CN113315667B - State estimation method of time-lag complex network system under outlier detection - Google Patents
State estimation method of time-lag complex network system under outlier detection Download PDFInfo
- Publication number
- CN113315667B CN113315667B CN202110853538.7A CN202110853538A CN113315667B CN 113315667 B CN113315667 B CN 113315667B CN 202110853538 A CN202110853538 A CN 202110853538A CN 113315667 B CN113315667 B CN 113315667B
- Authority
- CN
- China
- Prior art keywords
- time
- complex network
- matrix
- formula
- outlier
- 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.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
- H04L43/0805—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters by checking availability
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
- G06F17/12—Simultaneous equations, e.g. systems of linear equations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Pure & Applied Mathematics (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Computing Systems (AREA)
- Environmental & Geological Engineering (AREA)
- Operations Research (AREA)
- Emergency Protection Circuit Devices (AREA)
Abstract
The invention discloses a state estimation method of a time-lag complex network system under outlier detection. The method comprises the following steps: establishing a state equation of a time-lag complex network system, and converting the state equation into a time-lag-free system equation; establishing a detection method of the intermittent occurrence outliers based on the non-time-lag system equation, and detecting all outliers of the time-lag complex network system in the operation process; according to the detection result of the outlier, obtaining an estimator parameter by using a convex optimization technology; and constructing an estimator, substituting the estimator parameters into the estimator, and calculating the estimated value of the time-lag complex network system state. The invention considers the influence of intermittent occurrence outliers in a time-lag complex network system and establishes a detection method, which can effectively detect all the output interfered by outliers; in addition, the invention also considers the influence of bounded noise, and utilizes the output which is not influenced by the outlier to generate the estimated value of the system state based on the convex optimization technology, thereby better meeting the application requirements of the actual industry.
Description
Technical Field
The invention belongs to the field of state estimation, and particularly relates to a state estimation method of a time-lag complex network system under outlier detection.
Background
The time-lapse complex network is formed by connecting a large number of nodes which are coupled with each other through a given topology, and can be used for simulating systems in many industries, such as a smart grid, an intelligent transportation network, the world wide web and the like. For time-lag complex network systems, many internal variables are generally unavailable due to the fact that coupling nodes are more, the size is larger and the like.
The importance of the state estimation technology for reconstructing the system state by using the output of the time-lag complex network system is self-evident in consideration of the important role of the state variables of the time-lag complex network system on monitoring the system operation. The core of the state estimation technology of the system state is to generate an estimated value of the state meeting the specific system performance by using model information and measurement information of the system.
In practice, a time-lag complex network system is usually interfered by external amplitude bounded noise, and if the processing is improper, the accuracy of state estimation is greatly influenced; on the other hand, outliers contained in the system measurements are typically large in amplitude (compared to noise) and occur intermittently, and if the outliers are not properly processed, they will be much more disruptive to the estimation than the noise.
Disclosure of Invention
The invention aims to provide a state estimation method of a time-lag complex network system under outlier detection, which fully considers the influence of outliers and bounded noise of external amplitude values to ensure the state estimation precision of the time-lag complex network system.
In order to achieve the purpose, the invention adopts the following technical scheme:
a state estimation method of a time-lag complex network system under outlier detection comprises the following steps:
and (3) calculating a state estimation value of the time-lag complex network system by utilizing an estimator related to the detection result according to the detection result of the intermittent occurrence outlier obtained in the step (2) and the estimator parameter solved in the step (3).
The invention has the following advantages:
as described above, the present invention provides a state estimation method for a time-lapse complex network system under outlier detection, which considers the influence of outliers occurring intermittently, and uses a high-order keeper to establish a detection method, and can give a high-precision estimation at each time no matter whether outliers in system measurement are detected at these times, thereby satisfying the requirements of the actual industry.
Drawings
FIG. 1 is a block diagram illustrating a flow chart of a method for estimating a state of a time-lag complex network system under outlier detection according to an embodiment of the present invention;
FIG. 2 is a schematic diagram illustrating the existence time of the outliers contained in the measurement output of the first set of sensors compared with the corresponding outlier detection results in accordance with an embodiment of the present invention;
FIG. 3 is a schematic diagram illustrating the existence time of the outliers contained in the measurement output of the second set of sensors compared with the corresponding outlier detection results according to an embodiment of the present invention;
FIG. 4 is a schematic diagram illustrating the existence time of the outliers contained in the measurement output of the third group of sensors in comparison with the corresponding outlier detection results in accordance with an embodiment of the present invention;
FIG. 5 is a schematic diagram illustrating a comparison between the system state (first component) of the first node and the estimated system state (first component) obtained by the estimation method according to the present invention;
FIG. 6 is a schematic diagram illustrating a comparison between the system state (second component) of the first node and the estimated system state (second component) obtained by the estimation method according to the present invention;
FIG. 7 is a schematic diagram showing a comparison between the system state (first component) of the second node and the estimated system state (first component) obtained by the estimation method according to the present invention;
fig. 8 is a schematic diagram showing a comparison between the system state (second component) of the second node and the estimated system state (second component) obtained by the estimation method according to the present invention;
FIG. 9 is a schematic diagram illustrating a comparison between the system state (first component) of the third node and the estimated system state (first component) obtained by the estimation method according to the present invention;
fig. 10 is a schematic diagram showing a comparison between the system state (second component) of the third node and the system estimated state (second component) obtained by the estimation method according to the present invention;
fig. 11 is a schematic diagram illustrating a comparison between the system state (first component) of the fourth node and the estimated system state (first component) obtained by the estimation method according to the present invention;
fig. 12 is a schematic diagram showing a comparison between the system state (second component) of the fourth node and the system estimated state (second component) obtained by the estimation method according to the present invention;
FIG. 13 is a schematic diagram illustrating a comparison of real output values of three nodes with estimator input values after a high-order keeper proposed by the method of the present invention;
fig. 14 shows a comparison of estimated error values for an estimated signal based on the method of the invention and based on the more applied lunberg estimator method.
Detailed Description
The invention is described in further detail below with reference to the following figures and detailed description:
as shown in fig. 1, a state estimation method for a time-lag complex network system under outlier detection includes the following steps:
The state equation of the time-lag complex network system is shown as formula (1);
wherein the content of the first and second substances,to represent complex network systemsA node; k represents a sampling instant;to represent complex network systemsAn n-dimensional state vector of each node at time k;is shown asQ-dimensional signal to be estimated at time k for each node.
j denotes the jth node of the complex network system, j =1,2, …, N denotes the number of nodes of the complex network system.
xj,kAn n-dimensional state vector representing the jth node of the complex network system at time k.
To represent complex network systemsCoupling between individual nodes and jth node ifIf there is a coupling between one node and the jth node, then>0, otherwise, the first step is to perform the following steps,=0。
is shown asMaximum skew of individual nodes;an adjacency matrix representing a complex network system whenWhen = j, there are。
Matrix arrayA parameter matrix representing the system, reflecting the influence of the system state at the current moment on the system state at the next moment; matrix arrayA parameter matrix representing the system, reflecting the influence of the system state at the historical moment on the system state at the next moment;a coupling matrix that is process noise;reflecting the relationship between the signal to be estimated and the system state.
the bounded process noise, representing the dimension r, for any sampling instant k,satisfies the inequality:wherein, in the step (A),representing a known normal, the operator | | · | | represents taking the euclidean norm for the vector "·".
If the system has measurements of d nodes, 1< d < N, the measurement equation of the system is shown in formula (2).
Wherein the content of the first and second substances,representing the system measurement output in m dimensions;indicating being boundedDimensional measurement noise, for any time k, satisfies,Is a known normal number.
CiAnd DiAre all known constant matrices.
oi,kAs field value of intermittent occurrence, oi,kThe following model is used for description, namely, the formula (3) shows.
In the formula (I), the compound is shown in the specification,representing a unit step function;denotes the j (th)1The amplitude of the secondary occurrence outliers, for any time k, is satisfied。
Wherein the content of the first and second substances,represents a known normal number;denotes the j (th)1The occurrence time of the individual outliers is,denotes the j (th)1The disappearance moment of the individual outliers;andsatisfies the inequality:
Order toDenotes the j (th)1Duration intervals of individual outliers; order toDenotes the j (th)11 time outlier disappearance and jth1The time interval between the occurrences of the secondary outliers can then be given by the following equation (4).
Definition ofHas an initial value of,Is initially ofThen there is(ii) a For arbitrary j1The following formula (5) and formula (6) are given.
Wherein the content of the first and second substances,is a known normal number; the physical meaning of equation (5) is jth in the system measurement1The duration interval of each outlier is not more than。
Wherein the content of the first and second substances,is a known normal number; the physical meaning of the formula (6) is j 11 time outlier disappearance and jth1The time interval between the occurrence of the sub-outliers is not shorter than. Let the state vector of the complex network system be:
then there are:
A = diag{A1,A2,…,AN},B = diag{B1,B2,…,BN},E = diag{E1,E2,…,EN};
diag {. } represents a block diagonal matrix,') ""represents the kronecker product of the matrix; 1i-1、1N-iA column vector representing all elements as 1; order to。
The following non-skew system is obtained:
to representI represents an identity matrix;an initial value representing the state of the dead time system,is divided intoIndividual blocks, i.e.。
Finally, by adopting an observable decomposition technology, the time-lag-free system (8) is transformed into:
in the formula (I), the compound is shown in the specification,is ni,1Dimension vector, representing the system's inability to observe at time kThe sub-states of (1);is ni,2A dimension vector representing a partial state that can be observed by the system at time k;
And 2, establishing a detection method of the intermittent occurrence outliers based on the time-lag-free system equation obtained in the step 1, and detecting all outliers of the time-lag complex network system in the operation process.
Based on an energy-based decomposition theory in a linear system theory and a Karley-Hamilton theorem in a matrix theory, a detection function and a detection threshold of a field value are given, whether the system measurement is influenced by the field value is judged by whether the detection function value exceeds the detection threshold, and all field values of the time-lag complex network in the operation process are detected.
For the ith measurement of a complex network system, a time k and a natural number s are given, and a detection function f is giveni(k, s) is defined as:
wherein the content of the first and second substances,the calculation is iterated by the following formula.
Wherein j is3 {0,1,…,ni,2-1},Representing a matrix of constantsCharacteristic polynomial ofThe coefficient of (d) is calculated by the following formula.
Where det (-) represents the determinant of the matrix "·"; defining a detection threshold value ofThen, thenObtained by solving equation (11).
Wherein r represents process noiseThe dimension of (a);represents ni,2A zero matrix of rows Nr columns,represents ni,2A zero matrix of rows Nrs columns.
All matrix norms referred to above are Frobenius norms.
for all j2Not less than 0, the occurrence time and disappearance time of the outlier are determined byAndare given one by one.
Formula (12) gives a method of detecting a field value occurring intermittently according to the detection function in formula (10) and the detection threshold in formula (11); for the ith sensor node, let j2=0,1, …, having:
1) j th2Detection of occurrence time of the secondary outlier:
let equation (10) detect function fiS =0 in (k, s) to give fi(k,0) not earlier than time of dayAnd satisfies the detection conditionIs the j-th time point k2The time of appearance of the secondary outliers.
2) J th2Detection of disappearance time of the secondary outlier:
satisfies the detection conditionOrder of (1)The smallest natural number s is the j2The disappearance of the secondary outliers.
According to the detection result of the outlier, a high-order keeper is adopted, and then the input of the estimator is represented as:
wherein the content of the first and second substances,indicating that no outliers are detected at time k, the estimator uses the received system measurements;indicating that the outlier is detected at time k, the estimator input is taken according to a high order hold mechanism。
Wherein the content of the first and second substances,representing the input signal of the estimator.
Equation (14) represents the time of dayLinearly combining if outliers are detected in the estimator inputAs input to the estimator.
Wherein, in the step (A),indicating the occurrence time of the outlier closest to the current time k.
Wherein, aggregateThe element (b) represents the occurrence time of all the outliers not later than the current time k;representation collectionThe maximum value of the occurrence time of the field value at the current time k.
The invention identifies the system output interfered by the outlier by establishing the outlier detection method, if the outlier is detected in the system measurement output at a certain time or a certain time, a high-order retainer is adopted at the moment, and the estimation precision of the estimator is effectively ensured.
And 3, calculating to obtain estimator parameters by utilizing a convex optimization technology according to the detection result of the system outlier.
The analysis of the estimation performance is based on the detection result of the system outlier in step 2, and the analysis result shows that when a specific linear matrix inequality has a solution, the estimation error index of the system is finally bounded.
The inventive method thus gives the estimator parameters by solving an optimization problem (the function to be optimized is the final bound) that satisfies the inequality constraints described previously, which as above can be achieved by means of convex optimization techniques.
Estimator parametersIs given by solving the solution of the optimization problem (17) with the linear matrix inequality constraints shown in equations (15) and (16), where equation (15), equation (16), and equation (17) are as follows:
MTM<P 1 (16)
in the formula (I), the compound is shown in the specification,, 0<λ<1 represents a given constant;P 1 = diag{P 1,1, P 1,2,…, P 1,Nand (c) the step of (c) in which,P 1,1, P 1,2,…, P 1,Nall are positive definite matrix variables to be solved;P 2the positive definite matrix variable to be solved.
wherein the calculation formula (18) calculates the estimator parametersWhen i =1,2 …, d, only the solved matrix needs to be usedP 1The first d sub-blocks ofP 1,1,P 1,2,,P 1,dThat is (due to d)<N)。
And 4, calculating a state estimation value of the time-lag complex network system by using the estimator.
And (3) calculating a state estimation value of the time-lag complex network system by using a Luneberg type estimator related to the detection result according to the detection result of the intermittent occurrence outlier obtained in the step (2) and the estimator parameter calculated in the step (3).
Using the input of the estimator given in equation (13) and given in equation (18)Is estimated from the estimated parametersCalculating a state estimation value of the time-lag complex network system, wherein a specific calculation formula is shown as a formula (19);
in the formula (19)Is 1,2, … …, N, N represents the number of nodes of the complex network, but forThe estimated value of the state has different calculation modes.
In particular, for the first d nodes of the system, i.e. forWhen values are taken in the set {1,2, … …, d }, corresponding sensor measurement information existsCan be used for the treatment of the diseases,the value of (2) is calculated by adopting a first formula in a formula (19);
for the (d + 1) th to (N) th nodes of the system, no corresponding measurement information is available,is calculated by the second formula in formula (19), and at this time,values are taken from the set { d +1, d +2, … …, N }.
Wherein the content of the first and second substances,to representIs determined by the estimated value of (c),to representAn estimated value of (d);is shown asAt time k for a nodeqAn estimate of a signal to be estimated of the dimension;
under the interference of bounded noise and outlier, the index of the estimated value of the signal to be estimated calculated by the formula (19) is finally bounded, and the final bound is the minimum value calculated by the optimization problem (17), namely。
The method of the invention considers the influence of intermittent occurrence outliers in a time-lag complex network system and establishes a detection method, which can effectively detect all the output interfered by outliers; the invention also considers the influence of bounded noise, and utilizes the output which is not influenced by the outlier to generate the estimated value of the system state based on the convex optimization technology, thereby better meeting the application requirements of the actual industry.
Because the complex network model can describe many practical systems, such as artificial neural networks, social relationship networks, power systems, information networks, and the like, the types of sensors used are different for different systems, such as electric meters for circuit systems; for moving objects, the sensors may be position sensors, velocity sensors, and the like.
When the method provided by the invention is applied, only the parameters of the corresponding system are needed to be used, and if the parameters related to the noise cannot be accurately obtained, the parameters which are relatively conservative (such as the upper bound of the amplitude of the noise is larger) can be used.
The state estimation method of the time-lag complex network system under outlier detection proposed by the invention is explained in combination with simulation verification in the following to verify the validity of the method proposed by the invention.
Taking the simulation step length as 150, a four-node complex network system is used, wherein three groups of sensors corresponding to the first three nodes of the system sample the system (considering the situation that part of nodes have no measurement output, the fourth node of the system does not use the corresponding sensor, namely no measurement output). The parameters of the used time-lapse complex network are:
C1=[0.6 0.9],C2=[0.9 -0.6], C3=[0.7 0.3];
M1=I,M2=1.1I,M3=0.9I,M4=0.8I;
D1=0.3, D2=0.4,D3=0.4;
Ξ=0.165I;
the process noise and measurement noise added are: omega1,k= 1sin(k), ω2,k= 2sin(1.2k), ω3,k= 3cos(0.8k0.8), ω4,k= 4cos(0.9k0.9+1.5);
By using the state estimation method provided by the invention, the estimation value is generated by using MATLAB software and is compared with the real value of each state variable of the time-lag complex network.
Fig. 2 to 4 show graphs of occurrence time of the outliers contained in the measurement output of the three sets of sensors and corresponding outlier detection results, wherein the upper half of fig. 2 to 4 shows the actual occurrence time and amplitude of the outliers of the first, second, and third nodes, and the lower half of fig. 2 to 4 shows the adopted outlier detection results, and the outlier detected is 1, otherwise 0.
As can be seen from fig. 2 to 4, the outlier detection method of the present invention can accurately detect the appearance time and the disappearance time of the outlier.
Fig. 5 to 12 show the trajectories of the system states of the first, second, third and fourth total four nodes and the estimated state of the system based on the estimation method proposed by the present invention (the solid line represents the true value of the system, and the dotted line represents the estimated value of the system).
FIGS. 5-6 are graphs showing the effect of estimating two components of the state of the first node; FIGS. 7-8 are graphs of the estimated effect of two components of the state of the second node; FIGS. 9-10 are graphs of the estimated effect of two components of the state of the third node; FIGS. 11-12 are graphs showing the effect of estimating two components of the state of the fourth node. It can be easily seen from fig. 5 to 12 that:
the estimated value of the system state obtained by the method of the invention can better track the true value of the system state.
Fig. 13 represents the actual output values of the first, second, and third nodes and the estimator input values after the high-order hold proposed by the present invention is adopted, the solid line represents the actual output trajectory of the system, and the dotted line represents the output trajectory after the high-order hold is adopted.
As can be seen from fig. 13, when the measured signal obtained by the estimator is detected to contain outliers, the input of the estimator generated by the advanced keeper of the present invention is very close to the real output of the system.
When the field value is included in the measurement, the measurement information is unavailable, and a common processing mode aiming at the problem is to adopt a zero input mechanism (the input of the estimator is zero and only uses model information for prediction) or a zero-order keeping mechanism (the input of the estimator adopts the input at the last moment), wherein the difference between the input of the estimator and the real output of the system which is not influenced by the field value is larger, which will certainly influence the estimation performance.
Fig. 14 shows a comparison of estimated error values based on the method of the invention and based on the more applied lunberg estimator method. The solid line represents the estimated error value trajectory based on the method of the invention and the dashed line represents the estimated error value trajectory based on the lunberg observer. As can be seen from fig. 14, the estimation error of the estimator according to the present invention is much lower than that of the traditional roberg estimator, and the effect is better than that of the traditional roberg estimator.
It should be understood, however, that the description herein of specific embodiments is not intended to limit the invention to the particular forms disclosed, but on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention as defined by the appended claims.
Claims (1)
1. A state estimation method of a time-lag complex network system under outlier detection is characterized by comprising the following steps:
step 1, establishing a state equation of a time-lag complex network system, and converting the state equation into a time-lag-free system equation;
the state equation of the time-lag complex network system is shown as formula (1);
wherein the content of the first and second substances,to represent complex network systemsA node; k represents a sampling instant;to represent complex network systemsAn n-dimensional state vector of each node at time k;is shown asQ-dimensional signals to be estimated of each node at k time;
j represents the jth node of the complex network system, j =1,2, …, N represents the number of nodes of the complex network system;
xj,krepresenting an n-dimensional state vector of a jth node of a complex network system at time k;
To represent complex network systemsCoupling between individual nodes and jth node ifIf there is a coupling between one node and the jth node, then>0, otherwise, the first step is to perform the following steps,=0;
is shown asMaximum skew of individual nodes;an adjacency matrix representing a complex network system whenWhen = j, there are;
matrix arrayA parameter matrix representing the system, reflecting the influence of the system state at the current moment on the system state at the next moment; matrix arrayA parameter matrix representing the system, reflecting the influence of the system state at the historical moment on the system state at the next moment;a coupling matrix that is process noise;reflecting the relation between the signal to be estimated and the system state;
the measurement of d nodes of the system can be obtained, and if d is more than 1 and less than N, the measurement equation of the system is shown as a formula (2);
wherein the content of the first and second substances,representing the system measurement output in m dimensions;indicating being boundedDimensional measurement noise, for any time k, satisfies,Is a known normal number;
Ciand DiAre all known constant matrices;
oi,kas field value of intermittent occurrence, oi,kThe following model is used for description, namely the formula (3);
in the formula (I), the compound is shown in the specification,representing a unit step function;denotes the j (th)1The amplitude of the secondary occurrence outliers, for any time k, is satisfied;
Wherein the content of the first and second substances,represents a known normal number;denotes the j (th)1The occurrence time of the individual outliers is,denotes the j (th)1The disappearance moment of the individual outliers;andsatisfies the inequality:
order toDenotes the j (th)1Duration intervals of individual outliers; order toDenotes the j (th)11 time outlier disappearance and jth1The time interval between occurrences of the secondary outliers, then the following equation (4) can be obtained;
definition ofHas an initial value of,Is initially ofThen there is(ii) a For arbitrary j1The following formula (5) and formula (6);
wherein the content of the first and second substances,is a known normal number; the physical meaning of equation (5) is jth in the system measurement1The duration interval of each outlier is not more than;
Wherein the content of the first and second substances,is a known normal number; the physical meaning of the formula (6) is j11 time outlier disappearance and jth1Time intervals between occurrences of sub-outliersNot shorter than;
Let the state vector of the complex network system be:
then there are:
A = diag{A1,A2,…,AN},B = diag{B1,B2,…,BN},E = diag{E1,E2,…,EN};
diag {. } represents a block diagonal matrix,') ""represents the kronecker product of the matrix; 1i-1、1N-iA column vector representing all elements as 1;
wherein:
to representI represents an identity matrix;an initial value representing the state of the dead time system,is divided intoIndividual blocks, i.e.;
Finally, by adopting an observable decomposition technology, the time-lag-free system (8) is transformed into:
in the formula (I), the compound is shown in the specification,is ni,1A dimension vector representing a sub-state that the system cannot observe at time k;is ni,2A dimension vector representing a partial state that can be observed by the system at time k;
step 2, establishing a detection method of the intermittent occurrence outliers based on the time-lag-free system equation obtained in the step 1, and detecting all outliers of the time-lag complex network system in the operation process;
for the ith measurement of a complex network system, a time k and a natural number s are given, and a detection function f is giveni(k, s) is defined as:
wherein the content of the first and second substances,the calculation is carried out iteratively by the following formula;
wherein j is3 {0,1,…,ni,2-1},Representing a matrix of constantsCharacteristic polynomial ofThe coefficient of (d) is calculated by the following formula;
where det (-) represents the determinant of the matrix "·"; defining a detection threshold value ofThen, thenSolving by a formula (11);
wherein r represents process noiseThe dimension of (a);represents ni,2A zero matrix of rows Nr columns,represents ni,2A zero matrix of rows Nrs columns;
All the matrix norms referred to above are Frobenius norms;
for all j2Not less than 0, the occurrence time and disappearance time of the outlier are determined byAndgiving out the raw materials one by one;
formula (12) gives a method of detecting a field value occurring intermittently according to the detection function in formula (10) and the detection threshold in formula (11); for the ith sensor node, let j2=0,1, …, having:
1) j th2Detection of occurrence time of the secondary outlier:
let equation (10) detect function fiS =0 in (k, s) to give fi(k,0) not earlier than time of dayAnd satisfies the detection conditionIs the j-th time point k2The appearance time of the secondary outlier;
2) j th2Detection of disappearance time of the secondary outlier:
satisfies the detection conditionOrder of (1)The smallest natural number s is the j2The disappearance moment of the secondary outlier;
according to the detection result of the outlier, a high-order keeper is adopted, and then the input of the estimator is represented as:
wherein the content of the first and second substances,indicating that no outliers are detected at time k, the estimator uses the received system measurements;indicating that the outlier is detected at time k, the estimator input is taken according to a high order hold mechanism;
Wherein the content of the first and second substances,representing an input signal of an estimator; equation (14) shows that at time k, if a outlier is detected in the estimator input, the linear combinations will be
wherein, in the step (A),indicating the occurrence time of the outlier closest to the current time k;
wherein, aggregateThe element (b) represents the occurrence time of all the outliers not later than the current time k;representation collectionMaximum value in occurrence time of field value not later than current time k;
step 3, according to the detection result of the outlier, calculating by using a convex optimization technology to obtain an estimator parameter;
estimator parametersIs given by solving the solution of the optimization problem (17) with the linear matrix inequality constraints shown in equations (15) and (16), where equation (15), equation (16), and equation (17) are as follows:
MTM<P 1 (16)
in the formula (I), the compound is shown in the specification,, 0<λ<1 represents a given constant;P 1 = diag{P 1,1, P 1,2,…, P 1,Nand (c) the step of (c) in which,P 1,1, P 1,2,…, P 1,Nall are positive definite matrix variables to be solved;P 2a positive definite matrix variable to be solved;
step 4, calculating a state estimation value of the time-lag complex network system by using an estimator;
calculating a state estimation value of the time-lag complex network system by using an estimator related to the detection result according to the detection result of the intermittent occurrence outlier obtained in the step 2 and the estimator parameter solved in the step 3;
calculating a state estimation value of the time-lag complex network system by using the input of the estimator given in the formula (13) and the estimator parameter given in the formula (18), wherein the specific calculation formula is shown as a formula (19);
wherein the content of the first and second substances,to representIs determined by the estimated value of (c),to representAn estimated value of (d);is shown asAt time k for a nodeqAn estimate of a signal to be estimated of the dimension;
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110853538.7A CN113315667B (en) | 2021-07-28 | 2021-07-28 | State estimation method of time-lag complex network system under outlier detection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110853538.7A CN113315667B (en) | 2021-07-28 | 2021-07-28 | State estimation method of time-lag complex network system under outlier detection |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113315667A CN113315667A (en) | 2021-08-27 |
CN113315667B true CN113315667B (en) | 2021-10-15 |
Family
ID=77381711
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110853538.7A Active CN113315667B (en) | 2021-07-28 | 2021-07-28 | State estimation method of time-lag complex network system under outlier detection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113315667B (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111025914A (en) * | 2019-12-26 | 2020-04-17 | 东北石油大学 | Neural network system remote state estimation method and device based on communication limitation |
CN112180899A (en) * | 2020-09-30 | 2021-01-05 | 山东科技大学 | State estimation method of system under intermittent anomaly measurement detection |
CN112257012A (en) * | 2020-10-10 | 2021-01-22 | 山东科技大学 | State estimation method of time-varying system under influence of abnormal signal |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100274102A1 (en) * | 2009-04-22 | 2010-10-28 | Streamline Automation, Llc | Processing Physiological Sensor Data Using a Physiological Model Combined with a Probabilistic Processor |
JP6477199B2 (en) * | 2015-04-23 | 2019-03-06 | 沖電気工業株式会社 | Vibration state estimation device, vibration state estimation method, and program |
US11146579B2 (en) * | 2018-09-21 | 2021-10-12 | General Electric Company | Hybrid feature-driven learning system for abnormality detection and localization |
CN109709579B (en) * | 2019-02-21 | 2023-02-14 | 哈尔滨工程大学 | GNSS satellite ephemeris fault detection method based on user ranging error real-time estimation |
CN112418051B (en) * | 2020-11-18 | 2024-05-03 | 温州大学 | State estimation method for nonlinear dynamic system under non-Gaussian noise |
-
2021
- 2021-07-28 CN CN202110853538.7A patent/CN113315667B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111025914A (en) * | 2019-12-26 | 2020-04-17 | 东北石油大学 | Neural network system remote state estimation method and device based on communication limitation |
CN112180899A (en) * | 2020-09-30 | 2021-01-05 | 山东科技大学 | State estimation method of system under intermittent anomaly measurement detection |
CN112257012A (en) * | 2020-10-10 | 2021-01-22 | 山东科技大学 | State estimation method of time-varying system under influence of abnormal signal |
Non-Patent Citations (6)
Title |
---|
SASDN时滞不确定性分析的顶层设计;葛磊蛟等;《电力工程技术》;20200531;全文 * |
Ultimately_Bounded_Filtering_Subject_to_Impulsive_Measurement_Outliers;Lei Zou,etc;《IEEE》;20210518;全文 * |
Uplink Channel Estimation for Massive MIMO Systems With Impulsive Noise;Lei Zhou,etc;《IEEE》;20201211;全文 * |
具有分布式时滞与一步随机时滞系统的H∞滤波;赵忠义等;《山东科技大学学报(自然科学版)》;20170815;全文 * |
基于状态观测器的历史数据建模方法研究;董泽等;《系统仿真学报》;20180108(第01期);全文 * |
时滞系统的自抗扰控制综述;王丽君等;《控制理论与应用》;20131215(第12期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN113315667A (en) | 2021-08-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zhang et al. | Machine tool thermal error modeling and prediction by grey neural network | |
Darouach et al. | Data reconciliation in generalized linear dynamic systems | |
Tavakolpour-Saleh et al. | Parametric and nonparametric system identification of an experimental turbojet engine | |
CN109145516B (en) | Analog circuit fault identification method based on improved extreme learning machine | |
CN106873568B (en) | Sensor fault diagnosis method based on H infinity robust Unknown Input Observer | |
CN111698257B (en) | Industrial information physical system security detection method for multi-class malicious attacks | |
Henry et al. | The prism: efficient signal processing for the internet of things | |
CN113315667B (en) | State estimation method of time-lag complex network system under outlier detection | |
Zhang et al. | Improved Renyi entropy benchmark for performance assessment of common cascade control system | |
CN111553226B (en) | Method for extracting river monitoring section water surface width based on remote sensing interpretation technology | |
Horvath et al. | Sensor fault diagnosis of inland navigation system using physical model and pattern recognition approach | |
Zhang et al. | Iterative learning control for a class of parabolic system fault diagnosis | |
CN111412391B (en) | Pipe network leakage detection method and system | |
CN111108738B (en) | Data processing device, data analysis device, data processing system, and method for processing data | |
CN107563103B (en) | Consistency filter design method based on local conditions | |
CN111062139A (en) | Method for estimating water flow situation of urban underground drainage pipe network | |
CN112948755B (en) | Telemetry sine parameter interpretation method | |
Robinson et al. | Model-based prognosis using an explicit degradation model and Inverse FORM for uncertainty propagation | |
CN110598334B (en) | Performance degradation trend prediction method based on collaborative derivation related entropy extreme learning machine | |
CN106936628B (en) | It is a kind of meter and sensor fault fractional order network system situation estimation method | |
CN114490596A (en) | Method for cleaning transformer oil chromatographic data based on machine learning and neural network | |
Zhu et al. | Fault detection for discrete systems based on a descriptor system method | |
Yang et al. | Robust weighted fusion Kalman estimators for systems with uncertain noise variances, multiplicative noises, missing measurements, packets dropouts and two‐step random measurement delays | |
CN111048203B (en) | Brain blood flow regulator evaluation device | |
Zhang et al. | Distributed input and state estimation for linear discrete-time systems in sensor networks with missing measurements |
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 |