A dynamic state estimation method for distribution network against random data packet loss
Through Markov jump model and Kalman filtering method, a measurement data compensation strategy is constructed, which solves the problem of failure of dynamic state estimation of distribution network when data packet loss is solved, and resists denial of service attacks are realized, ensuring state estimation is ensured. accuracy and stability.
Patent Information
- Application Number
- CN202011473339.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-12-15
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2040-12-15
AI Technical Summary
Dynamic status estimates of distribution networks are easily failed when facing data packet loss, especially in denial of service attacks, where data packet loss cannot be effectively tolerated, resulting in increased security risks.
The Markov jump model and Markov chain are used to characterize the data packet loss process, define the Markov chain transfer probability, set up the measured data packet loss compensation strategy, and combine the improved traceless Kalman filtering and extended Kalman filtering Methods: A distribution network state estimation model is constructed with a fusion measurement data compensation strategy, and state prediction and update.
Effectively resist the random packet loss problem of data caused by denial of service attacks, ensure the accuracy and stability of distribution network state estimation, and avoid state estimation failure.
Smart Images

Figure CN112670981B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of dynamic state estimation of distribution networks, and in particular to a dynamic state estimation method for distribution networks capable of resisting random data packet loss. Background Art
[0002] With the in-depth development of smart distribution networks, intelligent devices such as measurement, communication, computing and control have an increasingly greater impact on distribution networks. As distribution networks become more and more intelligent, network attacks can more easily invade information devices. Although network attacks cannot directly attack physical devices in distribution networks, they can damage information devices and make them perceive incorrect data, thereby issuing incorrect instructions to physical devices, which brings great security risks to distribution networks.
[0003] The distribution network state estimation uses the measurement data provided by the Supervisory Control and Data Acquisition (SCADA) system as the data source. State estimation can reduce measurement errors, eliminate bad data, and output data that the distribution network measurement equipment cannot perceive, providing support for economic dispatch, automatic power generation control, optimal power flow and other decision-making analysis of the power system. Malicious network attacks such as denial of service attacks can cause packet loss in the measurement data source of state estimation.
[0004] Therefore, it is hoped that there is a state estimation algorithm that can tolerate data packet loss and will not cause state estimation failure when data packet loss occurs. Summary of the invention
[0005] In order to solve the above technical problems, the present invention proposes a distribution network dynamic state estimation method for resisting random data packet loss, which can effectively resist the problem of random data packet loss caused by denial of service attacks;
[0006] A method for estimating dynamic state of a distribution network against random data packet loss comprises the following steps:
[0007] Step 1: Use the Markov jump model to characterize the packet loss of distribution network measurement data, propose a measurement data loss compensation strategy, and establish a Markov chain model;
[0008] The Markov chain is a Markov process The parameter set T is a discrete time set, The state space composed of the values of is a discrete state I = {m0,m1…,m k ,m k+1 ,…}, with a random process If for any integer and any m0,m1…,mk,mk +1∈I, whose conditional probability satisfies Then it is called is a Markov chain. Pr represents the probability, the parameter set T is a discrete time set, and t k is a value in the parameter set T, The state space composed of the values of is a discrete state, I = {m0,m1…,m k ,m k+1 ,…}, I is the state set, is a set of integers.
[0009] Step 2: Establish a model of SCADA measurement data packet loss;
[0010] Assuming the sampling period is a, the set of received measurements without DoS attack and without loss is defined as S1 = {Z1, Z2, Z3, ···Z k ···}, the set of received measurements after the attack caused data packet loss is in The number of consecutive measurements lost is represented by the set I = {0, 1, 2, ..., M}, where M represents the maximum number of measurement data lost. Assume that the most recently received measurement is The next received measurement is make but express The number of measurements lost between
[0011] Step 3: Define the Markov chain transition probability. In the complete probability space, if the data packet loss process It takes values in the finite probability space M = {0,1,2,…,M} and is a discrete Markov random process with a transition probability of It is called a Markov packet loss process. The transition probability constraint is
[0012] Step 4: Characterize the probability of random transition of packet loss in the measurement process. The random transition probability of packet loss in the measurement process under a denial of service attack is expressed as T gives the energy constraint of data packet loss; p ij It represents the probability that the measurement loss caused by the DoS attack is transferred from i-1 to j-1. The specific expansion form of the transfer probability is The sum of the elements in each row is 1, that is, p ij ≥0, P is the transition probability matrix, p ij are the elements of the matrix P;
[0013] Step 5: Set the measurement data packet loss compensation strategy; for measurement compensation, use the successfully transmitted measurement to compensate for the lost measurement. arrive Lost in between For measurement data Indicates; for Z k The measured data after compensation is expressed as Z dk Indicates, where k represents the sampling time, t k If Z is the time it takes to successfully transmit dk It is expressed as follows:
[0014]
[0015] Step 6: Construct a distribution network state estimation model integrating the measurement data compensation strategy. The distribution network state estimation model selects nonlinear measurement equations and state equations. The distribution network state estimation model integrating the measurement data compensation strategy is as follows: the first is the state equation, and the second is the measurement equation.
[0016] x k =f(x k-1 )+q k-1
[0017] Z k =Z dk =h(x k )+r k
[0018] The state quantity x k It is the quantity to be sought in n-dimensional band, including bus voltage amplitude and bus voltage phase angle, x k-1 is x k The state quantity obtained at the last sampling moment, quantity measurement Z k It is an m-dimensional measurement that is measured by SCADA and then collected into the energy management system, including branch active and reactive power flow, bus active and reactive power and voltage amplitude; Z dk is the measurement of Z k Compensation after a denial of service attack. f(·) is the state prediction function, h(·) is the measurement matrix, which is determined by the admittance matrix of the bus system and Kirchhoff's law, q k-1 ,r k They are prediction error and measurement error, which are independent of each other;
[0019] Step 7: Construct a prediction function model; the state prediction function f(·) uses Holt's two-parameter exponential smoothing method, x k+1∣k is the state quantity at time k+1 predicted by the state quantity at time k, x k∣k-1 is the state quantity at time k predicted by the state quantity at time k-1, is an estimated value, S k and b k is an intermediate variable, α L , β L is the smoothing parameter, with a value between 0 and 1:
[0020]
[0021] Step 8: Use the improved Unscented Kalman Filter (UKF) method to update the state prediction;
[0022] Step 8.1: At time k-1, the sampling particles are selected based on the unscented transformation using the symmetric sampling strategy method. The sampling points are selected as shown in the following formula;
[0023]
[0024] λ=α 2 (n+κ)-n
[0025]
[0026] where χ i,k-1 is the sampling point selected at time k-1, i represents the sampling point number, a total of 2n+1 are selected, n is the dimension of the state quantity, is the estimated value of the state at time k-1, is the state covariance at time k-1, λ is the scaling factor, W i m is the weight of the mean, W i c is the weight of the variance, κ is the redundancy set to 0, and α is the parameter that changes the distribution distance of the particles;
[0027] Step 8.2: Substitute the sampling points sampled by the unscented transformation method into the defined measurement equation and state equation to obtain the predicted state value and predicted measurement value at the predicted time k. is the predicted quantity measure obtained using the UKF method, is the predicted state quantity obtained by the UKF method;
[0028]
[0029] where χ i,k is the state prediction function f(·) and the sampling point χ selected at time k-1 i,k-1 The predicted sampling point at time k is obtained; Z i,k is the predicted measurement point at time k;
[0030] Step 8.3: By finding the covariance of the state estimate and the measurement estimate Estimated variance of the quantity measured at time k Find the Kalman gain matrix based on the UKF method yes The inverse matrix of T is the transpose of the matrix, R k is the measurement noise covariance matrix.
[0031]
[0032] Step 8.4: At time k, sample to obtain a new measurement value Z k Afterwards, the compensation strategy adopted for the measured data is used according to step 5, Z dk is the measured value after compensation. Then, according to step 8, the Kalman gain correction state quantity based on the UKF method is obtained. is the state quantity after correction, and the state quantity update steps are as follows.
[0033]
[0034] Step 9: Combine the extended Kalman filter (EKF) method with data packet loss compensation to update the state prediction and obtain the estimated value of the distribution network state;
[0035] Step 9.1: Use the extended Kalman filter (EKF) method to re-predict the state estimate obtained by UKF to improve the prediction accuracy, as shown in the following formula.
[0036]
[0037] Step 9.2: Correct and obtain the final state estimate as shown below.
[0038]
[0039] Among them, Q k is the system noise covariance matrix, R k is the measurement noise covariance matrix, F k G K is the Jacobian matrix of the system model, H k and U k is the Jacobian matrix of the measurement equation, is the Kalman gain obtained by the EKF method, is the predicted state covariance matrix obtained by the EKF method, and finally are the state covariance matrix and estimated value at time k respectively.
[0040] The beneficial effects produced by the present invention are:
[0041] This technical solution provides a distribution network dynamic state estimation method for resisting random data packet loss, which can effectively resist the problem of random data packet loss caused by denial of service attacks. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] Figure 1 This is a flow chart of a method for estimating dynamic state of a distribution network for resisting random data packet loss according to the present invention;
[0043] Figure 2 A flow chart of a data packet loss modeling compensation strategy in an embodiment of the present invention;
[0044] Figure 3 This is a diagram showing an IEEE 30-node calculation example in an embodiment of the present invention;
[0045] Figure 4 It is a simulation result diagram in an embodiment of the present invention;
[0046] in Figure 4 -a is the voltage amplitude curve corresponding to bus No. 5, Figure 4 -b is the voltage phase angle curve corresponding to bus No. 5. DETAILED DESCRIPTION
[0047] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.
[0048] A dynamic state estimation method for distribution network against random data packet loss, such as Figure 1 As shown, the following steps are included:
[0049] Step 1: Use the Markov jump model to characterize the packet loss of distribution network measurement data, propose a measurement data loss compensation strategy, and establish a Markov chain model;
[0050] The Markov chain is a Markov process The parameter set T is a discrete time set, The state space composed of the values of is a discrete state I = {m0,m1…,m k ,m k+1 ,…}, with a random process If for any integer and any m0,m1…,m k ,m k+1 ∈I, whose conditional probability satisfies Then it is called is a Markov chain. Pr represents the probability, the parameter set T is a discrete time set, and t k is a value in the parameter set T, The state space composed of the values of is a discrete state, I = {m0,m1…,m k ,m k+1 ,…}, I is the state set, is a set of integers.
[0051] Step 2: Establish a model for SCADA measurement data packet loss, such as Figure 2 As shown;
[0052] Assuming the sampling period is a, the set of received measurements without DoS attack and without loss is defined as S1 = {Z1, Z2, Z3, ···Z k ···}, the set of received measurements after the attack caused data packet loss is in The number of consecutive measurements lost is represented by the set I = {0, 1, 2, ..., M}, where M represents the maximum number of measurement data lost. Assume that the most recently received measurement is The next received measurement is make but express The number of measurements lost between
[0053] Step 3: Define the Markov chain transition probability. In the complete probability space, if the data packet loss process It takes values in the finite probability space M = {0,1,2,…,M} and is a discrete Markov random process with a transition probability of It is called a Markov packet loss process. The transition probability constraint is
[0054] Step 4: Characterize the probability of random transition of packet loss in the measurement process. The random transition probability of packet loss in the measurement process under a denial of service attack is expressed as T gives the energy constraint of data packet loss; p ij It represents the probability that the measurement loss caused by the DoS attack is transferred from i-1 to j-1. The specific expansion form of the transfer probability is The sum of the elements in each row is 1, that is, p ij ≥0, P is the transition probability matrix, p ij are the elements of the matrix P;
[0055] Step 5: Set the measurement data packet loss compensation strategy; for measurement compensation, use the successfully transmitted measurement to compensate for the lost measurement. arrive Lost in between For measurement data Indicates; for Zk The measured data after compensation is expressed as Z dk Indicates, where k represents the sampling time, t k If Z is the time it takes to successfully transmit dk It is expressed as follows:
[0056]
[0057] Step 6: Construct a distribution network state estimation model integrating the measurement data compensation strategy. The distribution network state estimation model selects nonlinear measurement equations and state equations. The distribution network state estimation model integrating the measurement data compensation strategy is as follows: the first is the state equation, and the second is the measurement equation.
[0058] x k =f(x k-1 )+q k-1
[0059] Z k =Z dk =h(x k )+r k
[0060] The state quantity x k It is the quantity to be sought in n-dimensional band, including bus voltage amplitude and bus voltage phase angle, x k-1 is x k The state quantity obtained at the last sampling moment, quantity measurement Z k It is an m-dimensional measurement that is measured by SCADA and then collected into the energy management system, including branch active and reactive power flow, bus active and reactive power and voltage amplitude; Z dk is the measurement of Z k Compensation after a denial of service attack. f(·) is the state prediction function, h(·) is the measurement matrix, which is determined by the admittance matrix of the bus system and Kirchhoff's law, q k-1 ,r k They are prediction error and measurement error, which are independent of each other;
[0061] Step 7: Construct a prediction function model; the state prediction function f(·) uses Holt's two-parameter exponential smoothing method, x k+1∣k is the state quantity at time k+1 predicted by the state quantity at time k, x k∣k-1 is the state quantity at time k predicted by the state quantity at time k-1, is an estimated value, S k and b k is an intermediate variable, α L , β L is the smoothing parameter, with a value between 0 and 1:
[0062]
[0063] Step 8: Use the improved Unscented Kalman Filter (UKF) method to update the state prediction;
[0064] Step 8.1: At time k-1, the sampling particles are selected based on the unscented transformation using the symmetric sampling strategy method. The sampling points are selected as shown in the following formula;
[0065]
[0066] λ=α 2 (n+κ)-n
[0067]
[0068] where χ i,k-1 is the sampling point selected at time k-1, i represents the sampling point number, a total of 2n+1 are selected, n is the dimension of the state quantity, is the estimated value of the state at time k-1, is the state covariance at time k-1, λ is the scaling factor, W i m is the weight of the mean, W i c is the weight of the variance, κ is the redundancy set to 0, and α is the parameter that changes the distribution distance of the particles;
[0069] Step 8.2: Substitute the sampling points sampled by the unscented transformation method into the defined measurement equation and state equation to obtain the predicted state value and predicted measurement value at the predicted time k. is the predicted quantity measure obtained using the UKF method, is the predicted state quantity obtained by the UKF method;
[0070]
[0071] where χ i,k is the state prediction function f(·) and the sampling point χ selected at time k-1 i,k-1 The predicted sampling point at time k is obtained; Z i,k is the predicted measurement point at time k;
[0072] Step 8.3: By finding the covariance of the state estimate and the measurement estimate Estimated variance of the quantity measured at time k Find the Kalman gain matrix based on the UKF method yes The inverse matrix of T is the transpose of the matrix, Rk is the measurement noise covariance matrix.
[0073]
[0074] Step 8.4: At time k, sample to obtain a new measurement value Z k Afterwards, the compensation strategy adopted for the measured data is used according to step 5, Z dk is the measured value after compensation. Then, according to step 8, the Kalman gain correction state quantity based on the UKF method is obtained. is the state quantity after correction, and the state quantity update steps are as follows.
[0075]
[0076] Step 9: Update state prediction using the extended Kalman filter (EKF) method combined with data packet loss compensation;
[0077] Step 9.1: Use the extended Kalman filter (EKF) method to re-predict the state estimate obtained by UKF to improve the prediction accuracy, as shown in the following formula.
[0078]
[0079] Step 9.2: Correct and obtain the final state estimate as shown below.
[0080]
[0081] Among them, Q k is the system noise covariance matrix, R k is the measurement noise covariance matrix, F k G K is the Jacobian matrix of the system model, H k and U k is the Jacobian matrix of the measurement equation, is the Kalman gain obtained by the EKF method, is the predicted state covariance matrix obtained by the EKF method, and finally are the state covariance matrix and estimated value at time k respectively.
[0082] This embodiment uses the IEEE 30-node distribution network to verify the proposed dynamic state estimation method. The network topology is as follows: Figure 3 shown.
[0083] The mean of the results of multiple power flow calculations is used as the true value of the state, and the simulation of the measured real data is based on the true value and superimposed with a random noise error with a normal distribution standard deviation of 0.05.
[0084] The maximum number of packet losses M is set to 2, and the probability of setting the number of packet losses is uniformly random, that is, the probability of transferring from packet loss 0 to packet loss 0, 1, or 2 is equal, the probability of transferring from packet loss 1 to packet loss 0, 1, or 2 is equal, and the probability of transferring from packet loss 2 to packet loss 0, packet loss 1, or 2 is equal.
[0085]
[0086] The proposed algorithm is used for state estimation under the consideration of data packet loss, and 80 sampling periods are simulated.
[0087] Figure 4 It is the voltage amplitude and phase angle curve corresponding to bus No. 5. The solid line is the estimated state value calculated, and the dotted line is the true state value obtained through power flow calculation.
[0088] from Figure 4 The simulation results show that the estimated voltage amplitude and voltage phase fluctuate around the true value, with little difference from the true value, and no point with a large difference from the true value occurs, which proves that the proposed compensation strategy can still perform state estimation well after the attack occurs. This proves the effectiveness of the present invention.
[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.
Claims
1. A method for dynamic state estimation of a distribution network to resist random data packet loss, characterized in that: The following steps are involved: Step 1: Use the Markov jump model to characterize the packet loss of distribution network measurement data, propose a measurement data loss compensation strategy, and establish a Markov chain model; The Markov chain is a Markov process The parameter set T is a discrete time set, The state space composed of the values of is a discrete state I = {m0,m1…,m k ,m k+1 ,…}, with a random process If for any integer and any m0,m1…,m k ,m k+1 ∈I, its conditional probability satisfies Then it is called is a Markov chain, where Pr represents probability, parameter set T is a discrete time set, t k is a value in the parameter set T, The state space composed of the values of is a discrete state, I = {m0,m1…,m k ,m k+1 ,…}, I is the state set, is a set of integers; Step 2: Establish a model of SCADA measurement data packet loss; Assuming the sampling period is a, the set of received measurements without DoS attack and without loss is defined as S1 = {Z1, Z2, Z3, ···Z k ···}, the set of received measurements after the attack caused data packet loss is in The number of consecutive measurements lost is represented by the set I = {0, 1, 2, ..., M}, where M represents the maximum number of measurement data lost. Assume that the most recently received measurement is The next received measurement is make but express The number of measurements lost between Step 3: Define the Markov chain transition probability. In the complete probability space, if the data packet loss process It takes values in the finite probability space M = {0,1,2,...,M} and is a discrete Markov random process with a transition probability of It is called a Markov packet loss process, where the transition probability constraint is Step 4: Characterize the probability of random transition of packet loss in the measurement process. The random transition probability of packet loss in the measurement process under a denial of service attack is expressed as T gives the energy constraint of data packet loss; p ij It represents the probability that the measurement loss caused by the denial of service attack is transferred from i-1 to j-1. The specific expansion form of the transfer probability is The sum of the elements in each row is 1, that is, p ij ≥0, P is the transition probability matrix, p ij is an element of the matrix P; Step 5: Set the measurement data packet loss compensation strategy; for measurement compensation, use the successfully transmitted measurement to compensate for the lost measurement. arrive Lost in between For measurement data Indicates; for Z k The measured data after compensation is expressed as Z dk Indicates, where k represents the sampling time, t k If Z is the time it takes to successfully transmit dk It is expressed as follows: Step 6: Construct a distribution network state estimation model integrating the measurement data compensation strategy. The distribution network state estimation model selects nonlinear measurement equations and state equations. The distribution network state estimation model integrating the measurement data compensation strategy is as follows: the first is the state equation, and the second is the measurement equation. x k =f(x k-1 )+q k-1 Z k =Z dk =h(x k )+r k The state quantity x k It is the quantity to be sought in n-dimensional band, including bus voltage amplitude and bus voltage phase angle, x k-1 is x k The state quantity obtained at the last sampling moment, quantity measurement Z k It is an m-dimensional measurement that is measured by SCADA and then collected into the energy management system, including branch active and reactive power flow, bus active and reactive power and voltage amplitude; Z dk is the measurement of Z k Compensation after a denial of service attack, f(·) is the state prediction function, h(·) is the measurement matrix, which is determined by the admittance matrix of the bus system and Kirchhoff's law, q k-1 ,r k They are prediction error and measurement error, which are independent of each other; Step 7: Construct a prediction function model; the state prediction function f(·) uses Holt's two-parameter exponential smoothing method, x k+1∣k is the state quantity at time k+1 predicted by the state quantity at time k, x k∣k-1 is the state quantity at time k predicted by the state quantity at time k-1, is an estimated value, S k and b k is an intermediate variable, α L , β L is the smoothing parameter, with a value between 0 and 1: Step 8: Use the improved unscented Kalman filter UKF method to update the state prediction; Step 9: Combine the extended Kalman filter (EKF) method with data packet loss compensation to update the state prediction and obtain the estimated value of the distribution network state.
2. A method for estimating dynamic state of a distribution network against random data packet loss according to claim 1, characterized in that: Step 8 specifically includes: Step 8.1: At time k-1, the sampling particles are selected based on the unscented transformation using the symmetric sampling strategy method. The sampling points are selected as shown in the following formula; λ=a 2 (n+k)-n where χ i,k-1 is the sampling point selected at time k-1, i represents the sampling point number, a total of 2n+1 are selected, n is the dimension of the state quantity, is the estimated value of the state at time k-1, is the state covariance at time k-1, λ is the scaling factor, W i m is the weight of the mean, W i c is the weight of the variance, κ is the redundancy set to 0, and α is the parameter that changes the distribution distance of the particles; Step 8.2: Substitute the sampling points sampled by the unscented transformation method into the defined measurement equation and state equation to obtain the predicted state value and predicted measurement value at the predicted time k. is the predicted quantity measure obtained using the UKF method, is the predicted state quantity obtained by the UKF method; where χ i,k is the state prediction function f(·) and the sampling point χ selected at time k-1 i,k-1 The predicted sampling point at time k is obtained; Z i,k is the predicted measurement point at time k; Step 8.3: By finding the covariance of the state estimate and the measurement estimate Estimated variance of the quantity measured at time k Find the Kalman gain matrix based on the UKF method yes The inverse matrix of T is the transpose of the matrix, R k is the measurement noise covariance matrix; Step 8.4: At time k, sample to obtain a new measurement value Z k Afterwards, the compensation strategy adopted for the measured data is used according to step 5, Z dk is the measured value after compensation, and then the Kalman gain correction state quantity based on the UKF method is obtained according to step 8, is the state quantity after correction, and the state quantity update steps are as follows:
3. A method for estimating dynamic state of a distribution network against random data packet loss according to claim 2, characterized in that: Step 9 specifically includes: Step 9.1: Use the extended Kalman filter (EKF) method to re-predict the state estimate obtained by UKF to improve the prediction accuracy, as shown in the following formula: Step 9.2: Correct the final state estimate as shown below: Among them, Q k is the system noise covariance matrix, R k is the measurement noise covariance matrix, F k G K is the Jacobian matrix of the system model, H k and U k is the Jacobian matrix of the measurement equation, K kEKF is the Kalman gain obtained by the EKF method, is the predicted state covariance matrix obtained by the EKF method, and finally are the state covariance matrix and estimated value at time k respectively.
