CN115496126A - Multi-energy microgrid state recognition method based on GRA (generalized grammes) relevance algorithm and SVM (support vector machine) - Google Patents
Multi-energy microgrid state recognition method based on GRA (generalized grammes) relevance algorithm and SVM (support vector machine) Download PDFInfo
- Publication number
- CN115496126A CN115496126A CN202210993171.3A CN202210993171A CN115496126A CN 115496126 A CN115496126 A CN 115496126A CN 202210993171 A CN202210993171 A CN 202210993171A CN 115496126 A CN115496126 A CN 115496126A
- Authority
- CN
- China
- Prior art keywords
- data
- aggregation
- equipment
- state
- gra
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
- H02J3/46—Controlling of the sharing of output between the generators, converters, or transformers
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application provides a multi-energy micro-grid state recognition method based on a GRA (generalized grammes) correlation algorithm and an SVM (support vector machine), which is characterized in that original output data of various devices in the multi-energy micro-grid and operation parameters of the devices are obtained; constructing the original output data of each type of equipment into time sequence data; determining typical states of various devices based on the time series data and a Kmeans aggregation algorithm; determining the correlation degree between the typical state of various types of equipment and the operating parameters thereof by using a GRA correlation degree algorithm; establishing a contact characteristic library between the operation states of various devices and the operation parameters thereof according to the association degree; and identifying the current operation state of the target equipment by utilizing an SVM classifier based on the contact feature library. According to the method, the parameter with the maximum correlation degree with the typical states of various devices in the multi-energy microgrid is calculated through a GRA algorithm, a contact feature library is formed, and the current running state of the target device is further identified more quickly and effectively through an SVM classifier.
Description
Technical Field
The application relates to the technical field of multi-energy micro-grid state recognition, in particular to a multi-energy micro-grid state recognition method based on a GRA (generalized GRA) correlation algorithm and an SVM (support vector machine).
Background
In order to improve the overall use efficiency of energy sources and the ability to consume renewable energy sources, the demand for the interconnection, fusion and complementary integration of various energy systems is increasingly urgent in recent years. The multi-energy microgrid consisting of a power system, a natural gas system, a thermodynamic system, energy conversion equipment serving as a coupling link and the like becomes a development trend in the energy field and becomes a necessary choice for building a clean, low-carbon, safe and efficient modern energy system. Along with the fact that the coupling degree of electric cold and heat in the multi-energy microgrid is continuously strengthened, the number of devices is continuously increased, and the requirements for optimal control of the devices are also continuously increased.
At present, certain achievements are formed in the aspects of research, development and application of multi-energy micro-grid data acquisition and control technology at home and abroad, but in a general view, the problems of insufficient systematization degree, immature product research and development and the like still exist in the existing research, and the practical application of multi-energy micro-grid interaction is difficult to support. Therefore, a high-efficiency and rapid identification method for states of each device in the multi-energy microgrid needs to be researched so as to provide a data base for optimization control and support coordination and complementation of electric cooling and heating devices in the multi-energy microgrid.
Disclosure of Invention
An object of the embodiment of the application is to provide a method for identifying the state of a multi-energy microgrid based on a GRA relevance algorithm and an SVM, so as to solve the problem of how to quickly and efficiently identify the current state of each device of the multi-energy microgrid. The specific technical scheme is as follows: in a first aspect, a method for identifying a state of a multi-energy microgrid based on a GRA association algorithm and a SVM is provided, and the method includes:
acquiring original output data of various devices in the multi-energy microgrid and operating parameters of the various devices;
constructing the original output data of each type of equipment into time sequence data;
determining typical states of various types of equipment based on the time series data and a Kmeans aggregation algorithm;
utilizing a GRA (Gray relationship Analysis) algorithm to determine the correlation degree between the typical states of various types of equipment and the operating parameters thereof;
establishing a contact feature library between the operation states of various devices and the operation parameters thereof according to the association degree;
and identifying the current operation state of the target equipment by utilizing an SVM (Support Vector Machine) classifier based on the contact feature library.
Optionally, the original output data is a grid QS (Quality Safety) file with a sampling interval of T, a number of N and a number of M sampling objects, and the constructing of the original output data of each type of device into time sequence data includes the following steps:
identifying attribution equipment of various types of original output data according to data header information in the QS file of the power grid;
carrying out name matching on the identified data attribution equipment and actual equipment so as to establish a corresponding relation between original output data and the actual equipment name;
based on the corresponding relation between the original output data and the actual equipment name, constructing a time sequence according to the sampling time of the original output data;
the constructed time sequence is expressed as a matrix S;
in the formula, v M,N The nth sampled data for the mth device.
Optionally, the determining the typical states of various types of devices based on the time-series data and the Kmeans aggregation algorithm includes the following steps:
determining a clustering number K value in time sequence data of various devices;
selecting K initial aggregation centers in the time series data based on the K values;
calculating Euclidean distances from each original output data in the time sequence data to K initial aggregation centers respectively, and classifying the Euclidean distances into a cluster corresponding to the initial aggregation center with the minimum Euclidean distance;
calculating the Euclidean distance from each original output data to the initial aggregation center based on the following formula:
in the formula (I), the compound is shown in the specification,for all sets of operational state data of the ith device,for the initial set of aggregation centers, p, of the ith device ij For the jth operating state data, p, of the ith device ik The output of the kth initial aggregation center of the ith equipment;
updating the aggregation center of each cluster in the newly composed cluster by the following formula;
in the formula u im An aggregation center for an mth cluster of the ith device; n is a radical of m The number of data included in the mth cluster, p il The first running state data in the mth cluster of the ith device;
performing iterative training until the average error criterion function is converged and clustering is finished to obtain a clustering sequence S of the time sequence S 1 And will S 1 Determining the state as a typical state of the equipment;
in the formula, N 2 The number of the clustered output data is,for the m device N after clustering 2 Individual historical force data.
Optionally, the determining a value of the clustering number K in the time-series data of each type of device includes the following steps:
determining a clustering number K value based on the following sum of squared errors formula:
in the formula: p is a radical of i Data are collected to represent the running state of the equipment i; m is a unit of ij Representing the aggregation centers of each class; and with the increase of K, the aggregation degree of each aggregation cluster is gradually increased, the error Sum of Squares (SSE) is gradually reduced, when K reaches the real aggregation number, the return of the aggregation degree is sharply reduced, the corresponding SSE is reduced to the maximum value, and the size of the aggregation number K can be determined.
Optionally, the determining the correlation between the typical states of various types of devices and the operating parameters thereof by using the GRA algorithm includes the following steps:
establishing a typical state sequence of various devices as a mother sequence;
the following formula is used to calculate the gray correlation coefficient of the typical state with each parameter:
in the formula, ζ i (k) Is the kth grey correlation coefficient of the ith parameter and the mother sequence, xi (k) is the kth data of the ith parameter, x0 (k) is the kth data of the mother sequence, rho is a resolution coefficient and takes a value between 0 and 1,taking the minimum value of the absolute value of the ith parameter and the mother sequence,taking the maximum value of the absolute value of the ith parameter and the mother sequence;
calculating a correlation coefficient average value according to the grey correlation coefficient;
and constructing a correlation sequence based on the average value of the correlation coefficients.
Optionally, the identifying the current operating state of the target device by using the SVM classifier based on the contact feature library includes the following steps:
acquiring real-time operation data of target equipment;
classifying the real-time operation data by using an SVM classifier;
and identifying the current operation state of the target equipment according to the classification result and the contact feature library.
Optionally, the classifying the real-time operation data by using the SVM classifier includes the following steps:
selecting a kernel function;
mapping the real-time operating data to a high-dimensional feature space using the kernel function;
mapping a high-dimensional feature space and then constructing a classification hyperplane;
solving the classification hyperplane to obtain an optimal discriminant function;
and classifying by using the optimal discriminant function.
Optionally, the method further comprises:
dividing the real-time operation data into a training set sample and a test set sample;
and testing the accuracy of the SVM classifier by using the test set samples.
Optionally, the SVM classifier employs a multi-classification SVM classifier.
The embodiment of the application has the following beneficial effects:
the embodiment of the application provides a method for identifying the state of a multi-energy micro-grid based on a GRA (generalized GRA) correlation algorithm and an SVM (support vector machine), wherein the method comprises the steps of acquiring original output data of various devices in the multi-energy micro-grid and operating parameters of the devices; constructing the original output data of each type of equipment into time sequence data; determining typical states of various devices based on the time series data and a Kmeans aggregation algorithm; determining the correlation degree between the typical state of various types of equipment and the operating parameters thereof by using a GRA algorithm; establishing a contact characteristic library between the operation states of various devices and the operation parameters thereof according to the association degree; and identifying the current operation state of the target equipment by utilizing an SVM classifier based on the contact feature library. According to the method and the device, the parameter with the maximum correlation degree with the typical states of various devices in the multi-energy micro grid is calculated through a GRA algorithm, a contact feature library is formed, and the current running state of the target device is further identified more quickly and effectively through an SVM classifier.
Of course, not all advantages described above need to be achieved at the same time in the practice of any one product or method of the present application.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious for those skilled in the art to obtain other drawings without inventive labor.
Fig. 1 is a flowchart of a method for identifying a state of a multi-energy microgrid based on a GRA association algorithm and an SVM according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without making any creative effort based on the embodiments in the present invention, belong to the protection scope of the present invention.
The embodiment of the application provides a method for recognizing a state of a multi-energy microgrid based on a GRA relevance algorithm and an SVM, and the method for recognizing a state of a multi-energy microgrid based on a GRA relevance algorithm and an SVM provided by the embodiment of the application is described in detail below with reference to specific implementation manners, and as shown in fig. 1, the specific steps are as follows:
step S101: and acquiring original output data of various devices in the multi-energy microgrid and operating parameters of the various devices.
In the embodiment of the present application, the various devices include a Combined Cooling and Heating and Power (CCHP) unit, a distributed photovoltaic Power plant, a gas boiler, an electric refrigerator, a thermoelectric energy storage device, and the like. The number of devices of each type is not limited.
Step S102: the raw output data for each type of device is constructed as time series data.
Optionally, the original output data is a power grid QS file with a sampling interval of T, a number of N and a number of sampling objects of M, and the constructing of the original output data of each type of device into time sequence data includes the following steps:
identifying attribution equipment of various types of original output data according to data header information in the QS file of the power grid;
name matching is carried out on the identified data attribution equipment and actual equipment so as to establish a corresponding relation between the original output data and the actual equipment name;
based on the corresponding relation between the original output data and the actual equipment name, constructing a time sequence according to the sampling time of the original output data;
the constructed time sequence is expressed as a matrix S;
in the formula, v M,N The nth sampled data for the mth device.
For the convenience of calculation in the subsequent steps, the data in the time series S may be normalized, for example, by using the following expression.
In the formula: s is i As an initial value of the operation state data of the ith device,the average value of the operation state data of the ith equipment is taken;and averaging the running state data after the normalization of the ith equipment.
Step S103: and determining typical states of various types of equipment based on the time series data and a Kmeans aggregation algorithm.
Optionally, the determining the typical states of various types of devices based on the time-series data and the Kmeans aggregation algorithm includes the following steps:
and determining the clustering number K value in the time sequence data of various devices.
In particular, the cluster number may be determined according to the elbow method, and the core indicator is the sum of the squares of the errors. The method comprises the following steps: determining a clustering number K value based on the following sum of squared errors formula:
in the formula: p is a radical of formula i Acquiring data representing the operating state of the equipment i; m is a unit of ij Representing the aggregation centers of each class; along with the increase of K, the aggregation degree of each aggregation cluster is gradually increased, the error Sum of Squares (SSE) is gradually reduced, when K reaches the real aggregation number, the return of the aggregation degree is sharply reduced, the corresponding SSE is reduced to the maximum value, and the size of the aggregation number K can be determined at the moment, similar to the elbow of a relation graph of the SSE and the K. Namely, the corresponding K value at the maximum value of SSE is determined as the cluster number.
Selecting K initial aggregation centers in the time series data based on the K values;
calculating Euclidean distances from each original output data in the time sequence data to K initial aggregation centers respectively, and enabling the Euclidean distances to be in a cluster corresponding to the initial aggregation center with the minimum Euclidean distance;
calculating the Euclidean distance from each original output data to the initial aggregation center based on the following formula:
in the formula (I), the compound is shown in the specification,for all running state data sets of the ith device,initial set of aggregation centers for the ith device, p ij For the jth operating status data, p, of the ith device ik The output of the kth initial aggregation center of the ith device;
updating the aggregation center of each cluster in the newly composed cluster by the following formula;
in the formula u im An aggregation center of an mth cluster of an ith device; n is a radical of hydrogen m Is the number of data contained in the mth cluster, p il The operation state data of the ith device in the mth cluster is obtained;
repeating the steps to carry out iterative training until the average error criterion function is converged and clustering is finished to obtain a clustering sequence S of the time sequence S 1 And then S is 1 Determining the device as a typical state of the device;
in the formula, N 2 The number of the clustered output data is,for the m device N after clustering 2 Individual historical force data.
In one example, if there are R-class devices, for example, then the typical state sequence for each class of device is obtained as S 1 、S 2 、…、S R 。
Step S104: the GRA algorithm is used to determine the correlation between the typical state of various types of equipment and its operating parameters.
Optionally, the determining the correlation between the typical states of various types of devices and the operating parameters thereof by using the GRA algorithm includes the following steps:
establishing a typical state sequence of various devices as a mother sequence;
the following formula is used to calculate the gray correlation coefficient of the typical state with each parameter:
in the formula, ζ i (k) Is the correlation coefficient of the ith parameter and the kth grey of the mother sequence, xi (k) is the kth data of the ith parameter, x0 (k) is the kth data of the mother sequence, rho is a resolution coefficient, the value is between 0 and 1,taking the minimum value of the absolute value of the ith parameter and the mother sequence,taking the maximum value of the absolute value of the ith parameter and the mother sequence;
calculating a correlation coefficient average value according to the grey correlation coefficient;
and constructing a correlation sequence based on the average value of the correlation coefficients. For example, the order may be by relevance size.
Step S105: and establishing a contact characteristic library between the operation states of various devices and the operation parameters thereof according to the association degree.
In this step, a link feature library between the operation states of various devices and the operation parameters thereof can be established through the operation parameters with the maximum correlation degree with the typical states.
Step S106: and identifying the current operation state of the target equipment by utilizing an SVM classifier based on the contact feature library.
Optionally, the identifying, based on the contact feature library, the current operating state of the target device by using the SVM classifier includes the following steps:
acquiring real-time operation data of target equipment;
classifying the real-time operation data by using an SVM classifier;
and identifying the current operation state of the target equipment according to the classification result and the contact feature library.
The corresponding relation between the operation parameters and the operation states is included in the contact characteristic library, and the corresponding operation states can be found quickly and efficiently based on the corresponding relation and the classified real-time operation data.
Optionally, the classifying the real-time operation data by using the SVM classifier includes the following steps:
a kernel function is selected.
In the embodiment of the application, a proper kernel function can be selected according to the data characteristics of the multi-energy microgrid device, and the selection of the kernel function has different degrees of influence on the final classification effect of the SVM.
The general kernel functions of the SVM are shown in table 1, and suitable kernel functions can be selected according to data characteristics.
TABLE 1SVM commonly used Kernel function
And mapping the real-time running data to a high-dimensional feature space by using the kernel function.
By selecting a suitable kernel function, the input vector is mapped from the input space to a high-dimensional feature space, so that the input data becomes linearly separable. By performing calculation according to a linear algorithm in a high-dimensional feature space, a linear optimal classification hyperplane can be determined in the space, and the hyperplane corresponds to a nonlinear optimal classification hyperplane in the original input data feature space. This approach can also be used for the case of linear inseparability, making the input data linearly separable.
And mapping the high-dimensional feature space and then constructing a classification hyperplane.
Solving the classification hyperplane to obtain an optimal discriminant function;
and classifying by using the optimal discriminant function.
In the embodiment of the present application, the optimal discriminant function is solved in two cases, one is a case where the input data is linearly separable, and the other is a case where the input data is linearly inseparable.
First, for the linear separable case, a classification hyperplane can be constructed as shown in the formula:
ω T x + b =0 formula (7)
In the formula, omega T And b is the normal vector and intercept of the hyperplane, respectively.
It can make the sample of two kinds of classifications in the training set correctly fall on this hyperplane both sides to it is correct to fall into two kinds with the sample:
in the formula, a binary variable y is equal to { -1,1} and is used as a learning target to represent a negative class and a positive class. When y is i When =1, represents x i Belong to a first category; when y is i When =1, represents x i Belonging to the second category. In the formula, ω T ·x i Representing the inner product of two vectors, plane omega T ·x i + b =1 and ω T ·x i + b = -1 is the boundary hyperplane in the classification problem.
The resulting decision function is:
f(x)=sgn(ω T x + b) formula (9)
For the solution of the optimal classification hyperplane, the distance between the two boundary hyperplanes is maximized. Because two boundaries exceed the plane omega T ·x i The distance from + b =1 to the origin is | b-1|/| | | ω | T ·x i The distance from + b = -1 to the origin is | b +1|/| | ω | | two boundary hyperplanes are parallel, so the distance between them is 2/| | ω |.
Solving the optimal classification hyperplane, i.e., maximizing 2/| ω |, is equivalent to minimizing | ω | | survival 2 2, solving can therefore translate to a quadratic programming problem as follows:
solving the linear programming problem by applying a Lagrange multiplier method, wherein the Lagrange expression of the problem is as follows:
in the formula, alpha i The Lagrange coefficient is more than or equal to 0. Let L (ω, b, α) take the partial differential of ω and b to 0, then:
the dual problem can be obtained according to Lagrange dual theory:
the solution of the dual problem can adopt a quadratic programming method to set a final optimal solutionThe corresponding optimum ω can be solved * And b * Comprises the following steps:
in the formula, x r And x s Respectively, a pair of support vectors in either of two categories. The optimal discriminant function that can be obtained for final classification is:
when linearity is not available, part of sample points fall between two boundary hyperplanes, and if the previous constraint condition is still met, the optimal solution cannot be obtained, so that a relaxation variable xi needs to be introduced i I =1,2, \8230;, n. For outliers that fall within the interval, the constraint is satisfied by the corresponding slack variable, while for points on either side of the interval the corresponding slack variable is 0, which is equivalent to a superordinate classification of the approximationAnd solving the plane, wherein partial training samples are allowed to be positioned in the interval, and the boundary hyperplane interval in the two modes is maximized as much as possible.
For the linear inseparable condition, solving the quadratic programming problem of the optimal classification hyperplane equivalence is as follows:
wherein c is a penalty function used for controlling the scale of the external points and suppressing the noise data, and the larger the value of the penalty function is, the larger the penalty of the model for the classification error is. The problem solving method is the same as the linear time-divisible solving method described above.
Selecting a kernel function phi (-) and mapping the training sample to a high-dimensional feature space, and applying the linear time-sharing calculation method to obtain a corresponding optimal discriminant function as follows:
when the linear inseparability occurs after the kernel function is adopted to map the data of each device of the multi-energy microgrid to a high-dimensional feature space, the processing method is the same as the linear inseparability in the input space, and the solution can also be realized by introducing a relaxation variable.
Optionally, the method further comprises:
dividing the real-time operation data into a training set sample and a test set sample;
and testing the accuracy of the SVM classifier by using the test set samples. So as to select a more suitable kernel function and train the SVM classifier.
In one example, the recognition accuracy accurve calculation formula is:
in the formula, n is the lumped sample number of the test samples, and n' is the correctly identified sample number.
On the basis, the calculation formula of the average accuracy rate avg _ accuracy of the identification classification result evaluation index is as follows:
in the formula, K represents the number of classes in the test sample, and the average accuracy rate is the average value of the identification accuracy rates of the samples in all the classes.
Optionally, the SVM classifier employs a multi-classification SVM classifier. And realizing the state recognition of the multiple types of resources in the multi-energy microgrid by constructing a multi-classification SVM.
At present, methods for constructing a multi-classification SVM are mainly divided into two methods, one method is to adopt an optimal method to directly combine and solve parameters of a plurality of optimal classification hyperplanes in the multi-classification SVM. The second method is indirect, which is to implement multi-classification SVM by combining a plurality of two-classification SVMs with each other, and there are typically one-to-one and one-to-many methods.
1) One-to-one method: the method is based on labels of all sample points in input training samples, two-classification SVM is trained between every two training samples of the classes, and finally, the two-classification SVM are combined together to form the multi-classification SVM. When multiple classes exist in the training sample, multiple two-class SVMs need to be constructed. In practical application, each two-classification SVM classifies sample data every time the sample data is input, which is equivalent to a voting mechanism, and the number of votes obtained in which class is the largest, namely the number of times of classification in which class is the largest after sub-classification, is the final sample class.
2) One-to-many method: the one-to-many method is different from the one-to-one method in that the SVM is constructed between every two categories, and a two-classification SVM classifier is constructed by training between each category and all the rest samples. When a plurality of classes exist in the training sample, only a plurality of two-classification SVM classifiers need to be constructed. In practical application, each SVM classifier is used for classifying the input samples, and the classification result corresponding to the maximum decision function value is the final class of the input samples.
And selecting a proper method to construct the multi-classification SVM classifier according to the actual data requirements of the multi-energy microgrid.
It should be understood that parts of the specification not set forth in detail are well within the prior art.
It is noted that, in this document, relational terms such as "first" and "second," and the like, are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrases "comprising a," "8230," "8230," or "comprising" does not exclude the presence of additional like elements in a process, method, article, or apparatus that comprises the element.
The above description is merely exemplary of the present application and is presented to enable those skilled in the art to understand and practice the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (9)
1. A multi-energy microgrid state recognition method based on a GRA correlation algorithm and an SVM is characterized by comprising the following steps:
acquiring original output data of various devices in the multi-energy microgrid and operating parameters of the various devices;
constructing the original output data of each type of equipment into time sequence data;
determining typical states of various types of equipment based on the time series data and a Kmeans aggregation algorithm;
determining the correlation degree between the typical state of various types of equipment and the operating parameters thereof by using a GRA algorithm;
establishing a contact feature library between the operation states of various devices and the operation parameters thereof according to the association degree;
and identifying the current operation state of the target equipment by utilizing an SVM classifier based on the contact feature library.
2. The method of claim 1, wherein the raw output data is a QS file of a power grid with a sampling interval of T, a number of N and a number of M sampling objects, and the constructing the raw output data of each type of device into time-series data comprises:
identifying attribution equipment of various types of original output data according to data header information in the QS file of the power grid;
name matching is carried out on the identified data attribution equipment and actual equipment so as to establish a corresponding relation between the original output data and the actual equipment name;
based on the corresponding relation between the original output data and the actual equipment name, constructing a time sequence according to the sampling time of the original output data;
the constructed time sequence is expressed as a matrix S;
in the formula, v M,N The nth sampled data for the mth device.
3. The method of claim 2, wherein the determining the typical states of various types of devices based on the time series data and the Kmeans aggregation algorithm comprises: determining a clustering number K value in time sequence data of various devices;
selecting K initial aggregation centers in the time series data based on the K values;
calculating Euclidean distances from each original output data in the time sequence data to K initial aggregation centers respectively, and enabling the Euclidean distances to be in a cluster corresponding to the initial aggregation center with the minimum Euclidean distance;
calculating the Euclidean distance from each original output data to the initial aggregation center based on the following formula:
in the formula (I), the compound is shown in the specification,for all sets of operational state data of the ith device,for the initial set of aggregation centers, p, of the ith device ij For the jth operating status data, p, of the ith device ik The output of the kth initial aggregation center of the ith equipment;
in the newly composed clusters, updating the aggregation center of each cluster by the following formula;
in the formula u im An aggregation center for an mth cluster of the ith device; n is a radical of hydrogen m The number of data included in the mth cluster, p il The first running state data in the mth cluster of the ith device;
performing iterative training until the average error criterion function is converged and clustering is finished to obtain a clustering sequence S of the time sequence S 1 And will S 1 Determining the device as a typical state of the device;
4. The method of claim 3, wherein determining the K value of the cluster number in the time-series data of each type of device comprises:
determining a clustering number K value based on the following sum of squared errors formula:
in the formula: p is a radical of i Acquiring data representing the operating state of the equipment i; m is ij Representing the aggregation centers of each class; and with the increase of K, the aggregation degree of each aggregation cluster is gradually increased, the error Sum of Squares (SSE) is gradually reduced, when K reaches the real aggregation number, the return of the aggregation degree is sharply reduced, the corresponding SSE is reduced to the maximum value, and the size of the aggregation number K can be determined.
5. The method for recognizing the state of the multi-energy microgrid based on a GRA (GRA correlation) algorithm and an SVM (support vector machine) as claimed in claim 1, wherein the step of determining the correlation between the typical states of various types of equipment and the operating parameters thereof by using the GRA algorithm comprises the following steps:
establishing a typical state sequence of various devices as a mother sequence;
the following formula is used to calculate the gray correlation coefficient of the typical state with each parameter:
in the formula, ζ i (k) Is the kth grey correlation coefficient of the ith parameter and the mother sequence, xi (k) is the kth data of the ith parameter, x0 (k) is the kth data of the mother sequence, rho is a resolution coefficient and takes a value between 0 and 1,taking the minimum value of the absolute value of the ith parameter and the mother sequence,taking the maximum value of the absolute value of the ith parameter and the mother sequence;
calculating a correlation coefficient average value according to the grey correlation coefficient;
and constructing a correlation sequence based on the average value of the correlation coefficients.
6. The method as claimed in claim 1, wherein the step of identifying the current operating state of the target device by using an SVM classifier based on the contact feature library comprises the following steps:
acquiring real-time operation data of target equipment;
classifying the real-time operation data by using an SVM classifier;
and identifying the current operation state of the target equipment according to the classification result and the contact feature library.
7. The GRA correlation algorithm and SVM based multi-energy microgrid state recognition method of claim 6, wherein the classification of the real-time operation data by an SVM classifier comprises the steps of:
selecting a kernel function;
mapping the real-time operating data to a high-dimensional feature space using the kernel function;
mapping a high-dimensional feature space and then constructing a classification hyperplane;
solving the classification hyperplane to obtain an optimal discriminant function;
and classifying by using the optimal discriminant function.
8. The method of claim 6, wherein the method further comprises:
dividing the real-time operation data into a training set sample and a test set sample;
and testing the accuracy of the SVM classifier by using the test set samples.
9. The GRA relevancy algorithm and SVM based multi-energy microgrid state recognition method of claim 6, wherein the SVM classifier employs a multi-classification SVM classifier.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210993171.3A CN115496126A (en) | 2022-08-18 | 2022-08-18 | Multi-energy microgrid state recognition method based on GRA (generalized grammes) relevance algorithm and SVM (support vector machine) |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210993171.3A CN115496126A (en) | 2022-08-18 | 2022-08-18 | Multi-energy microgrid state recognition method based on GRA (generalized grammes) relevance algorithm and SVM (support vector machine) |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115496126A true CN115496126A (en) | 2022-12-20 |
Family
ID=84465880
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210993171.3A Pending CN115496126A (en) | 2022-08-18 | 2022-08-18 | Multi-energy microgrid state recognition method based on GRA (generalized grammes) relevance algorithm and SVM (support vector machine) |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115496126A (en) |
-
2022
- 2022-08-18 CN CN202210993171.3A patent/CN115496126A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107766990B (en) | Method for predicting power generation power of photovoltaic power station | |
CN106845717B (en) | Energy efficiency evaluation method based on multi-model fusion strategy | |
CN109325607A (en) | A kind of short-term wind power forecast method and system | |
CN112395289B (en) | Distributed photovoltaic data layered storage method and system | |
CN117034179B (en) | Abnormal electric quantity identification and tracing method and system based on graph neural network | |
CN113988215B (en) | Power distribution network metering cabinet state detection method and system | |
Yoo et al. | Attention-based autoregression for accurate and efficient multivariate time series forecasting | |
CN111428766B (en) | Power consumption mode classification method for high-dimensional mass measurement data | |
Kumar et al. | Cloud-based electricity consumption analysis using neural network | |
CN115204698A (en) | Real-time analysis method for power supply stability of low-voltage transformer area | |
CN111324790A (en) | Load type identification method based on support vector machine classification | |
Miraftabzadeh et al. | Knowledge Extraction from PV Power Generation with Deep Learning Autoencoder and Clustering-Based Algorithms | |
CN117493922A (en) | Power distribution network household transformer relation identification method based on data driving | |
CN116307111A (en) | Reactive load prediction method based on K-means clustering and random forest algorithm | |
Rouwhorst et al. | Improving clustering-based forecasting of aggregated distribution transformer loadings with gradient boosting and feature selection | |
CN115496126A (en) | Multi-energy microgrid state recognition method based on GRA (generalized grammes) relevance algorithm and SVM (support vector machine) | |
CN115481788B (en) | Phase change energy storage system load prediction method and system | |
Liu et al. | Embedded scenario clustering for wind and photovoltaic power, and load based on multi-head self-attention | |
CN113487194B (en) | Electric power system dispatcher grade evaluation system based on text classification | |
CN115149528A (en) | Intelligent electric energy meter distributed prediction method based on big data non-intrusive technology | |
Zheng et al. | [Retracted] Application Based on Artificial Intelligence in Substation Operation and Maintenance Management | |
Kou et al. | Machine learning based models for fault detection in automatic meter reading systems | |
Chen et al. | Consumer-transformer relationship identification based on UMAP and mini batch K-means | |
Liang et al. | KNN_SVM based information transmission method for renewable energy power generation | |
Li et al. | Residential Photovoltaic Power Forecasting Considering Division of Weather Type Index Interval |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |