CN111160386A - Data-driven power island detection method based on active reactive power disturbance - Google Patents
Data-driven power island detection method based on active reactive power disturbance Download PDFInfo
- Publication number
- CN111160386A CN111160386A CN201911189108.9A CN201911189108A CN111160386A CN 111160386 A CN111160386 A CN 111160386A CN 201911189108 A CN201911189108 A CN 201911189108A CN 111160386 A CN111160386 A CN 111160386A
- Authority
- CN
- China
- Prior art keywords
- reactive power
- neural network
- back propagation
- data
- island
- 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.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 46
- 238000013528 artificial neural network Methods 0.000 claims abstract description 62
- 238000012549 training Methods 0.000 claims abstract description 33
- 238000012706 support-vector machine Methods 0.000 claims abstract description 28
- 230000004927 fusion Effects 0.000 claims abstract description 26
- 230000001537 neural effect Effects 0.000 claims abstract description 6
- 239000002245 particle Substances 0.000 claims description 36
- 238000000034 method Methods 0.000 claims description 23
- 230000003044 adaptive effect Effects 0.000 claims description 21
- 238000010248 power generation Methods 0.000 claims description 18
- 238000004364 calculation method Methods 0.000 claims description 12
- 238000005457 optimization Methods 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 11
- 230000007246 mechanism Effects 0.000 abstract description 5
- 230000006870 function Effects 0.000 description 16
- 238000004891 communication Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R21/00—Arrangements for measuring electric power or power factor
- G01R21/001—Measuring real or reactive component; Measuring apparent energy
- G01R21/003—Measuring reactive component
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Probability & Statistics with Applications (AREA)
- Power Engineering (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a data driving power island detection method based on active reactive power disturbance, which comprises the following steps: firstly, carrying out reactive disturbance on a power-controlled single-phase grid-connected inverter by using pre-designed intermittent reactive power to obtain a corresponding data set under the condition that a pseudo-island effect exists; then, performing mechanism analysis on the single-phase inverter to obtain a characteristic data set; combining a k-means algorithm with logic operation to obtain fusion data, training a self-adaptive back propagation neural network by using a characteristic data set and the fusion data, taking the output of the self-adaptive back propagation neural network as a support vector machine to obtain a training set, training the support vector machine, and finally obtaining the self-adaptive back propagation neural network-support vector machine (ABPS) classifier. The island detection method solves the problem of high detection cost of the traditional island detection method, and additional large disturbance and large undetectable area are not required to be applied; and technical references are provided for accurate island detection, subsequent island effect utilization and the like.
Description
Technical Field
The invention belongs to the field of power island effect detection, and particularly relates to a data driving power island detection method based on active reactive power disturbance.
Background
Renewable energy sources such as photovoltaic energy, wind energy, geothermal energy, wave energy and the like are increasingly widely applied to distributed power generation, and provide electric energy for daily life of people. In addition, distributed power generation systems that can supply power to local loads and feed the power back to the distribution grid are taking an increasingly important position in the power industry. However, distributed power generation systems still face many challenges in practical applications, such as anti-islanding protection issues. Islanding refers to the situation that a power grid trips due to failure or maintenance, and a distributed power generation system cannot detect power failure in time. Thus, islanding will result in an uncontrollable self-powered system consisting of a distributed power generation system and local loads. The island may endanger the safety of maintenance personnel, the service life of an electric power system and an electric power facility and the like, so that the research on the solution of island protection and the reduction of island influence have very important significance.
In the prior art, anti-islanding protection methods can be divided into three main categories, namely passive methods, communication-based methods and active methods. Passive island detection methods generally rely on measuring key parameters such as voltage, frequency, phase, etc. to determine whether an island event has occurred. If the parameter exceeds a predefined threshold, islanding is detected using over/under voltage protection, over/under frequency protection (OFP/UFP), and phase jump protection. The voltage threshold is typically set to 85% -110% of normal, while the frequency threshold is set to 49.5Hz 50.5Hz for a 50Hz system. There are other passive methods such as those based on power factor variation and voltage harmonic distortion. The passive detection method has the advantages of no harm to the quality of electric energy and easy realization. However, larger non-detection regions are formed. NDZ is defined as a series of load parameters for which the anti-islanding method cannot detect islanding. The detection method based on communication has no harm to the quality of electric energy, but the former has no NDZ, so the detection method is superior to a passive detection method. However, the communication-based approach is quite costly and therefore not widely applicable, since peripheral devices are required to detect islanding conditions.
Disclosure of Invention
The purpose of the invention is as follows: in order to solve the problems of high cost, large detection error and the like in the prior art, the invention provides a data driving power island detection method based on active reactive power disturbance.
The technical scheme is as follows: the invention provides a data driving electric power island detection method based on active reactive power disturbance, which is used for judging whether an island exists in a distributed power generation system and specifically comprises the following steps:
step 1: calculating to obtain reactive power injected into an inverter according to the common point frequency output by a phase-locked loop end in the distributed power generation system; taking the common point frequency and the corresponding reactive power injected into the inverter as a group of characteristic data sets; collecting the frequency of a common point and corresponding reactive power to obtain n groups of characteristic data sets;
step 2: sequentially carrying out data fusion on the n groups of characteristic data sets to obtain n groups of fusion data;
and step 3: calculating the weight and the threshold of the self-adaptive back propagation neural network by utilizing a particle swarm algorithm based on n groups of fusion data; meanwhile, in each particle swarm iterative computation, the n groups of characteristic data sets in the step 1 are used as a training set of the self-adaptive back propagation neural network to obtain a trained self-adaptive back propagation neural network;
and 4, step 4: taking the output of the trained self-adaptive back propagation neural network as the input of a support vector machine, and training the support vector machine to obtain a decision function; finally, the establishment of a training self-adaptive back propagation neural network-support vector machine classifier is completed;
and 5: changing a common point frequency fluctuation mode and a PLC load of an inverter to obtain a group of new characteristic data sets, and classifying the group of new characteristic data sets by using a self-adaptive back propagation neural network-support vector machine classifier so as to judge whether an island exists in the power system, wherein the method specifically comprises the following steps: and taking the new test data as a training set of the trained adaptive back propagation neural network to obtain the output of the adaptive back propagation neural network, and calculating whether the output is island data or not by using a decision function so as to judge whether the power generation system is in an island or not.
Further, a specific method for calculating reactive power injected into the inverter in step 1 is as follows:
where f' represents the common point frequency of the phase-locked loop output, frFor the resonant frequency, Q, of the inverter RLC loadfIs the quality factor, Q, of the inverterDGAnd PDGRepresenting reactive power and active power injected into the inverter, respectively.
Further, a specific method for sequentially performing data fusion on the n groups of feature data sets in step 2 is as follows:
carrying out k-means calculation on the common point frequency in the n groups of characteristic data sets and the reactive power injected into the inverter to obtain a classification result of each common point frequency and the reactive power injected into the inverter; and performing numerical value distribution on each classification result;
performing data fusion on the D-th group of feature data sets by using the following formula, wherein D is 1,2, … n;
L=ClusterR∧Clusterf
wherein ClusterRThe value is obtained after the numerical value distribution is carried out on the classification result of the reactive power; clusterfThe value is a value obtained by carrying out numerical value distribution on the classification result of the public point frequency; when L is {0,1}, and is 0, it indicates that the power system under the set of characteristic data sets is not islanded; a time of 1 represents the power under the set of feature data setsThe system is islanded.
Further, classifying the reactive power data set injected into the inverter by using a k-means algorithm, wherein the classification result is two types, and the first type is as follows: the reactive power injected into the inverter is not in the intermittent period; the second type is: the reactive power injected is in the intermittent period; according to the value of the common point frequency when the island occurs, classifying the common point frequency data set by using a k-means algorithm, wherein the classification result is three types, and the first type is as follows: common point frequency when islanding does not occur; the second type is: islanding occurs and reactive power injected into the inverter is at a common point frequency of the intermittent period; the third type is: islanding occurs and reactive power injected into the inverter is not at a common point frequency of an intermittent time period; the values of L after the numerical assignment based on the above classification results are shown in table 1 below:
TABLE 1
The first type of reactive power is 0 after being subjected to numerical value distribution, and the second type of reactive power is 1 after being subjected to numerical value distribution; the first class of common point frequencies is numerically assigned a value of 0, the second class is numerically assigned a value of 1, and the third class is numerically assigned a value of 2.
Further, the specific steps of step 3 are as follows:
step 3.1: initializing parameters in a self-adaptive back propagation neural network and a particle swarm algorithm, wherein the size of a particle swarm is M;
step 3.2, carrying out the iterative computation for the t time, and sequentially taking the initial particles as the weight and the threshold of the self-adaptive back propagation neural network, wherein M self-adaptive back propagation neural networks exist;
step 3.3: taking n groups of characteristic data sets as training sets, sequentially inputting the training sets into an mth self-adaptive back propagation neural network, and carrying out epochs training on the n groups of characteristic data sets by the neural network to obtain n outputs;
step 3.4: making a mean square error between the Nth output and the Nth fusion data, wherein N is 1,2, …, N; obtaining n groups of mean square deviations, and taking the sum of the n groups of mean square deviations as a group of fitness values;
step 3.5, judging whether M is larger than M, if so, turning to step 3.6; if not, m +1, turning to step 3.3;
step 3.6: comparing the M groups of fitness values, and judging whether the fitness values smaller than Y exist in the M groups of fitness values, wherein Y is a preset numerical value; if yes, stopping the particle swarm algorithm, selecting a group of particles corresponding to the minimum fitness value from the fitness value group smaller than Y as a global optimal solution, and taking the global optimal solution as a weight and a threshold of the adaptive back propagation neural network to obtain the adaptive back propagation neural network with the determined weight and threshold, namely obtaining the trained adaptive back propagation neural network; if not, judging whether t reaches the maximum iteration times; if not, taking a group of corresponding particles with the minimum fitness value in the iterative calculation as the individual optimal solution of the iterative calculation, comparing the individual optimal solution with the historical global optimal solution, selecting the group of corresponding particles with the minimum fitness value as the global optimal solution of the iterative calculation, and turning to the step 3.7; if so, stopping iterative computation, and selecting a global optimal solution as a weight value and a threshold value of the self-adaptive back propagation neural network to obtain the self-adaptive back propagation neural network with the determined weight value and threshold value;
and 3.7, updating the speed and the position of each particle according to the global optimal solution and the individual optimal solution of the iterative computation, taking the updated particles as initial particles, t +1, and turning to the step 3.2.
Further, the specific method of step 4 is as follows:
outputting x of the trained self-adaptive back propagation neural networki(i ═ 1,2, … n) as training data for the support vector machine and establish a hyperplane, describing the decision function as:
y(xi)=ωTφ(xi)+b
where ω represents the normal vector of the hyperplane, b represents the deviation,for finding x for mapping functionsiInner product of (d);
if the decision function satisfies the conditionX is theniThe corresponding group of characteristic data is non-island data, which indicates that the power generation system under the data does not have an island; if the decision function satisfies the conditionX is theniThe corresponding group of characteristic data is island data, which indicates that the power generation system under the data has an island;
based on yiE { -1, 1} with each xiAnd (3) associating, initializing a regularization parameter C, and expressing the hyperplane as a quadratic programming optimization problem:
ξ thereini≧ 0 denotes a group represented by xiA corresponding slack variable;
and (3) adopting a radial basis function as a kernel function of the training data, solving the quadratic programming optimization problem by utilizing a Lagrangian function, and rewriting a final decision function into:
y(xi)=ω*K(xi,x)+b*
wherein x ═ x1,x2,…,xn]ω, b is an optimum parameter, obtained from the KKT condition.
Has the advantages that: the method can well solve the problems of overlarge active disturbance, overlarge undetected areas and overhigh detection cost in the island detection of the grid-connected inverter of the power generation system, does not need to additionally apply large disturbance, does not depend on the traditional island detection threshold value, can obtain initial data for island judgment through data fusion, and finally further obtains island and non-island states through a support vector machine, thereby achieving the purpose of accurately detecting the island effect.
Drawings
FIG. 1 is an overall flow chart of the present invention;
FIG. 2 is a power system grid-connected inverter topology;
FIG. 3 is a common point frequency response of grid fluctuations, islanding and reactive power disturbances;
FIG. 4 is a schematic diagram of a back propagation neural network;
fig. 5 is a clustering result with respect to common point frequencies obtained using the k-means method.
Fig. 6 is a clustering result on reactive power injected into an inverter obtained by using a k-means method.
FIG. 7, where (a) is the assigned value for the clustered common point frequencies and (b) is the assigned value for the clustered reactive power disturbances;
fig. 8 is a result of logical operations on n sets of feature data sets.
Fig. 9, where (a) is the common point frequency collected under case 1,2 and (b) is the reactive power injected into the inverter collected under case 1, 2;
fig. 10, wherein (a) is a result of islanding detection for case 1 using only the support vector machine, and wherein (b) is a result of islanding detection for case 2 using only the support vector machine;
fig. 11, wherein (a) is a result of performing islanding detection on case 1 using the present invention, and wherein (b) is a result of performing islanding detection on case 2 using the present invention.
Detailed Description
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate an embodiment of the invention and, together with the description, serve to explain the invention and not to limit the invention.
The embodiment shown in fig. 1 provides a data-driven power island detection method based on active reactive power disturbance, which specifically includes the following steps:
(1) defining corresponding characteristic variables according to an inverter island generation mechanism, wherein a common point frequency output by a phase-locked loop and modulated reactive power injected into an inverter are selected as inverter observation variables, and the modulated reactive power is reactive power obtained according to the common point frequency; the method comprises the steps of collecting a common point frequency and corresponding reactive power disturbance (reactive power injected into an inverter), and using the common point frequency and the corresponding reactive power disturbance as a group of characteristic data sets to obtain n groups of characteristic data sets.
(2) Through mechanism analysis, it can be known that reactive power injected into an inverter has correlation with inverter common point frequency when an island occurs, therefore, in order to avoid false island false detection phenomenon caused by frequency fluctuation, n groups of characteristic data sets can be sequentially calculated through a logic operation, and thus n groups of fusion data used for training an Adaptive Back Propagation Neural Network (ABPNN) are obtained.
(3) And (3) combining a particle swarm algorithm, taking the n groups of characteristic data sets in the step (1) and the n groups of fusion data in the step (2) as a training set of the neural network, training the adaptive back propagation neural network to obtain the trained adaptive back propagation neural network, wherein the neural network can obtain fusion data used for testing subsequently, and a data fusion model based on the adaptive back propagation neural network is obtained.
(4) Establishing an island fault detection model based on a Support Vector Machine (SVM): for the island effect, two states, namely an island state and a non-island state, can be generated in the running process of the inverter, a Support Vector Machine (SVM) is adopted to classify the two states, and the output of a trained self-adaptive back propagation neural network is used as SVM training data; and finally obtaining the self-adaptive back propagation neural network-support vector machine (ABPS) classifier.
(5) A common point frequency fluctuation mode and RLC loads are changed through the constructed single-phase inverter to obtain a group of new characteristic data sets, an ABPS classifier is used for carrying out island and non-island state classification on the new characteristic data sets, and finally an island detection result is obtained under the condition that a pseudo-island effect exists.
As shown in fig. 2, in the single-phase grid-connected inverter built by Matlab/Simulink according to this embodiment, the parameters of the island test system obtained by the operation of the inverter are shown in table 1, and n sets of characteristic data sets simulating the change of the common point frequency (common point frequency output by the phase-locked loop) of the grid-connected inverter before and after the occurrence of the island effect and the reactive power injected into the inverter are also obtained.
TABLE 1 island test System parameters
In order to screen characteristic variables capable of well describing grid-connected inverter island changes, the mechanism analysis of the changes of related variables caused by the reactive power disturbance of an inverter is as follows:
wherein f' represents the frequency of the common point of the grid-connected inverter, frFor the resonant frequency, Q, of the inverter RLC loadfIs the quality factor, Q, of the inverterDGAnd PDGRepresenting reactive power and active power injected into the inverter, respectively. Because the power control of the inverter keeps active power, and the resonant frequency and the quality factor are constant values, a correlation exists between the common point frequency f' and reactive power injected into the inverter, and finally the common point frequency of the parallel inverter and the input reactive power are selected as characteristics;
as shown in fig. 2, a triangular intermittent reactive power disturbance (reactive power injected into the inverter) is used to apply interference to the grid-connected inverter, a corresponding common point frequency value is obtained from a phase-locked loop at the PCC, and the reactive power injected into the inverter is obtained through power detection of a control loop.
As shown in fig. 3, the frequency value at the common point may be affected by the grid voltage fluctuation and exceed the conventional detection threshold, which may cause false detection of the conventional islanding detection method. When the islanding effect occurs, the frequency at the common point firstly reaches the resonance frequency of the inverter and then changes along with the change of the injected reactive power.
According to the analysis, the frequency of the common point and the reactive power injected into the inverter are finally selected as characteristic variables.
Through mechanism analysis, it can be known that the reactive power injected into the inverter has correlation with the inverter common point frequency when an island occurs, and therefore, in order to avoid the false island false detection phenomenon caused by frequency fluctuation, data fusion can be performed on two groups of data sets through logical operation:
carrying out k-means calculation on the common point frequency in the n groups of characteristic data sets and the reactive power injected into the inverter to obtain a classification result of each common point frequency and the reactive power injected into the inverter; and performing numerical value distribution on each classification result;
performing data fusion on the D-th group of feature data sets by using the following formula, wherein D is 1,2, … n;
L=ClusterR∧Clusterf
wherein ClusterRThe value is obtained after the numerical value distribution is carried out on the classification result of the reactive power; clusterfThe value is a value obtained by carrying out numerical value distribution on the classification result of the public point frequency; when L is {0,1}, and is 0, it indicates that the power system under the set of characteristic data sets is not islanded; the case 1 indicates that the power system under the set of characteristic data is islanded.
FIG. 5 shows the result of k-means clustering the frequency of the common points: common point frequencies, can be divided into three categories, first category: frequency when islanding does not occur, second class: islanding and reactive power injected into the inverter is at a common point frequency of the intermittent period, and the third type: islanding occurs and reactive power injected into the inverter is not at a common point frequency of an intermittent time period; cluster 1 represents the effect when no islanding occurs and the influence of the power grid interference in the downlink section, cluster 2 represents that the category includes the influence of the resonance frequency on the common point frequency when islanding and the influence of the power grid uplink section interference, and cluster 3 represents that the category includes the influence of the power grid uplink section interference and the influence of the injected reactive power on the common point frequency.
FIG. 6 shows the result of the clustering analysis of the reactive power by k-means: the first type is: the reactive power injected into the inverter is not in the intermittent period; the second type is: the reactive power injected is in the intermittent period; when no island occurs, namely the frequency of a common point is 50Hz, the injected reactive power is triangular reactive power (the waveform is triangular wave); when the islanding occurs, the common point frequency is firstly clamped to the resonance frequency of the inverter, so that the reactive power injected into the inverter is smaller than that injected when the islanding does not occur; when the injected reactive power is in the intermittent period, the value of the injected reactive power is 0. In the figure, cluster 1 represents an intermittent period, and cluster 2 represents a disturbance period.
Fig. 7 shows, where (a) is the assigned value of the clustered common point frequency, it can be seen that the value of the first class is assigned to 0; the value of the second type after being distributed is 1, and the value of the third type after being distributed is 2;
(b) the distributed value of the reactive power injected into the inverter after clustering is known, the distributed value of the first type of value is 0, and the distributed value of the second type of value is 1;
fig. 8 shows the result of data fusion performed on n sets of feature data sets, where the value of L is shown in table 2,
TABLE 2
The specific method for establishing the data fusion model based on the adaptive back propagation neural network in the step 3 comprises the following steps: and (3) importing a training data set (n groups of characteristic data sets) and a logic operation output into Matlab 2018b, and simultaneously calling a self-adaptive back propagation neural network program to train data fusion model parameters of the neural network. As shown in fig. 4, the input data is two-dimensional frequency and reactive power, and the output data is one-dimensional fusion data. The corresponding operation is as follows:
step 3.1: initializing parameters in a self-adaptive back propagation neural network and a particle swarm algorithm, wherein the size of a particle swarm is M;
step 3.2, carrying out the iterative computation for the t time, and sequentially taking the initial particles as the weight and the threshold of the self-adaptive back propagation neural network, wherein M self-adaptive back propagation neural networks exist;
step 3.3: taking n groups of characteristic data sets as training sets, sequentially inputting the training sets into an mth self-adaptive back propagation neural network, and carrying out epochs training on the n groups of characteristic data sets by the neural network to obtain n outputs;
step 3.4: making a mean square error between the Nth output and the Nth fusion data, wherein N is 1,2, …, N; obtaining n groups of mean square deviations, and taking the sum of the n groups of mean square deviations as a group of fitness values;
step 3.5, judging whether M is larger than M, if so, turning to step 3.6; if not, m +1, turning to step 3.3;
step 3.6: comparing the M groups of fitness values, and judging whether the fitness values smaller than Y exist in the M groups of fitness values, wherein Y is a preset numerical value; if yes, stopping the particle swarm algorithm, selecting a group of particles corresponding to the minimum fitness value from the fitness value group smaller than Y as a global optimal solution, and taking the global optimal solution as a weight and a threshold of the adaptive back propagation neural network to obtain the adaptive back propagation neural network with the determined weight and threshold, namely obtaining the trained adaptive back propagation neural network; if not, judging whether t reaches the maximum iteration times; if not, taking a group of corresponding particles with the minimum fitness value in the iterative calculation as the individual optimal solution of the iterative calculation, comparing the individual optimal solution with the historical global optimal solution, selecting the group of corresponding particles with the minimum fitness value as the global optimal solution of the iterative calculation, and turning to the step 3.7; if so, stopping iterative computation, and selecting a global optimal solution as a weight value and a threshold value of the self-adaptive back propagation neural network to obtain the self-adaptive back propagation neural network with the determined weight value and threshold value;
and 3.7, updating the speed and the position of each particle according to the global optimal solution and the individual optimal solution of the iterative computation, taking the updated particles as initial particles, t +1, and turning to the step 3.2.
The parameters of the adaptive back propagation neural network used for the data fusion process are:
net.trainParam.epochs=2500,
net.trainParam.goal=5e-4,
net.trainParam.lr=0.1,
xSize=26,
maxgen=300,
the method comprises the steps of net, trainParam, epochs are maximum training times (epochs) of the neural network, net, trainParam, goal is required precision of the neural network training, net, trainParam, lr is learning rate of the neural network, xSize represents M, namely the number of particle group groups in the adaptive neural network, and maxgen represents maximum iteration times of the adaptive algorithm.
The specific method in the step 4 comprises the following steps: an island fault detection model based on a Support Vector Machine (SVM): outputting x of the trained self-adaptive back propagation neural networki(i ═ 1,2, … n) as the input of step 4 support vector machine, the concrete steps are:
(1) importing training data xi(i ═ 1, 2.. n), a hyperplane is created and the decision-like function is described as:
y(xi)=ωTφ(xi)+b
where ω represents the normal vector of the hyperplane and b represents the deviation.
(2) If the decision function satisfies the conditionX is theniThe corresponding group of characteristic data is non-island data, which indicates that the power generation system under the data does not have an island; if the decision function satisfies the conditionX is theniThe corresponding group of characteristic data is island data, which indicates that the power generation system under the data has an island;
(3) object yiE { -1, 1} with each xiAnd (4) associating. The regularization parameter c is initialized. The problem of finding a hyperplane appears as a Quadratic Programming (QP) optimization problem.
Therein, ξi≧ 0 denotes a group represented by xiA corresponding slack variable that measures the distance between the margin and the inseparable sample.
(4) A Radial Basis Function (RBF) is used as a kernel function for the training data. Lagrange functions are used to solve the QP problem. The optimal parameters ω, b are obtained from the Karush-Kuhn-tucker (kkt) condition, and then the final decision function is rewritten as:
y(xi)=ω*K(xi,x)+b*
wherein x ═ x1,x2,…,xn]。
As shown in fig. 9, where (a) is the frequency at the common point collected in case 1, case 2, and (b) is the reactive power injected into the inverter collected in case 1, case 2.
As shown in fig. 10, in (a), case 1 is subjected to islanding detection only by using a support vector machine; (b) island detection is performed on case 2 by using a support vector machine only.
As shown in fig. 11, (a) is a result of performing island detection on case 1 by using the ABPS of the present embodiment, and (b) is a result of performing island detection on case 2 by using the ABPS of the present embodiment.
Comparing fig. 10 with fig. 11, it can be seen that the influence caused by the pseudo-islanding effect cannot be avoided only by using the support vector machine for detection; when the pseudo-island effect occurs, the invention can normally detect the real island state.
It should be noted that the various features described in the above embodiments may be combined in any suitable manner without departing from the scope of the invention. The invention is not described in detail in order to avoid unnecessary repetition.
Claims (6)
1. A data driving power island detection method based on active reactive power disturbance is characterized in that the method is used for judging whether an island exists in a distributed power generation system, and specifically comprises the following steps:
step 1: calculating to obtain reactive power injected into an inverter according to the common point frequency output by a phase-locked loop end in the distributed power generation system; taking the common point frequency and the corresponding reactive power injected into the inverter as a group of characteristic data sets; collecting the frequency of a common point and corresponding reactive power to obtain n groups of characteristic data sets;
step 2: sequentially carrying out data fusion on the n groups of characteristic data sets to obtain n groups of fusion data;
and step 3: calculating the weight and the threshold of the self-adaptive back propagation neural network by utilizing a particle swarm algorithm based on n groups of fusion data; meanwhile, in each particle swarm iterative computation, the n groups of characteristic data sets in the step 1 are used as a training set of the self-adaptive back propagation neural network to obtain a trained self-adaptive back propagation neural network;
and 4, step 4: taking the output of the trained self-adaptive back propagation neural network as the input of a support vector machine, and training the support vector machine to obtain a decision function; finally, the establishment of a training self-adaptive back propagation neural network-support vector machine classifier is completed;
and 5: changing a common point frequency fluctuation mode and a PLC load of an inverter to obtain a group of new characteristic data sets, and classifying the group of new characteristic data sets by using a self-adaptive back propagation neural network-support vector machine classifier so as to judge whether an island exists in the power system, wherein the method specifically comprises the following steps: and taking the new test data as a training set of the trained adaptive back propagation neural network to obtain the output of the adaptive back propagation neural network, and calculating whether the output is island data or not by using a decision function so as to judge whether the power generation system is in an island or not.
2. The active reactive power disturbance-based data-driven power island detection method according to claim 1, wherein the specific method for calculating the reactive power injected into the inverter in the step 1 is as follows:
where f' represents the common point frequency of the phase-locked loop output, frFor the resonant frequency, Q, of the inverter RLC loadfIs the quality factor, Q, of the inverterDGAnd PDGRepresenting reactive power and active power injected into the inverter, respectively.
3. The active reactive power disturbance-based data-driven power island detection method according to claim 1, wherein the specific method for sequentially performing data fusion on the n groups of feature data sets in the step 2 is as follows:
carrying out k-means calculation on the common point frequency in the n groups of characteristic data sets and the reactive power injected into the inverter to obtain a classification result of each common point frequency and the reactive power injected into the inverter; and performing numerical value distribution on each classification result;
performing data fusion on the D-th group of feature data sets by using the following formula, wherein D is 1,2, … n;
L=ClusterR∧Clusterf
wherein ClusterRCounting for the result of the classification of the reactive powerA value after value assignment; clusterfThe value is a value obtained by carrying out numerical value distribution on the classification result of the public point frequency; when L is {0,1}, and is 0, it indicates that the power system under the set of characteristic data sets is not islanded; the case 1 indicates that the power system under the set of characteristic data is islanded.
4. The active reactive power disturbance-based data-driven power island detection method according to claim 3, wherein a k-means algorithm is used for classifying the reactive power data set injected into the inverter, and the classification result is two types, wherein the first type is as follows: the reactive power injected into the inverter is not in the intermittent period; the second type is: the reactive power injected is in the intermittent period; according to the value of the common point frequency when the island occurs, classifying the common point frequency data set by using a k-means algorithm, wherein the classification result is three types, and the first type is as follows: common point frequency when islanding does not occur; the second type is: islanding occurs and reactive power injected into the inverter is at a common point frequency of the intermittent period; the third type is: islanding occurs and reactive power injected into the inverter is not at a common point frequency of an intermittent time period; the values of L after the numerical assignment based on the above classification results are shown in table 1 below:
TABLE 1
The first type of reactive power is 0 after being subjected to numerical value distribution, and the second type of reactive power is 1 after being subjected to numerical value distribution; the first class of common point frequencies is numerically assigned a value of 0, the second class is numerically assigned a value of 1, and the third class is numerically assigned a value of 2.
5. The active reactive power disturbance-based data-driven power island detection method according to claim 3, wherein the specific steps of the step 3 are as follows:
step 3.1: initializing parameters in a self-adaptive back propagation neural network and a particle swarm algorithm, wherein the size of a particle swarm is M;
step 3.2, carrying out the iterative computation for the t time, and sequentially taking the initial particles as the weight and the threshold of the self-adaptive back propagation neural network, wherein M self-adaptive back propagation neural networks exist;
step 3.3: taking n groups of characteristic data sets as training sets, sequentially inputting the training sets into an mth self-adaptive back propagation neural network, and carrying out epochs training on the n groups of characteristic data sets by the neural network to obtain n outputs;
step 3.4: making a mean square error between the Nth output and the Nth fusion data, wherein N is 1,2, …, N; obtaining n groups of mean square deviations, and taking the sum of the n groups of mean square deviations as a group of fitness values;
step 3.5, judging whether M is larger than M, if so, turning to step 3.6; if not, m +1, turning to step 3.3;
step 3.6: comparing the M groups of fitness values, and judging whether the fitness values smaller than Y exist in the M groups of fitness values, wherein Y is a preset numerical value; if yes, stopping the particle swarm algorithm, selecting a group of particles corresponding to the minimum fitness value from the fitness value group smaller than Y as a global optimal solution, and taking the global optimal solution as a weight and a threshold of the adaptive back propagation neural network to obtain the adaptive back propagation neural network with the determined weight and threshold, namely obtaining the trained adaptive back propagation neural network; if not, judging whether t reaches the maximum iteration times; if not, taking a group of corresponding particles with the minimum fitness value in the iterative calculation as the individual optimal solution of the iterative calculation, comparing the individual optimal solution with the historical global optimal solution, selecting the group of corresponding particles with the minimum fitness value as the global optimal solution of the iterative calculation, and turning to the step 3.7; if so, stopping iterative computation, and selecting a global optimal solution as a weight value and a threshold value of the self-adaptive back propagation neural network to obtain the self-adaptive back propagation neural network with the determined weight value and threshold value;
and 3.7, updating the speed and the position of each particle according to the global optimal solution and the individual optimal solution of the iterative computation, taking the updated particles as initial particles, t +1, and turning to the step 3.2.
6. The active reactive power disturbance-based data-driven power island detection method according to claim 1, wherein the specific method of the step 4 is as follows:
outputting x of the trained self-adaptive back propagation neural networki(i ═ 1,2, … n) as training data for the support vector machine and establish a hyperplane, describing the decision function as:
y(xi)=ωTφ(xi)+b
where ω represents the normal vector of the hyperplane, b represents the deviation,for finding x for mapping functionsiInner product of (d);
if the decision function satisfies the conditionX is theniThe corresponding group of characteristic data is non-island data, which indicates that the power generation system under the data does not have an island; if the decision function satisfies the conditionX is theniThe corresponding group of characteristic data is island data, which indicates that the power generation system under the data has an island;
based on yiE { -1, 1} with each xiAnd (3) associating, initializing a regularization parameter C, and expressing the hyperplane as a quadratic programming optimization problem:
ξ thereini≧ 0 denotes a group represented by xiA corresponding slack variable;
and (3) adopting a radial basis function as a kernel function of the training data, solving the quadratic programming optimization problem by utilizing a Lagrangian function, and rewriting a final decision function into:
y(xi)=ω*K(xi,x)+b*
wherein x ═ x1,x2,…,xn]ω, b is an optimum parameter, obtained from the KKT condition.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911189108.9A CN111160386B (en) | 2019-11-28 | 2019-11-28 | Data-driven power island detection method based on active reactive power disturbance |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911189108.9A CN111160386B (en) | 2019-11-28 | 2019-11-28 | Data-driven power island detection method based on active reactive power disturbance |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111160386A true CN111160386A (en) | 2020-05-15 |
CN111160386B CN111160386B (en) | 2023-11-10 |
Family
ID=70556182
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911189108.9A Active CN111160386B (en) | 2019-11-28 | 2019-11-28 | Data-driven power island detection method based on active reactive power disturbance |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111160386B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114565058A (en) * | 2022-03-16 | 2022-05-31 | 广东电网有限责任公司 | Training method, device, equipment and medium for island detection model |
CN114705947A (en) * | 2022-03-16 | 2022-07-05 | 广东电网有限责任公司 | Island detection model training method, device, equipment and medium |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110066391A1 (en) * | 2009-09-11 | 2011-03-17 | University Of Cincinnati | Methods and systems for energy prognosis |
CN104316786A (en) * | 2014-10-10 | 2015-01-28 | 湖南大学 | Mixed isolated island detection method |
CN105515044A (en) * | 2015-12-22 | 2016-04-20 | 国家电网公司 | Black-start assistant decision-making system based on DTS |
-
2019
- 2019-11-28 CN CN201911189108.9A patent/CN111160386B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110066391A1 (en) * | 2009-09-11 | 2011-03-17 | University Of Cincinnati | Methods and systems for energy prognosis |
CN104316786A (en) * | 2014-10-10 | 2015-01-28 | 湖南大学 | Mixed isolated island detection method |
CN105515044A (en) * | 2015-12-22 | 2016-04-20 | 国家电网公司 | Black-start assistant decision-making system based on DTS |
Non-Patent Citations (2)
Title |
---|
李洋,李坤: "并网逆变器新型控制策略的研究" * |
谢星宇: "光伏并网逆变器孤岛检测技术研究" * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114565058A (en) * | 2022-03-16 | 2022-05-31 | 广东电网有限责任公司 | Training method, device, equipment and medium for island detection model |
CN114705947A (en) * | 2022-03-16 | 2022-07-05 | 广东电网有限责任公司 | Island detection model training method, device, equipment and medium |
Also Published As
Publication number | Publication date |
---|---|
CN111160386B (en) | 2023-11-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Alimi et al. | A review of machine learning approaches to power system security and stability | |
Trivedi et al. | Implementation of artificial intelligence techniques in microgrid control environment: Current progress and future scopes | |
CN107153845A (en) | A kind of isolated island detection method of the adaptive grid-connected photovoltaic system based on machine learning | |
Li et al. | Islanding fault detection based on data-driven approach with active developed reactive power variation | |
CN111160386B (en) | Data-driven power island detection method based on active reactive power disturbance | |
da Cunha et al. | Small-signal stability analysis in smart grids: An approach based on distributed decision trees | |
Zhang et al. | Data-driven security and stability rule in high renewable penetrated power system operation | |
Lala et al. | Adaptive tuning of power system stabilizer using a damping control strategy considering stochastic time delay | |
Xue et al. | Adaptive ultra-short-term wind power prediction based on risk assessment | |
Alyasiri et al. | RETRACTED ARTICLE: Passive islanding detection in microgrids using artificial neural networks | |
Yin et al. | Inspired lightweight robust quantum Q-learning for smart generation control of power systems | |
Shahriyari et al. | A Deep Learning-Based Approach for Comprehensive Rotor Angle Stability Assessment | |
Ali et al. | An Optimized Algorithm for Renewable Energy Forecasting Based on Machine Learning. | |
Ebrahimi et al. | Machine learning-aided anomaly detection for secure smartification of solar inverters | |
Zhang et al. | Prediction modeling of frequency response characteristic of power system based on historical data | |
KR20190071174A (en) | Method and system for short-term wind speed prediction based on pressure data | |
Wang et al. | A Comprehensive Review on the Development of Data-Driven Methods for Wind Power Prediction and AGC Performance Evaluation in Wind-Thermal Bundled Power Systems | |
Xiong et al. | Forecasting ultra‐short‐term wind power by multiview gated recurrent unit neural network | |
Vantuch et al. | Power quality prediction designed as binary classification in AC coupling Off-Grid system | |
Yu et al. | Study on a comprehensive indicator and online classification of early warning of low frequency oscillation in power system | |
Bâra et al. | Comparative analysis between wind and solar forecasting methods using artificial neural networks | |
Elamine et al. | Multi-agent architecture for smart micro-grid optimal control using a hybrid BP-PSO algorithm for wind power prediction | |
Haque et al. | Machine learning in renewable energy systems for smart cities | |
Jie et al. | Investigation on Power System Dynamic Monitoring System Using Artificial Neural Networks | |
Raman et al. | Conditional generative adversarial networks for dynamic control-parameter selection in power systems |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |