CN116527060A - Information compression and anomaly detection method based on event trigger sampling - Google Patents

Abstract

The invention discloses an information compression and anomaly detection method based on event-triggered sampling, including the following steps: S1. Determine system data sampling event trigger conditions and thresholds, and complete data sampling of measurement variables; S2. Trigger sampling measurement variables according to events, and design events Triggering the Kalman filter to reconstruct the data; S3, using the residual analysis data to perform fault detection. The present invention adopts the information compression and anomaly detection method based on event-triggered sampling, which can reconstruct the sampled data on the basis of reducing the transmission pressure of the transmission channel, and perform system fault detection according to the reconstructed data. The accuracy and security of subsequent system performance analysis are guaranteed.

Description

Information Compression and Anomaly Detection Method Based on Event-Triggered Sampling

technical field

The invention relates to the technical field of data processing, in particular to an information compression and anomaly detection method based on event-triggered sampling.

Background technique

At present, in the field of industrial process control, data collection and monitoring are mostly realized through periodic sampling. Periodic sampling samples signals at fixed time intervals. Since each sampling node has nothing to do with the current working state, there is a serious waste of communication resources. question. For example, satellite signal monitoring systems, power grid monitoring systems, earthquake monitoring stations, and environmental quality signal acquisition systems will generate a large amount of redundant data during long-term operation, resulting in waste of system communication resources.

On the whole, time-triggered sampling control systems are too conservative in sampling frequency. Specifically, since the update frequency of the measured variables is fixed, in order to take into account the worst possible estimation results, a relatively conservative, that is, a relatively small sampling period is generally selected. It can be seen that the disadvantage of using this sampling method is that even when the estimator has obtained the ideal estimation accuracy and no longer needs to sample any new measurement updates, the measured variables will still be updated at a faster frequency. At the same time, when the period sampling period is small, too many transmission signals will increase the amount of data and occupy the resources of the data transmission channel, and the requirements for the capacity of the memory and the bandwidth of data transmission are relatively high. For example, when an unmanned vehicle is working in a ring with complex terrain, it does not need to transmit the monitoring data such as the remaining power and temperature of the unmanned vehicle to the remote control system at all times. Transmission at all times, so that the energy consumption of the unmanned vehicle can be saved, and the unmanned vehicle can work longer.

Contents of the invention

Aiming at the problem of massive data transmission in long-running process control systems, the present invention provides an information compression and anomaly detection method for event-triggered sampling, which realizes the compressed sampling of a large amount of sensor data and the data recovery of event-triggered state estimators without significant Under the premise of reducing the validity of data, minimize the transmission of redundant measurement variables due to the long-term operation of the system, reduce the number of sampling transmission updates, and realize the reconstruction of data signals through event-triggered state estimators, reducing channel load and signal The energy consumed by the transmission and fault detection can be performed by reconstructing the data.

In order to achieve the above object, the present invention provides an information compression and anomaly detection method based on event-triggered sampling, including the following steps:

S1. Determine the trigger condition and threshold of the system data sampling event, and complete the data sampling of the measured variable;

S2. According to event-triggered sampling measurement variables, design event-triggered Kalman filter to reconstruct data;

S3. Using the residual analysis data to perform fault detection.

Preferably, the system data described in step S1 includes control signals and measurement/controlled signals;

Among them, after the measurement/controlled signal is acquired by the sensor, it is judged whether to transmit according to the event trigger condition of the communication channel, and the event trigger condition of the communication channel is as follows:

In the formula, γ _k is a trigger variable with a value of 0 or 1, which represents the event trigger information sampled in the k-th step; y _k is the measured variable in the k-th step measured by the sensor; y _k-1 is measured by the sensor Step k-1 measures variables;

At this time, the condition is triggered when the difference between the sensor's current measured value y _k and the last transmitted measured value y _k-1 exceeds the threshold δ, and then the sensor will sample and transmit the system monitoring data.

Preferably, the steps for obtaining the threshold δ are as follows:

First sample to obtain the average communication rate

In the formula, N is the length of the time series;

Then, the estimated error is obtained by event-triggered state estimator

In the formula, Represents the estimated value of the system measurement variable _yk obtained via the event-triggered state estimator;

Finally, change the threshold δ to obtain the average communication rate under different thresholds δ with estimated error/> Plot the average communication rate /> with estimated error/> Select the event trigger condition threshold δ through the intersection point of the balance curve or the closest distance between the two.

Preferably, step S2 specifically includes the following steps:

S21, establishing the state space equation of the system;

S22, system variable initialization;

S23. After the event-triggered Kalman filter is used for state estimation, the system measurement value is reconstructed.

Preferably, in step S21, the control signal of the system is defined as u=[u ₁ ,u ₂ ,...,u _n ]∈R ⁿ , where R ⁿ is an n-dimensional real number matrix; the measurement/controlled signal definition is y=[y ₁ ,y ₂ ,...,y _m ]∈R ^m , R ^m is an m-dimensional real number matrix; and considering that the system measurement variable becomes a discrete signal after sampling, combining the control signal u _k and sampling measurement / The measured variable y _k obtained by the controlled signal establishes a discrete linear time-invariant system of the system:

In the formula, A is the system matrix, B is the control matrix, C is the observation matrix, D is the direct transfer matrix; x _k ∈ R ⁿ is the state signal/variable; u ₁ , u ₂ ,...,u _n are the estimated For the device, it is a known deterministic input control signal; w _k and v _k are used to represent process noise and measurement noise, respectively.

Preferably, step S22 specifically includes the following steps:

S221. Set initial state variable x ₀ =[a ₁ ,a ₂ ,…,a _n ] ^T , process noise and measurement noise/> Where a ₁ , a ₂ ,…, a _n , b ₁ , c ₁ are all non-negative constants;

S222. Establish an event-triggered state estimator according to the established discrete linear time-invariant system.

Preferably, step S222 specifically includes the following steps:

S2221. According to the normal working state of the system, zero-initialize the priori estimated state covariance matrix Prior_Sigma and the priori estimated state vector Prior_xhat; and make the posteriorly estimated state vector Poster_xhat=x ₀ Matrix Poster_Sigma=w _k ;

S2222. According to the system state space equation, the posterior estimation state vector Poster_xhat and the state covariance matrix Poster_xhat, the prior estimation variables are predicted as follows:

Prior_xhat(:,k+1)=A*Poster_xhat(:,k)+B*u _k (5)

Prior_Sigma(:,k+1)=A*Poster_Sigma(:,k)*A ^T +sigmw (6)

In the formula, Prior_xhat(:,k+1) represents the state vector of the prior estimation of sampling at step k+1; Poster_xhat(:,k) represents the state vector of posterior estimation of sampling at step k; Prior_Sigma(:,k +1) represents the state covariance matrix of the prior estimation of sampling at step k+1; Poster_Sigma(:,k) represents the state covariance matrix of posterior estimation of sampling at step k; sigmw represents the process noise variance;

S2223. According to the value of the event-triggered information γ _k , the a posteriori estimated state vector Poster_xhat and the a posteriori estimated state covariance matrix Poster_Sigma are combined with the event-triggered Kalman filter to estimate and update the system state vector.

Step S23 specifically includes the following steps:

After the event-triggered Kalman filter is used for state estimation, the measured variables of the system are reconstructed. According to the state space equation of the system, the reconstructed variables of the k-th step system can be obtained as follows:

Preferably, step S2223 specifically includes the following steps:

When the event-triggered information γ _k =1, the measured variables are measured and sampled by the system sensors to the event-triggered state estimator for state estimation. At this time, the steps to update the system state vector estimation according to the predicted prior estimated variables and measured variables are as follows:

The first step is to set the observation noise covariance matrix: H=sigmv;

The second step is to calculate the optimal event-triggered Kalman filter gain matrix:

K(:,k+1)＝Prior_Sigma(:,k+1)*C ^T /(C*Prior_Sigma(:,k+1)*C ^T +H) (8)

In the formula, K(:,k+1) represents the event-triggered Kalman filter gain matrix sampled at step k+1;

In the third step, the state vector of the posterior estimate is corrected by the difference between the a priori estimated value of the k+1 step and the measured variable of the k+1 step:

Poster_xhat(:,k+1)＝Prior_xhat(:,k+1)+K(:,k+1)*(y _k+1 -C*Prior_xhat(:,k+1)) (9)

In the formula, K(:,k+1)*(y _k+1 -C*Prior_xhat(:,k+1)) represents the correction part between the prior estimated state vector and the measured value, where K(:, k+1) represents the event-triggered Kalman filter gain matrix sampled at step k+1, and the Kalman gain matrix K(:,k+1) sampled at step k+1 is used as a weighting factor to balance the measured value and The influence of the state vector estimated a priori in the update;

The fourth step is to update the state covariance matrix of the posterior estimate:

Poster_Sigma(:,k+1)=(I-K(:,k+1)*C)*Prior_Sigma(:,k+1) (10);

When the event trigger information γ _k = 0, since the measured variable is transmitted to the event-triggered state estimator without being measured by the system sensor, the sampling value at the previous moment is used for state estimation. At this time, the steps for updating the system state vector estimation are as follows:

The first step is to set the observation noise covariance matrix: H=sigmv+1/Pi, which means that the size of the observation noise covariance matrix at this time is determined by the system noise and the probability false alarm rate Pi of the event-triggered state estimator. At this time, due to There is a lack of measured variables, so the noise matrix is increased to compensate for the estimation error;

The second step is to calculate the optimal event-triggered Kalman filter gain matrix:

K(:,k+1)=Prior_Sigma(:,k+1)*C ^T /(C*Prior_Sigma(:,k+1)*C ^T +H) (11);

The third step is to calculate the state vector measurement value of the posterior estimate:

Poster_xhat(:,k+1) = Prior_xhat(:,k+1) (12);

The fourth step is to update the state covariance matrix of the posterior estimate:

Poster_Sigma(:,k+1)=(I−K(:,k+1)*C)*Prior_Sigma(:,k+1) (13).

Preferably, step S3 specifically includes the following steps:

S31. Perform residual analysis according to the reconstructed data and measurement variables, and the residual Res is defined as follows:

S32. Set the alarm threshold as the standard deviation S _y of the overall measured variable:

In the formula, M represents the length of the system sampling data, y _i represents the system measurement variable of the ith sampling, Indicates the mean value of all system sampling data;

S33. Comparing the residual Res with the alarm threshold: if the residual Res exceeds the alarm threshold _Sy , it is judged that the system is malfunctioning at this moment; otherwise, it is judged that the system is running normally at this moment.

The present invention has the following beneficial effects:

1. Through event-triggered sampling and designing an event-triggered state estimator, the proposed event-triggered sampling, compared with the traditional data compression method based on periodic sampling, because its sampling control period is not fixed, but is triggered by a specific event, so it significantly reduces sampling frequency and communication volume.

2. Considering data reconstruction, using the implicit information at the event-triggered sampling time, an event-triggered Kalman filter is proposed. When the data is transmitted to the event-triggered Kalman filter, the received value is directly used; when the event triggers the Kalman filter When the Mann filter does not receive data, the measurement variables hidden in the event trigger conditions can be mined to reconstruct the data, thereby effectively improving the performance of the system.

3. After the data reconstruction is completed, use the residual analysis and the set alarm threshold to detect the fault of the abnormal signal.

The technical solutions of the present invention will be described in further detail below with reference to the accompanying drawings and embodiments.

Description of drawings

Fig. 1 is the flow diagram of the information compression and anomaly detection method based on event-triggered sampling of the present invention;

Fig. 2 is the simulation result figure of the rotational speed signal event triggering sampling of the embodiment of the present invention;

Fig. 3 is the simulation result diagram of the rotational speed data reconstruction of the embodiment of the present invention;

Fig. 4 is a graph of fault detection simulation results of rotational speed reconstruction data according to an embodiment of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages disclosed in the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below in conjunction with the accompanying drawings and the embodiments. It should be understood that the specific embodiments described here are only used to explain the embodiments of the present invention, and are not intended to limit the embodiments of the present invention. Based on the embodiments in the present application, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present application. Examples of the described embodiments are shown in the drawings, wherein like or similar reference numerals designate like or similar elements or elements having the same or similar functions throughout.

It should be noted that the terms "comprising" and "having" and any variations thereof are intended to cover a non-exclusive inclusion, for example, a process, method, system, product or server comprising a series of steps or units is not necessarily limited to expressly instead of those steps or elements listed, may include other steps or elements not explicitly listed or inherent to the process, method, product or apparatus.

Like numbers and letters denote similar items in the following figures, so that once an item is defined in one figure, it does not require further definition and explanation in subsequent figures.

In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "upper", "lower", "inner", "outer" etc. is based on the orientation or positional relationship shown in the drawings, or the The usual orientation or positional relationship of the invention product in use is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, therefore It should not be construed as a limitation of the present invention.

In the description of the present invention, it should also be noted that, unless otherwise clearly specified and limited, the terms "setting", "installation" and "connection" should be interpreted in a broad sense, for example, it can be a fixed connection or an optional connection. Detachable connection, or integral connection; it can be mechanical connection or electrical connection; it can be direct connection or indirect connection through an intermediary, and it can be the internal communication of two components. Those of ordinary skill in the art can understand the specific meanings of the above terms in the present invention in specific situations.

As shown in Figure 1, the information compression and anomaly detection method based on event-triggered sampling includes the following steps:

S1. Determine the trigger condition and threshold of the system data sampling event, and complete the data sampling of the measured variable;

In this embodiment, considering the signals of the satellite signal monitoring system, the grid monitoring system, and the unmanned vehicle system (the following systems specifically refer to this type of system), the key data of this type of system can be divided into two categories: control signals And measured/controlled signals (the / here and the other / in the text all represent or). After the controlled signal is acquired by the sensor, it is necessary to judge whether to transmit through the event trigger condition of the communication channel. For example, in the sensor data acquisition system of the unmanned vehicle system, since variables such as the speed of the unmanned vehicle generally only change significantly when starting or shifting gears, deterministic event trigger condition sampling is selected to reduce unnecessary rotational speed data transmission.

Preferably, the system data described in step S1 includes control signals and measurement/controlled signals;

Among them, after the measurement/controlled signal is acquired by the sensor, it is judged whether to transmit according to the event trigger condition of the communication channel, and the event trigger condition of the communication channel is as follows:

In the formula, γ _k is a trigger variable with a value of 0 or 1, which represents the event trigger information sampled in the k-th step; y _k is the measured variable in the k-th step measured by the sensor; y _k-1 is measured by the sensor Step k-1 measures variables;

At this time, the condition is triggered when the difference between the sensor's current measured value y _k and the last transmitted measured value y _k-1 exceeds the threshold δ, and then the sensor will sample and transmit the system monitoring data.

The average communication rate of the system represents the redundancy problem of sampled data, and the general Positively correlated with the threshold δ in the event trigger mechanism. In order to reduce the communication rate of the system, it is generally realized by increasing the threshold δ in the event trigger condition. Preferably, the steps for obtaining the threshold δ are as follows:

First sample to obtain the average communication rate

In the formula, N is the length of the time series;

If δ is set too large, too little data acquired by the system state estimator is not conducive to subsequent data recovery and fault detection, so the definition-estimation error

In the formula, Represents the estimated value of the system measurement variable _yk obtained via the event-triggered state estimator;

Finally, change the threshold δ to obtain the average communication rate under different thresholds δ with estimated error/> Plot the average communication rate /> with estimated error/> Select the event trigger condition threshold δ through the intersection point of the balance curve or the closest distance between the two.

S2. According to event-triggered sampling measurement variables, design event-triggered Kalman filter to reconstruct data;

Preferably, step S2 specifically includes the following steps:

S21, establishing the state space equation of the system;

Preferably, in step S21, the control signal of the system is defined as u=[u ₁ ,u ₂ ,...,u _n ]∈R ⁿ , where R ⁿ is an n-dimensional real number matrix; the measurement/controlled signal definition is y=[y ₁ ,y ₂ ,...,y _m ]∈R ^m , R ^m is an m-dimensional real number matrix; and considering that the system measurement variable becomes a discrete signal after sampling, combining the control signal u _k and sampling measurement / The measured variable y _k obtained by the controlled signal establishes a discrete linear time-invariant system of the system:

In the formula, A is the system matrix, B is the control matrix, C is the observation matrix, D is the direct transfer matrix; x _k ∈ R ⁿ is the state signal/variable; u ₁ , u ₂ ,...,u _n are the estimated For the device, it is a known deterministic input control signal; w _k and v _k are used to represent process noise and measurement noise, respectively.

S22, system variable initialization;

Preferably, step S22 specifically includes the following steps:

S221. Set initial state variable x ₀ =[a ₁ ,a ₂ ,…,a _n ] ^T , process noise and measurement noise/> Where a ₁ , a ₂ ,…, a _n , b ₁ , c ₁ are all non-negative constants;

S222. Establish an event-triggered state estimator according to the established discrete linear time-invariant system.

Preferably, step S222 specifically includes the following steps:

S2221. According to the normal working state of the system, zero-initialize the priori estimated state covariance matrix Prior_Sigma and the priori estimated state vector Prior_xhat; and make the posteriorly estimated state vector Poster_xhat=x ₀ Matrix Poster_Sigma=w _k ;

S2222. According to the system state space equation, the posterior estimation state vector Poster_xhat and the state covariance matrix Poster_xhat, the prior estimation variables are predicted as follows:

Prior_xhat(:,k+1)=A*Poster_xhat(:,k)+B*u _k (5)

Prior_Sigma(:,k+1)=A*Poster_Sigma(:,k)*A ^T +sigmw (6)

In the formula, Prior_xhat(:,k+1) represents the state vector of the prior estimation of sampling at step k+1; Poster_xhat(:,k) represents the state vector of posterior estimation of sampling at step k; Prior_Sigma(:,k +1) represents the state covariance matrix of the prior estimation of sampling at step k+1; Poster_Sigma(:,k) represents the state covariance matrix of posterior estimation of sampling at step k; sigmw represents the process noise variance;

S2223. According to the value of the event-triggered information γ _k , the a posteriori estimated state vector Poster_xhat and the a posteriori estimated state covariance matrix Poster_Sigma are combined with the event-triggered Kalman filter to estimate and update the system state vector.

Preferably, step S2223 specifically includes the following steps:

When the event-triggered information γ _k =1, the measured variables are measured and sampled by the system sensors to the event-triggered state estimator for state estimation. At this time, the steps to update the system state vector estimation according to the predicted prior estimated variables and measured variables are as follows:

The first step is to set the observation noise covariance matrix: H=sigmv;

The second step is to calculate the optimal event-triggered Kalman filter gain matrix:

K(:,k+1)＝Prior_Sigma(:,k+1)*C ^T /(C*Prior_Sigma(:,k+1)*C ^T +H) (8)

In the formula, K(:,k+1) represents the event-triggered Kalman filter gain matrix sampled at step k+1;

In the third step, the state vector of the posterior estimate is corrected by the difference between the a priori estimated value of the k+1 step and the measured variable of the k+1 step:

Poster_xhat(:,k+1)＝Prior_xhat(:,k+1)+K(:,k+1)*(y _k+1 -C*Prior_xhat(:,k+1)) (9)

In the formula, K(:,k+1)*(y _k+1 -C*Prior_xhat(:,k+1)) represents the correction part between the prior estimated state vector and the measured value, where K(:, k+1) represents the event-triggered Kalman filter gain matrix sampled at step k+1, and the Kalman gain matrix K(:,k+1) sampled at step k+1 is used as a weighting factor to balance the measured value and The influence of the state vector estimated a priori in the update;

The fourth step is to update the state covariance matrix of the posterior estimate:

Poster_Sigma(:,k+1)=(I-K(:,k+1)*C)*Prior_Sigma(:,k+1) (10);

When the event trigger information γ _k = 0, since the measured variable is transmitted to the event-triggered state estimator without being measured by the system sensor, the sampling value at the previous moment is used for state estimation. At this time, the steps for updating the system state vector estimation are as follows:

The first step is to set the observation noise covariance matrix: H=sigmv+1/Pi, which means that the size of the observation noise covariance matrix at this time is determined by the system noise and the probability false alarm rate Pi of the event-triggered state estimator. At this time, due to There is a lack of measured variables, so the noise matrix is increased to compensate for the estimation error;

The second step is to calculate the optimal event-triggered Kalman filter gain matrix:

K(:,k+1)=Prior_Sigma(:,k+1)*C ^T /(C*Prior_Sigma(:,k+1)*C ^T +H) (11);

The third step is to calculate the state vector measurement value of the posterior estimate:

Poster_xhat(:,k+1) = Prior_xhat(:,k+1) (12);

The fourth step is to update the state covariance matrix of the posterior estimate:

Poster_Sigma(:,k+1)=(I−K(:,k+1)*C)*Prior_Sigma(:,k+1) (13).

S23. After state estimation is performed using an event-triggered Kalman filter, the system measurement value is reconstructed;

Step S23 specifically includes the following steps:

After the event-triggered Kalman filter is used for state estimation, the measured variables of the system are reconstructed. According to the state space equation of the system, the reconstructed variables of the k-th step system can be obtained as follows:

It can be seen that the event-triggered Kalman filter can obtain a more accurate state estimation value by continuously estimating and updating the system state, thereby providing a benchmark value or reference value for the observed signal.

S3. Using the residual analysis data to perform fault detection.

Preferably, step S3 specifically includes the following steps:

S31. Perform residual analysis according to the reconstructed data and measurement variables, and the residual Res is defined as follows:

S32. In order to reflect the overall fluctuation degree of the system data, here, according to the actual measured value of the system, set the alarm threshold as the standard deviation S _y of the overall measured variable:

In the formula, M represents the length of the system sampling data, y _i represents the system measurement variable of the ith sampling, Indicates the mean value of all system sampling data;

S33. Comparing the residual Res with the alarm threshold: if the residual Res exceeds the alarm threshold _Sy , it is judged that the system is malfunctioning at this moment; otherwise, it is judged that the system is running normally at this moment.

As shown in FIG. 2 , FIG. 3 and FIG. 4 , they are the simulation results of the embodiment of the present invention. Figure 2 shows the event-triggered sampling effect for the rotational speed signal. It can be seen that under the sampling method proposed by the present invention, the sampling data points are far less than the original data, which realizes the data redundancy processing problem and can effectively save data. The resource occupation of the transmission channel. Figure 3 shows the effect of reconstructing the motor speed data of the unmanned vehicle through the event-triggered state estimator. It can be seen that through the proposed event-triggered Kalman filter reconstruction method, the reconstructed data basically coincides with the original data, which is a good example for the follow-up Signal fault detection provides an important basis. Figure 4. According to event triggering Kalman filter to analyze the residual error of the reconstructed unmanned vehicle speed data, and then use the residual error and the alarm threshold to compare the fault detection, which realizes the mining of hidden information of the system sampling signal and ensures the safety safe operation of the vehicle system.

Therefore, the present invention adopts the above information compression and anomaly detection method based on event-triggered sampling, and mines the measurement variables hidden in the event-triggered sampling data by setting an event-triggered event-triggered state estimator, thereby effectively improving system performance.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that: it still Modifications or equivalent replacements can be made to the technical solutions of the present invention, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims (10)

1. The information compression and anomaly detection method based on event trigger sampling is characterized by comprising the following steps of: the method comprises the following steps:

s1, determining a triggering condition and a threshold value of a system data sampling event, and completing data sampling of a measurement variable;

s2, sampling a measurement variable according to event triggering, and designing event triggering Kalman filter to reconstruct data;

and S3, performing fault detection by using residual analysis data.

2. The event triggered sampling based information compression and anomaly detection method of claim 1, wherein: the system data described in step S1 includes control signals and measurement/control signals;

after the measurement/controlled signal is acquired by the sensor, whether transmission is performed or not is judged through the event triggering condition of the communication channel, and the event triggering condition of the communication channel is as follows:

wherein, gamma _k A trigger variable of 0 or 1, which represents event trigger information of the kth sample; y is _k Is the kth step measured variable measured by the sensor; y is _k-1 Is the k-1 step measured variable measured by the sensor;

at this time, at the current measurement value y of the sensor _k And the last transmitted measurement y _k-1 The difference value exceeds the threshold delta, the condition is triggered, and the sensor samples and transmits the system monitoring data.

3. The event triggered sampling based information compression and anomaly detection method of claim 2, wherein: the step of obtaining the threshold delta is as follows:

first sampling to obtain average communication rate

Wherein N is the length of the time series;

then, the estimated error is obtained by the event-triggered state estimator

In the method, in the process of the invention,representing a system measurement variable y acquired via an event triggered state estimator _k Is a function of the estimated value of (2);

finally, changing the threshold value delta to obtain the average communication rate under different threshold values deltaError of estimation->Plotting average communication rateError of estimation->And (3) selecting an event triggering condition threshold delta from the balance curve between the two points or from the closest point of the balance curve.

4. The event triggered sampling based information compression and anomaly detection method of claim 1, wherein: the step S2 specifically comprises the following steps:

s21, establishing a state space equation of the system;

s22, initializing a variable of the system;

s23, reconstructing a system measured value after performing state estimation by using an event-triggered Kalman filter.

5. The event triggered sampling based information compression and anomaly detection method of claim 4, wherein: in step S21, the control signal of the system is defined as u= [ u ] ₁ ,u ₂ ,...,u _n ]∈R ⁿ ，R ⁿ Is a real matrix of n dimensions; the measurement/controlled signal is defined as y= [ y ] ₁ ,y ₂ ,...,y _m ]∈R ^m ，R ^m Is a real matrix of m dimension; and consider that the system measurement variable becomes discrete signal after sampling, combine with control signal u _k And a measurement variable y obtained by sampling the measurement/controlled signal _k Establishing a discrete linear time invariant system of the system:

wherein A is a system matrix, B is a control matrix, C is an observation matrix, and D is a direct transfer matrix; x is x _k ∈R ⁿ Is a status signal/variable; u (u) ₁ ,u ₂ ,...,u _n Is a deterministic input control signal that is known to the estimator; w (w) _k And v _k Respectively for representing process noise and measurement noise.

6. The event triggered sampling based information compression and anomaly detection method of claim 4, wherein: the step S22 specifically includes the following steps:

s221, setting initial state variable x ₀ ＝[a ₁ ,a ₂ ,…,a _n ] ^T Process noiseMeasurement noise->Wherein a is ₁ ,a ₂ ,…,a _n ,b ₁ ,c ₁ Are all non-negative constants;

s222, establishing an event trigger state estimator according to the established discrete linear time invariant system.

7. The event triggered sampling based information compression and anomaly detection method of claim 6, wherein: step S222 specifically includes the following steps:

s2221, initializing a state covariance matrix Prior_Sigma estimated in Prior and a state vector Prior_xhat estimated in Prior according to the normal working and running states of the system; and let the posterior estimated state vector Poster_xhat=x ₀ State covariance matrix of posterior estimation master_sigma=w _k ；

S2222, predicting prior estimated variables according to a system state space equation, a posterior estimated state vector Poster_xhat and a state covariance matrix Poster_xhat as follows:

Prior_xhat(:,k+1)＝A*Poster_xhat(:,k)+B*u _k (5)

Prior_Sigma(:,k+1)＝A*Poster_Sigma(:,k)*A ^T +sigmw (6)

where Prior_xhat (: k+1) represents the a priori estimated state vector of the k+1 th step sample; poster_xhat (: k) represents the state vector of the posterior estimate of the kth sample; prior_Sigma (: k+1) represents the state covariance matrix of the Prior estimate of the k+1th step sample; poster_Sigma (: k) represents the state covariance matrix of the posterior estimate of the kth sample; sigmw represents the process noise variance;

s2223, triggering information gamma according to event _k And (3) carrying out system state vector estimation update on the state vector Poster_xhat of posterior estimation and the state covariance matrix Poster_Sigma of posterior estimation in combination with an event-triggered Kalman filter.

8. The event triggered sampling based information compression and anomaly detection method of claim 6, wherein: the step S23 specifically includes the following steps:

after the state estimation is carried out by using the event-triggered Kalman filter, the system measurement variables are reconstructed, and the reconstruction variables of the system in the kth step can be obtained according to the state space equation of the system as follows:

9. the event triggered sampling based information compression and anomaly detection method of claim 7, wherein: step S2223 specifically includes the steps of:

when event triggering information gamma _k When=1, the measured variable is measured and sampled by the system sensor to be in the event-triggered state estimator for state estimation, and at this time, the system state vector estimation updating step according to the predicted prior estimated variable and the measured variable is as follows:

step one, setting an observed noise covariance matrix: h=sigmv;

secondly, calculating an optimal event triggering Kalman filter gain matrix:

K(:,k+1)＝Prior_Sigma(:,k+1)*C ^T /(C*Prior_Sigma(:,k+1)*C ^T +H) (8)

wherein K (wherein, k+1) represents an event-triggered Kalman filter gain matrix of the (k+1) th step of sampling;

third, correcting the state vector of the posterior estimation by using the prior estimation value of the k+1 step and the difference value of the measured variable of the k+1 step:

Poster_xhat(:,k+1)＝Prior_xhat(:,k+1)+K(:,k+1)*(y _k+1 -C*Prior_xhat(:,k+1)) (9)

wherein, K (: k+1): (y _k+1 -C: k+1) represents the correction part between the a priori estimated state vector and the measured value, wherein K (: k+1) represents the k+1-th sampled event triggered kalman filter gain matrix, k+1-th sampled kalman gain matrix K (: k+1) as a weighting factor for balancing the effect of the measured value and the a priori estimated state vector in the updating;

fourth, updating the state covariance matrix of the posterior estimation:

Poster_Sigma(:,k+1)＝(I-K(:,k+1)*C)*Prior_Sigma(:,k+1) (10)；

when event triggering information gamma _k Because the measured variable is not transmitted to the event-triggered state estimator through the measurement of the system sensor, the state estimation is performed by using the sampled value of the previous step, and the update procedure of the system state vector estimation is as follows:

step one, setting an observed noise covariance matrix: h=sigmv+1/Pi, which indicates that the magnitude of the observed noise covariance matrix is determined by the system noise and the probability false alarm rate Pi of the event-triggered state estimator, and the noise matrix is increased to compensate the estimation error due to lack of measurement variables;

secondly, calculating an optimal event triggering Kalman filter gain matrix:

K(:,k+1)＝Prior_Sigma(:,k+1)*C ^T /(C*Prior_Sigma(:,k+1)*C ^T +H) (11)；

thirdly, calculating a state vector measured value of posterior estimation:

Poster_xhat(:,k+1)＝Prior_xhat(:,k+1) (12)；

fourth, updating the state covariance matrix of the posterior estimation:

Poster_Sigma(:,k+1)＝(I-K(:,k+1)*C)*Prior_Sigma(:,k+1) (13)。

10. the event triggered sampling based information compression and anomaly detection method of claim 9, wherein: the step S3 specifically comprises the following steps:

s31, carrying out residual analysis according to the reconstructed data and the measurement variables, wherein the residual Res is defined as follows:

s32, setting the alarm threshold value as the standard deviation S of the whole measurement variable _y ：

Where M represents the length of the system sample data, y _i The system measurement variable representing the ith sample,representing the average value of all the system sampling data;

s33, comparing the residual Res with an alarm threshold value: if residual Res exceeds alarm threshold S _y Judging that the system fails at the moment; otherwise, judging that the system is running normally at the moment.