generator active power variation delta P', inputting an identification model, and carrying out fixed order, identification and conversion to obtain a frequency variation

Frequency variation +.>

Inputting a swinging equation, calculating an inertia time constant H, and inputting the inertia time constant H into a dynamic frequency equation to calculate the inertial center frequency of the system +.>

The formula for the rocking equation is as follows:

wherein D is the damping coefficient of the generator, and w is the initial angular frequency of the generator.

The dynamic frequency equation is as follows:

wherein H is _sys Is the sum of inertia time constants H and D of all generator sets in the system _i Damping system for ith generator, w _i Is the initial angular frequency of the ith generator.

Because of a large amount of noise in the measured power grid data under the normal operation condition of the power system, when the fitting degree of the identification model is high, the fitting phenomenon may occur. In order to avoid the problem that a large amount of identification data may cause overfitting of an identification model, in the embodiment of the application, the trend removal and low-pass filtering noise reduction pretreatment are required for the active power variation delta P of the generator.

The red-pool information quantity criterion (Akaike Information Criterion, AIC) is a standard for measuring the fitting superiority of a statistical model, and is based on the concept of entropy, so that the complexity of an estimated model and the superiority of the model fitting data can be weighed.

System recognition is the determination of a mathematical model describing the behavior of a system from the input-output time function of the system. The objective of building a mathematical model by recognition is to estimate important parameters characterizing the behavior of the system, build a model that mimics the behavior of a real system, predict the future evolution of the system output using the inputs and outputs of the system that are currently measurable, and design the controller. The main problem in analysing the system is to determine the output signal based on the input time function and the characteristics of the system.

For the order of the identification model of the power system, the expression effect of the identification model is not accurate enough due to the too low order, and the too high order can reach the occurrence of the overfitting phenomenon. In order to avoid deviation of the power system to the identification result due to the identification model order, AIC is used in the embodiment of the application to remove trend and reduce noise by low-pass filtering, and the change quantity delta P of the active power of the generator is preprocessed ^′ An appropriate system recognition model order is determined, which focuses on the external property expression of the measured data.

AIC is another application of the principle of maximum likelihood as one of the most common model selection criteria. According to the principle of maximum likelihood, a model approximating the actual process is found such that the probability distribution of the output of the model approximates as closely as possible to the probability distribution of the output of the actual process. The AIC performance index expression formula is as follows:

AIC(n)＝Nlnρ _N +2N (3)

wherein N is the number of parameters in the identification model, ρ _n Is the variance of the prediction error of the nth order model. The meaning of AIC can be understood as: probability density function estimated by AIC model and actual probability densityAn estimate of the Kullback-Leibler distance between the degree functions. The AIC performance index combines model complexity and flexibility to adapt to data, and the best-fit model corresponds to the smallest AIC value.

After the identification model is obtained, the identification model needs to be converted into a first-order transfer function form, the fitting degree of the identification model is reduced in the order reduction process, errors are introduced, output result deviation is caused, and the higher the order of the identification model is, the larger the errors caused by the order reduction of the identification model are. In order to avoid the reduction of the fitting degree of the AIC fixed-order identification model in the order reduction process, the embodiment of the application converts the discrete time identification model obtained by identification into a continuous time model, then inputs a step into the identification model, and outputs a frequency variation of an output value within 0.5-2 s after the step response

Substituting the inertial time constant H into the swing equation to calculate the inertial center frequency of the system according to the inertial time constant>

In step 102, feature extraction is performed on generator parameters based on a dynamic prediction model, so as to obtain a first frequency characteristic, and the dynamic prediction model is constructed based on a particle swarm algorithm and support vector regression.

The mechanical power P of the generator obtained in the step 101 _m Electromagnetic power P _e And inputting the damping coefficient D and the initial angular frequency w into a dynamic prediction model, and extracting the characteristics of the dynamic prediction model to obtain a first frequency characteristic.

The particle swarm algorithm (Particle Swarm optimization, PSO) is translated into a particle swarm optimization algorithm, a particle swarm optimization algorithm or a particle swarm optimization algorithm, and is a random search algorithm based on swarm cooperation developed by simulating the foraging behavior of the bird swarm. It is generally considered to be one of the cluster intelligence (Swarm intelligence, SI).

Support vector regression (support vector regression, SVR) is a supervised learning algorithm used to predict discrete values. In SVR, the best fit line is the most point hyper-plane.

The essence of SVR is to solve the classification problem. For a given training sample set s= { (x) _i ,y _i )|x _i ∈R ⁿ ,y _i E R }, i=1, 2, …, l, where x _i Inputting a feature vector for n-dimension of the ith sample, y _i For the classification category of the ith sample, R is the real space, R ⁿ And (3) solving the optimal classification hyperplane for an n-dimensional real space, wherein l is the total number of samples. In contrast to the classification problem of the support vector machine, the output of the regression problem of the support vector machine is no longer a discrete value but becomes a continuous value. The regression problem of the support vector machine is to find the implicit regression function according to limited observed data, namely, find the mapping relation f from the input space to the output space: r is R ⁿ R, i.e. solving regression functions

Wherein w is a weight coefficient, b is a bias vector matrix,>

for nonlinear mapping input into high-dimensional space, x is the input feature vector.

In the embodiment of the application, v-SVR is adopted as a dynamic prediction model, an influence factor v (v is more than or equal to 0) of an insensitive loss coefficient is introduced, and the prediction precision is improved by solving proper v. The optimization problem to be solved in this case is:

wherein C is penalty factor, ζ _i And

epsilon is the insensitivity loss factor for the relaxation variable.

Constructing Lagrange function to obtain the dual problem of the formula (4):

wherein alpha is _i And

are Lagrange multipliers, K (x _i ,x _j ) For corresponding to transform->

Is a kernel function of (a).

Solving convex quadratic optimization programming problem to obtain alpha _i And

is a solution to (a). Thus, the regression prediction function can be constructed as:

in one possible implementation, the first frequency characteristic includes a first maximum frequency change rate, a first transient frequency extremum, and a first quasi-steady state frequency, and step 102 may specifically include:

and inputting the mechanical power, the electromagnetic power, the damping coefficient, the initial angular frequency and the system inertial center frequency into a dynamic prediction model to obtain a first frequency characteristic.

Transient frequency extremum f _nadir And frequency rate of change (rate of change of frequency, rocofs) are commonly used as triggering signals for protective elements and control devices in electrical networks, often by maximum frequency rate of change after a power system is disturbed

Transient frequency extremum f _nadir And a quasi-steady-state frequency f _ss To determine whether the frequency of the power system can be kept stable. Thus, in the embodiment of the present application, the mechanical power P of the generator obtained in step 101 _m Electromagnetic power P _e Inputting the damping coefficient D and the initial angular frequency w into a dynamic prediction model to obtain a first frequency characteristicSex f _SVR Wherein the first frequency characteristic f _SVR Comprising the following steps: first maximum rate of frequency change->

First transient frequency extremum->

And a first quasi-steady state frequency->

In one possible implementation, the training process of the dynamic prediction model may include:

Acquiring mechanical power, electromagnetic power, damping coefficient, initial angular frequency and system inertial center frequency, and acquiring a standard sample set by adopting a characteristic screening method according to the mechanical power, the electromagnetic power, the damping coefficient, the initial angular frequency and the system inertial center frequency, wherein the standard sample set comprises a training sample and a test sample;

training the support vector regression by taking the training sample as input and the corresponding characteristic of the training sample as output;

and optimizing the test sample by adopting a particle swarm algorithm according to the trained support vector regression to obtain a dynamic prediction model.

Since the dynamic frequency after disturbance of the power system is directly related to the motion equation of the rotor of the generator, for a multi-unit power system, the dynamic frequency equation is shown in the formula (2), and the electromagnetic power P generated by the generator in the system in a steady state is considered _e Balanced with the power consumed in the grid, there are:

wherein P is _lossij For the line loss from node i to node j in the power system, P _lj And the active power consumed by the jth load is m is the number of generator nodes of the power system, and n is the number of generator nodes of the power system.

The mechanical power P to be obtained _m Electromagnetic power P _e Damping coefficient D, initial angular frequency w and system inertial center frequency

And after the standard sample set is obtained through feature screening, training the SVR model by using a training sample in the standard sample set, and performing performance test on the trained SVR model by using another part of test samples. The method comprises the steps of optimizing parameters of an SVR model in a training process, adopting a particle swarm algorithm as an SVR model optimal parameter searching algorithm to obtain a dynamic prediction model, receiving power grid data measured by a power system after disturbance by using the trained dynamic prediction model, and rapidly predicting frequency characteristics after the power system is disturbed, namely a first frequency characteristic f _SVR Comprising a first maximum rate of frequency change +.>

Transient frequency extremum->

And a quasi-steady state frequency->

A specific training procedure can be seen in fig. 2.

In step 103, feature extraction is performed on generator parameters based on a steady-state prediction model, so as to obtain a second frequency characteristic, and the steady-state prediction model is constructed based on a lightweight gradient elevator and a production countermeasure network.

The mechanical power P of the generator obtained in the step 101 _m Electromagnetic power P _e And inputting the damping coefficient D and the initial angular frequency w into a dynamic prediction model, and extracting features of the dynamic prediction model to obtain a second frequency characteristic.

The lightweight gradient hoist (light gradient boosting machine, lightGBM) is a boosting integrated learning model based on a gradient hoisting tree (gradient boosting decision tree, GBDT) and uses a classification regression tree (classification and regression trees, CART) as a base learner. The CART algorithm recursively divides all samples to construct a binary tree, the feature space is divided into limited sub-areas, and the average value of all samples in the sub-areas is taken as the output of the sub-areas. The construction engineering of the CART algorithm is as follows:

(1) A root node is constructed containing all samples.

(2) The optimal splitting characteristic j and the dividing point s are selected, the least square error is used as the optimal dividing basis, and the calculation formula is as follows:

wherein y is _i As the actual value of the variable c ₁ C is the predicted value of the left node after splitting ₂ Is the predicted value of the right node after splitting. Traversing the variable j, scanning the dividing point s for the feature j, selecting (j, s) minimizing the formula (8), R ₁ And R is ₂ Refers to the left and right subregions.

(3) The output value is determined by dividing the subregion by the selected (j, s).

(4) Continuing to call steps (2) and (3) on the two sub-areas until a stop condition is met.

(5) Dividing the input space into N sub-regions R ₁ ,R ₂ ,…,R _n A regression tree is generated.

GBDT adopts boosting idea, and a new CART tree is generated in each iteration to fit the residual error of the previous round of result, so that the difference between the fitting value and the target value is smaller. And accumulating the CART tree output values generated by each round of iteration to obtain a final learning result. The lightGBM is improved on the basis of GBDT, a histogram algorithm is fused, mutually exclusive feature binding and single-side gradient sampling are performed, memory occupation is reduced, and training speed is improved. The use of a per-leaf growth strategy with depth limitation avoids the generation of deeper trees, preventing overfitting.

The production countermeasure network (generative adversarial network, GAN) comprises a generation model and a discrimination model, the specific structure of which is shown in fig. 3, and the true samples are learned in an unsupervised manner to simulate the data distribution condition for the purpose of generating similar sample data. The training process is as follows: firstly, a generating model is fixed, a real sample is input into a judging model to train to achieve a certain accuracy, then the judging model is fixed, a group of random variables are input into the generating model, and then the generating sample output by the generating model is input into the judging model to judge. And in the training process, the counter propagation gradient is simultaneously used for the generation model and the discrimination model, and model parameters are continuously updated until the discrimination model cannot distinguish a real sample from the generation sample. The training process can be expressed by the formula (11), and the formula (11) is as follows:

Wherein xP is as follows _data (x) For x taken from the true sample distribution zP _z (z) is z taken from the simulated sample distribution, G (z) is the generated model, D (x) is the discriminant model, and G and D are microtransmissible functions to ensure that errors of the model can be counter-propagated.

After the system topology is changed, the data dimension is changed into m dimension, n dimension random variable is input into a generator trained under a small number of samples of a new system, m dimension characteristics are output, and the generator enters a discriminator to check whether the data dimension is close to a real sample.

In one possible implementation, the second frequency characteristic includes a second maximum frequency change rate, a second transient frequency extremum, and a second quasi-steady state frequency, and step 103 may specifically include:

and inputting the mechanical power, the electromagnetic power, the damping coefficient, the initial angular frequency and the system inertial center frequency into a steady state prediction model to obtain a second frequency characteristic.

The application obtains the mechanical power P of the generator obtained in the step 101 _m Electromagnetic power P _e Inputting the damping coefficient D and the initial angular frequency w into a steady state prediction model to obtain a second frequency characteristic f _GBM Wherein the second frequency characteristic f _GBM Comprising the following steps: second maximum rate of frequency change

Second transient frequency extremum->

And a second quasi-steady state frequency->

In one possible implementation, the training process of the steady state prediction model may include:

Dividing mechanical power, electromagnetic power, damping coefficient, initial angular frequency and system inertial center frequency into a training set and a testing set;

inputting the training set into a maximum frequency change rate prediction model, calculating the importance of a first feature by adopting a lightweight gradient elevator, deleting the frequency feature with the lowest score according to an attention mechanism, and obtaining the maximum frequency change rate prediction model after feature processing;

inputting the training set into a transient frequency extremum prediction model, calculating a second feature importance degree by adopting a lightweight gradient elevator, deleting the frequency features with the lowest scores according to an attention mechanism, and obtaining a transient frequency extremum prediction model after feature processing;

inputting the training set into a quasi-steady state frequency prediction model, calculating a third feature importance degree by adopting a lightweight gradient elevator, deleting the frequency features with the lowest scores according to an attention mechanism, and obtaining a quasi-steady state frequency prediction model after feature processing;

and verifying the test set by adopting a production countermeasure network according to the maximum frequency change rate prediction model, the transient frequency extremum prediction model and the quasi-steady state frequency prediction model after feature processing to obtain a steady state prediction model.

The specific training process can be seen in fig. 4, in which in the embodiment of the present application, the mechanical power P is to be obtained _m Electromagnetic power P _e Damping coefficient D, initial angular frequency w and system inertial center frequency

The method is divided into a training set and a testing set: respectively inputting the training set into a maximum frequency change rate prediction model, a transient frequency extremum prediction model and a quasi-steady state frequency prediction model, respectively calculating a first feature importance, a second feature importance and a third feature importance by adopting a lightGBM, and then removing the corresponding lowest N according to an attention mechanism ₁ 、N ₂ 、N ₃ The method comprises the steps of judging whether the precision of a current prediction model is reduced to a first preset precision, a second preset precision and a third preset precision or not, if yes, determining a trained maximum frequency change rate prediction model, a transient frequency extremum prediction model and a quasi-steady frequency prediction model, and if no, N _i ＝ _i Returning +1 (i=1, 2, 3) to delete the frequency characteristic step with the lowest score according to the attention mechanism to continue execution until the current prediction model meets the preset requirement; and then respectively inputting the test set into a trained maximum frequency change rate prediction model, a transient frequency extremum prediction model and a quasi-steady state frequency prediction model for verification, and if the performance requirement of the model is met, determining a steady state prediction model according to the trained maximum frequency change rate prediction model, the transient frequency extremum prediction model and the quasi-steady state frequency prediction model.

In step 104, the first frequency characteristic and the second frequency characteristic are fused based on a fusion prediction model, so as to obtain a target frequency characteristic, and the fusion prediction model is constructed based on an adaptive neural fuzzy system.

Due to the first frequency characteristic f obtained from the dynamic predictive model _SVR And a steady state prediction model to obtain a second frequency characteristic f _GBM Taking into account the influence of all generator sets, loads and network structures in the power system on the frequency dynamic response, the method is characterized by predicting the resultThe accuracy increases with the increase of sample accuracy and data, and the disadvantage is that the prediction model does not involve the physical process of power system frequency response, and thus lacks in the reliability of the prediction result. The Adaptive neural Network-based Fuzzy Inference System (ANFIS) is a comprehensive algorithm which integrates the advantages of a neural Network and fuzzy logic, so that the strong self-learning and self-adaptation capabilities of the neural Network can be considered, and the capability of the fuzzy system for accurately expressing objective physical laws can be fully utilized. Therefore, in the embodiment of the present application, the ANFIS is used to construct the fusion prediction model, and the first frequency characteristic f obtained in step 102 is obtained _SVR And a second frequency characteristic f obtained in step 103 _GBM Inputting the fusion prediction model, and fusing to obtain the target frequency characteristic f _ANFIS Thereby obtaining more accurate and reliable frequency prediction results. The ANFIS is a novel neural network formed by organically combining fuzzy logic and the neural network, and the basic idea is that the parameters of a fuzzy reasoning system are adjusted by adopting a back propagation algorithm or a mixed algorithm combining the back propagation algorithm and a least square method based on a Sugeno fuzzy model, so that the designed fuzzy reasoning system can best simulate the relation between actual input and actual output.

In one possible implementation, the target frequency characteristic includes a target maximum frequency change rate, a target transient frequency extremum, and a target quasi-steady state frequency, and step 104 may specifically include:

and inputting the first frequency characteristic and the second frequency characteristic into a fusion prediction model to obtain the target frequency characteristic.

The embodiment of the application obtains the first frequency characteristic f obtained in the step 102 _SVR And a second frequency characteristic f obtained in step 103 _GBM Inputting the fusion prediction model, and fusing to obtain the target frequency characteristic f _ANFIS Wherein the target frequency characteristic f _ANFIS Comprising the following steps: target maximum rate of frequency change

Target transient frequency extremum- >

And a target quasi-steady state frequency +>

In one possible implementation, the training process of fusing the predictive models may include:

performing fuzzy extraction on the first frequency characteristic and the second frequency characteristic by adopting a fuzzy logic method to obtain a first fuzzy rule corresponding to the first frequency characteristic and a second fuzzy rule corresponding to the second frequency characteristic;

and fusing the first fuzzy rule and the second fuzzy rule based on the neural network to obtain a fused prediction model.

The structure of the fusion prediction model constructed in the embodiment of the present application is shown in fig. 5, where the ANFIS has 5 layers, and the inputs of the fusion prediction model are the first frequency characteristics f obtained by the dynamic prediction model respectively _SVR And a steady state prediction model to obtain a second frequency characteristic f _GBM The fusion process of two prediction results is divided into: the device comprises a blurring layer, a blurring rule layer, a normalization layer, an input connection layer and an output layer. In FIG. 5, A ₁ 、A ₂ 、A ₃ 、B ₁ 、B ₂ 、B ₃ Representing fuzzy sets, P _ord Representing the vector multiplied by N _orm Representing the norm, f ₁ -f ₆ Representing the fuzzy rule.

The main function of the blurring layer is to carry out blurring processing on each input by using a plurality of membership functions, and the embodiment of the application mainly selects a bell-shaped membership function and a Gaussian membership function. The output of the blurring layer after blurring processing is the blurring characteristic of each input variable, namely:

Wherein, subscript header of O is respectively 1, 2, 3 and 4 which respectively represent node numbers, O _1,A,a 、O _1,B,b For membership of fuzzy sets, usingSubscripts a and B represent the input quantities made up of the predicted results of the two sub-model methods,

and a and b are membership function serial numbers.

The main function of the fuzzy rule layer is to realize the product operation of all input fuzzy values, the output of each node of the layer represents the excitation intensity of a fuzzy rule, and the establishment of the fuzzy rule can realize the complete combination of fuzzy features of the prediction results of the two sub-model methods. Excitation intensity of each fuzzy rule which can be obtained through the fuzzy rule layer:

O _2,1 ＝O _1,A,a ·O _1,B,b ,a,b＝1,2,3 (13)

the main function of the normalization is to obtain a normalized excitation intensity by dividing the intensity of each output fuzzy rule of the previous layer by the total intensity of all fuzzy rules, namely:

O _3,1 ＝O _2,1 /∑ ₁ O _2,1 (14)

the main function of the input connection layer is to connect the output of the normalization layer to the input of the ANFIS, the output of which represents the final contribution of each rule to the total output, namely:

O _4,1 ＝O _3,1 (a ₁ f _SVR +b ₁ f _GBM +c ₁ ) (15)

wherein a is ₁ 、b ₁ 、c ₁ The set of parameters that are the nodes are referred to as the back-piece parameters.

The main function of the output layer is to add all the output signals of the previous layer, and the final obtained target frequency characteristic of the fusion prediction model is that:

f _ANFIS ＝∑ ₁ O _4,1 (16)

Based on the above, the input data of the fusion prediction model mainly comprises frequency characteristics of the dynamic prediction model and the steady state prediction model, and the output data is the target frequency characteristics generated by algorithm simulation. The precondition parameters are learned by adopting a back propagation gradient descent method, the conclusion parameters are determined by adopting a least square method, and the training sample precision condition is met by continuously cycling the learning process, so that a trained fusion prediction model is finally obtained.

In step 105, a prediction result is obtained based on the target frequency characteristic.

According to the target frequency characteristic f obtained in step 104 _ANFIS I.e. target maximum rate of frequency change

Target transient frequency extremum->

And a target quasi-steady state frequency +>

And determining whether the active frequency security situation of the current large power grid is stable.

In one possible implementation, step 105 may specifically include:

judging whether the target maximum frequency change rate is not greater than a first preset value, judging whether the target transient frequency extremum is not greater than a second preset value, and judging whether the target quasi-steady state frequency is not greater than a third preset value;

if the target maximum frequency change rate is not greater than the first preset value, the target transient frequency extremum is not greater than the second preset value, and the target quasi-steady state frequency is not greater than the third preset value, determining that the active frequency security situation of the large power grid is stable according to the target maximum frequency change rate, the target transient frequency extremum and the target quasi-steady state frequency;

And if at least one of the target maximum frequency change rate is not greater than a first preset value, the target transient frequency extremum is not greater than a second preset value and the target quasi-steady state frequency is not greater than a third preset value does not meet the requirements, determining that the active frequency security situation of the large power grid is unstable.

In the embodiment of the present application, the target frequency characteristic f obtained according to step 104 _ANFIS I.e. target maximum rate of frequency change

Target transient frequency extremum->

And a target quasi-steady state frequency +>

Respectively giving the target maximum frequency change rate

Target transient frequency extremum->

And a target quasi-steady state frequency +>

Setting a first preset value epsilon ₁ A second preset value epsilon ₂ And a third preset value epsilon ₃ Then judging the target maximum frequency change rate +.>

Whether or not it is not greater than a first preset value epsilon ₁ And, judge the target transient frequency extremum +.>

Whether or not it is not greater than a second preset value epsilon ₂ And, judge the target quasi-steady state frequency +.>

Whether or not it is not greater than a third preset value epsilon ₃ 。

If the target maximum frequency change rate

Not greater than a first preset value epsilon ₁ And, target transient frequency extremum

Not greater than a second preset value epsilon ₂ And, target quasi-steady state frequency +>

Not greater than a third preset value epsilon ₃ I.e.

The active frequency security situation of the current large power grid is determined to be stable.

If the target maximum frequency change rate

Not greater than a first preset value epsilon ₁ Target transient frequency extremum->

Not greater than a second preset value epsilon ₂ Target quasi-steady state frequency->

Not greater than a third preset value epsilon ₃ At least one of them does not meet the requirements, i.e

And if at least one of the active frequency safety situations is met, determining that the active frequency safety situation of the current large power grid is unstable.

According to the method, the first frequency characteristic and the second frequency characteristic are respectively obtained through the dynamic prediction model and the steady state prediction model, the first frequency characteristic and the second frequency characteristic are fused through the fusion prediction model to obtain the target frequency characteristic, the large power grid frequency security situation is judged according to the target frequency characteristic, and the target frequency characteristic becomes more accurate after fusion, so that the prediction accuracy of frequency dynamic response is improved.

It should be understood that the sequence number of each step in the foregoing embodiment does not mean that the execution sequence of each process should be determined by the function and the internal logic of each process, and should not limit the implementation process of the embodiment of the present application in any way.

The following are device embodiments of the present application, for details not described in detail therein, reference may be made to the corresponding method embodiments described above.

Fig. 6 shows a schematic structural diagram of a large grid frequency security situation prediction device provided in an embodiment of the present application, and for convenience of explanation, only a portion relevant to the embodiment of the present application is shown, which is described in detail below:

as shown in fig. 6, the large grid frequency security situation prediction apparatus 6 includes:

an acquisition module 61 for acquiring generator parameters including mechanical power, electromagnetic power, damping coefficient and initial angular frequency;

the first prediction module 62 is configured to perform feature extraction on generator parameters based on a dynamic prediction model, so as to obtain a first frequency characteristic, where the dynamic prediction model is constructed based on a particle swarm algorithm and support vector regression;

the second prediction model 63 is used for extracting characteristics of generator parameters based on a steady-state prediction model to obtain a second frequency characteristic, and the steady-state prediction model is constructed based on a lightweight gradient elevator and a production countermeasure network;

the fusion module 64 is configured to fuse the first frequency characteristic and the second frequency characteristic based on a fusion prediction model, so as to obtain a target frequency characteristic, where the fusion prediction model is constructed based on a self-adaptive neural fuzzy system;

the judging module 65 is configured to obtain a prediction result according to the target frequency characteristic.

The utility model provides a big electric wire netting frequency security situation prediction unit, this application obtains first frequency characteristic and second frequency characteristic respectively through dynamic prediction model and steady state prediction model, and rethread fuses prediction model to first frequency characteristic and second frequency characteristic and fuses and obtain target frequency characteristic, judges big electric wire netting frequency security situation according to target frequency characteristic, because target frequency characteristic becomes more accurate after the fusion to the prediction accuracy of frequency dynamic response has been improved.

In one possible implementation manner, after the obtaining module, the apparatus may further include:

the pretreatment module is used for obtaining the active power variation of the generator according to the difference between the mechanical power and the electromagnetic power, and carrying out pretreatment on the active power variation of the generator to obtain the pretreated active power variation of the generator;

the identification module is used for carrying out order determination, identification and conversion on the preprocessed active power variation of the generator based on an identification model to obtain a frequency variation, and the identification model is constructed based on a red pool information quantity criterion and a system identification method;

and the calculation module is used for calculating an inertial time constant for the frequency variation by adopting a rocking equation and determining the inertial center frequency of the system according to the inertial time constant.

In one possible implementation, the first frequency characteristic includes a first maximum frequency transformation rate, a first transient frequency extremum, and a first quasi-steady state frequency, and the first prediction module may be specifically configured to:

In one possible implementation, the second frequency characteristic includes a second maximum frequency change rate, a second transient frequency extremum, and a second quasi-steady state frequency, and the second prediction module is specifically configured to:

Inputting mechanical power, electromagnetic power, damping coefficient, initial angular frequency and system inertial center frequency into a steady state prediction model to obtain a second frequency characteristic;

the training process of the steady state prediction model comprises the following steps:

In one possible implementation, the target frequency characteristic includes a target maximum frequency change rate, a target transient frequency extremum, and a target quasi-steady state frequency, and the fusion module may specifically be configured to:

inputting the first frequency characteristic and the second frequency characteristic into a fusion prediction model to obtain a target frequency characteristic;

the training process of the fusion prediction model may include:

In one possible implementation manner, the judging module may specifically be configured to:

Fig. 7 is a schematic diagram of a terminal provided in an embodiment of the present application. As shown in fig. 7, the terminal 7 of this embodiment includes: a processor 70, a memory 71, and a computer program 72 stored in the memory 71 and executable on the processor 70. The processor 70, when executing the computer program 72, implements the steps of the embodiments of the large grid frequency security situation prediction method described above, such as steps 101 to 105 shown in fig. 1. Alternatively, the processor 70, when executing the computer program 72, performs the functions of the modules of the apparatus embodiments described above, such as the functions of the modules 61 through 65 shown in fig. 6.

By way of example, the computer program 72 may be partitioned into one or more modules that are stored in the memory 71 and executed by the processor 70 to complete the present application. The one or more modules may be a series of computer program instruction segments capable of performing the specified functions, which instruction segments are used for describing the execution of the computer program 72 in the terminal 7. For example, the computer program 72 may be divided into modules 61 to 65 shown in fig. 6.

The terminal 7 may be a computing device such as a desktop computer, a notebook computer, a palm computer, a cloud server, etc. The terminal 7 may include, but is not limited to, a processor 70, a memory 71. It will be appreciated by those skilled in the art that fig. 7 is merely an example of the terminal 7 and is not limiting of the terminal 7, and may include more or fewer components than shown, or may combine some components, or different components, e.g., the terminal may further include input and output devices, network access devices, buses, etc.

The processor 70 may be a central processing unit (Central Processing Unit, CPU), or may be another general purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a Field-programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The memory 71 may be an internal storage unit of the terminal 7, such as a hard disk or a memory of the terminal 7. The memory 71 may be an external storage device of the terminal 7, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card) or the like, which are provided on the terminal 7. Further, the memory 71 may also include both an internal storage unit and an external storage device of the terminal 7. The memory 71 is used for storing the computer program as well as other programs and data required by the terminal. The memory 71 may also be used for temporarily storing data that has been output or is to be output.

It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-described division of the functional units and modules is illustrated, and in practical application, the above-described functional distribution may be performed by different functional units and modules according to needs, i.e. the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-described functions. The functional units and modules in the embodiment may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit, where the integrated units may be implemented in a form of hardware or a form of a software functional unit. In addition, specific names of the functional units and modules are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present application. The specific working process of the units and modules in the above system may refer to the corresponding process in the foregoing method embodiment, which is not described herein again.

In the foregoing embodiments, the descriptions of the embodiments are emphasized, and in part, not described or illustrated in any particular embodiment, reference is made to the related descriptions of other embodiments.

Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

In the embodiments provided in the present application, it should be understood that the disclosed apparatus/terminal and method may be implemented in other manners. For example, the apparatus/terminal embodiments described above are merely illustrative, e.g., the division of the modules or units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection via interfaces, devices or units, which may be in electrical, mechanical or other forms.

The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.

The integrated modules/units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the present application implements all or part of the flow of the method of the foregoing embodiment, or may be implemented by a computer program to instruct related hardware, where the computer program may be stored in a computer readable storage medium, where the computer program, when executed by a processor, may implement the steps of the foregoing embodiments of the method for predicting a security situation of a large grid frequency. Wherein the computer program comprises computer program code which may be in source code form, object code form, executable file or some intermediate form etc. The computer readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), an electrical carrier signal, a telecommunications signal, a software distribution medium, and so forth. It should be noted that the computer readable medium may include content that is subject to appropriate increases and decreases as required by jurisdictions in which such content is subject to legislation and patent practice, such as in certain jurisdictions in which such content is not included as electrical carrier signals and telecommunication signals.

The above embodiments are only for illustrating the technical solution of the present application, and are not limiting; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present application, and are intended to be included in the scope of the present application.

Claims

1. The method for predicting the frequency security situation of the large power grid is characterized by comprising the following steps of:

2. The method for predicting security situation of large grid frequency according to claim 1, wherein after obtaining the generator parameters, the method further comprises:

and calculating an inertial time constant for the frequency variation amount by adopting a rocking equation, and determining the inertial center frequency of the system according to the inertial time constant.

3. The method for predicting the frequency security situation of the large power grid according to claim 2, wherein the first frequency characteristic includes a first maximum frequency change rate, a first transient frequency extremum and a first quasi-steady state frequency, the feature extraction is performed on the generator parameter based on the dynamic prediction model, and the first frequency characteristic is obtained, and the method comprises the following steps:

And inputting the mechanical power, the electromagnetic power, the damping coefficient, the initial angular frequency and the system inertia center frequency into a dynamic prediction model to obtain a first frequency characteristic.

4. A method of predicting security situations of a large grid frequency as recited in claim 3, wherein the training process of the dynamic prediction model comprises:

taking the training sample as input, taking the corresponding feature of the training sample as output, and training the support vector regression;

and optimizing the test sample by adopting a particle swarm algorithm according to the trained support vector regression to obtain the dynamic prediction model.

5. The method for predicting a frequency security situation of a large power grid according to claim 2, wherein the second frequency characteristic includes a second maximum frequency change rate, a second transient frequency extremum and a second quasi-steady state frequency, the extracting the generator parameter based on the steady state prediction model to obtain the second frequency characteristic includes:

Inputting the mechanical power, the electromagnetic power, the damping coefficient, the initial angular frequency and the system inertial center frequency into a steady-state prediction model to obtain a second frequency characteristic;

dividing the mechanical power, the electromagnetic power, the damping coefficient, the initial angular frequency and the system inertial center frequency into a training set and a testing set;

inputting the training set into a transient frequency extremum prediction model, calculating the importance of a second feature by adopting a lightweight gradient elevator, deleting the frequency feature with the lowest score according to an attention mechanism, and obtaining a transient frequency extremum prediction model after feature processing;

inputting the training set into a quasi-steady state frequency prediction model, calculating a third feature importance degree by adopting a lightweight gradient elevator, deleting frequency features with the lowest scores according to an attention mechanism, and obtaining a quasi-steady state frequency prediction model after feature processing;

And verifying the test set by adopting a production countermeasure network according to the maximum frequency change rate prediction model, the transient frequency extremum prediction model and the quasi-steady state frequency prediction model after the characteristic processing to obtain the steady state prediction model.

6. The method for predicting the frequency security situation of the large power grid according to claim 1, wherein the target frequency characteristic includes a target maximum frequency change rate, a target transient frequency extremum and a target quasi-steady state frequency, the fusing the first frequency characteristic and the second frequency characteristic based on the fusion prediction model to obtain the target frequency characteristic includes:

inputting the first frequency characteristic and the second frequency characteristic into a fusion prediction model to obtain the target frequency characteristic;

the training process of the fusion prediction model comprises the following steps:

and fusing the first fuzzy rule and the second fuzzy rule based on a neural network to obtain the fused prediction model.

7. The method for predicting the frequency security situation of a large power grid according to claim 6, wherein the obtaining a prediction result according to the target frequency characteristic includes:

if the target maximum frequency change rate is not greater than a first preset value, the target transient frequency extremum is not greater than a second preset value, and the target quasi-steady state frequency is not greater than a third preset value, determining that the active frequency security situation of the large power grid is stable according to the target maximum frequency change rate, the target transient frequency extremum and the target quasi-steady state frequency;

8. The utility model provides a big electric wire netting frequency security situation prediction unit which characterized in that includes:

9. A terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of the large grid frequency security situation prediction method according to any one of the preceding claims 1 to 7 when the computer program is executed.

10. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the large grid frequency security situation prediction method according to any one of the preceding claims 1 to 7.