Privacy protection state estimation method and system for data acquisition of intelligent electric meter

Abstract

The invention belongs to the technical field of state estimation of a power system, particularly relates to a privacy protection state estimation method and system for data acquisition of an intelligent electric meter, and aims to establish a discrete time dynamic system model of a power distribution network in order to solve the problem that state estimation precision and resolution are reduced when privacy is protected by an existing method based on data aggregation. Each intelligent electric meter performs local state prediction based on the global state estimation of the previous period, and calculates a local information matrix and a vector; applying noise disturbance satisfying a differential privacy mechanism to the information vector and the matrix; the disturbed information is uploaded to a control center; the control center fuses all information and solves to obtain high-precision global state estimation at the current moment; and finally, broadcasting a result to each electric meter for prediction of the next period. Through disturbance and fusion of information increment, efficient unification of strict user privacy protection and power distribution network node-level refined state estimation is realized, and meanwhile, the method has the advantages of being dispersed in calculation, efficient in communication and easy to deploy.

Description

Technical Field

[0001] This invention belongs to the field of power system state estimation technology, specifically relating to a privacy-preserving state estimation method and system for smart meter data acquisition. Background Technology

[0002] With the rapid development of smart grids, smart meters widely deployed in Advanced Metering Systems (AMIs) can collect detailed electricity consumption data from the user side at high frequencies, providing an important foundation for advanced applications such as distribution network state estimation, load forecasting, and fault diagnosis. State estimation, as a core component of grid sensing and control, processes noisy measurement data to calculate the true operating state of grid nodes, such as voltage and phase angle. Its accuracy and real-time performance directly affect grid dispatching and safe operation.

[0003] In existing technologies, to balance data availability and user privacy, privacy-preserving state estimation methods based on data aggregation are commonly used. This method involves encrypting or noise-perturbing measurements such as active and reactive power collected from electricity meters locally. The encrypted data from multiple meters is then aggregated by a concentrator to form regional-level aggregated measurements, which are uploaded to the control center. Finally, a weighted least squares algorithm is used for state estimation. While this method avoids direct exposure of raw user data to some extent, the aggregation of measurement information at the data acquisition end results in significant "coarsening" of the information uploaded to the control center. This leads to a loss of spatial resolution and node-level accuracy in state estimation, making it difficult to meet the needs of refined operation and monitoring of distribution networks.

[0004] Therefore, how to achieve high-precision, node-level state estimation that is similar to plaintext data state estimation while ensuring that users' electricity privacy is not leaked has become a key issue that urgently needs to be addressed in the field of smart grid data security and state awareness. Summary of the Invention

[0005] To address the problems mentioned in the background technology, this invention proposes a privacy-preserving state estimation method and system for smart meter data acquisition. By adopting "information aggregation" instead of the traditional "data aggregation" paradigm, each meter only uploads information matrices and vectors that have undergone differential privacy perturbation, rather than the original measurement data or simple aggregated values. This cuts off the privacy leakage path at the source. At the same time, the central system can reconstruct the dynamic state of the entire network with high accuracy by fusing this information, effectively solving the contradiction in existing methods where privacy protection inevitably comes at the cost of estimation accuracy.

[0006] The technical solution adopted by this invention to solve its technical problem is: to provide a method for estimating the privacy protection status of smart meter data collection, comprising the following steps:

[0007] S1: Establish a discrete-time dynamic system model of the distribution network, including state equations and measurement equations;

[0008] S2: In each estimation period, each smart meter performs local state prediction based on the global state estimation of the previous period, and calculates the local information matrix and information vector based on the local measurement data.

[0009] S3: Apply differential privacy perturbation to the information vector to generate a privacy-protected information vector, and perform synchronous perturbation on the information matrix;

[0010] S4: Each smart meter uploads the disturbed information matrix and information vector to the control center;

[0011] S5: The control center fuses all uploaded information matrices and information vectors to calculate the global state estimate;

[0012] S6: The control center broadcasts the global state estimate to each smart meter for local state prediction in the next cycle.

[0013] Furthermore, the discrete-time dynamic system model of the distribution network in S1 is as follows:

[0014] Equations of state:

[0015] x(k+1)=Ax(k)+ω(k);

[0016] Measurement equation:

[0017] y i (k)=C i x(k)+v i (k);

[0018] Where x(k) is the system state vector, A is the state transition matrix, ω(k) is the process noise, and y i (k) is the measured value of smart meter i, C i For measurement function, v i (k) represents the measurement noise.

[0019] Furthermore, the local state prediction formula in S2 is:

[0020]

[0021] in, For the local predicted state, κ ii Let N be the estimator gain matrix. i (k) represents the network topology coupling information item.

[0022] Furthermore, the formula for calculating the local information matrix in S2 is as follows:

[0023] Ω i (k)=H i (k)⊤R i−1 H i (k);

[0024] The formula for calculating the local information vector is:

[0025] q i (k)=H i (k)⊤R i−1 [z i (k)−h i (x i −(k))+H i (k)x i−(k) ];

[0026] Among them, H i (k) is the Jacobian matrix of the measurement function, R i To measure noise v i The covariance matrix of (k), z i (k) represents the actual measured value.

[0027] Furthermore, the differential privacy perturbation in S3 specifically refers to:

[0028] Calculate the information vector q i The sensitivity Δq of (k);

[0029] Generate Laplace noise vector η i (k), where ε is the privacy budget;

[0030] Calculate the information vector after perturbation: ;

[0031] And the information matrix Ω i (k) Applying the same perturbation yields Ω i (k).

[0032] Furthermore, the global state estimation calculation formula in S5 is as follows:

[0033] ;

[0034] .

[0035] Furthermore, the global state estimate broadcast in S6 includes the voltage magnitude and phase angle of each node.

[0036] This application also provides a privacy-preserving status estimation system for smart meter data collection, comprising:

[0037] Multiple smart meters are used to perform local state prediction, information calculation, and privacy perturbation;

[0038] The control center is used to receive information after disturbances and to perform global state fusion and broadcasting.

[0039] Compared with the prior art, the beneficial effects of the present invention are:

[0040] (1) This invention replaces the traditional “data aggregation” paradigm with “information aggregation”. Each meter only uploads information matrices and vectors that have been subjected to differential privacy perturbation, rather than the original measurement data or simple aggregated values. This cuts off the privacy leakage path from the source. At the same time, the center can reconstruct the dynamic state of the entire network with high accuracy by integrating this information, which effectively solves the contradiction that protecting privacy must sacrifice estimation accuracy in the existing methods.

[0041] (2) This invention eliminates the “information coarsening” process caused by regional data aggregation. The control center can directly integrate information from each node to obtain a fine state estimation result at the level of node voltage amplitude and phase angle that is similar to that based on plaintext data estimation, which significantly improves the spatial resolution and accuracy of power distribution network state perception.

[0042] (3) The distributed collaborative framework adopted in this invention distributes most of the computation (such as state prediction, information computation and privacy perturbation) to be executed locally on each smart meter, which greatly reduces the centralized computing burden and communication bandwidth requirements of the control center. It can still run efficiently when the system scales up, and is suitable for large-scale smart meter deployment scenarios.

[0043] (4) This invention does not require changes to the basic measurement functions and hardware of smart meters. It can be achieved by embedding algorithm software in the existing meter calculation module and upgrading the control center state estimator, which is convenient for deployment and promotion in the existing Advanced Metering System (AMI). Attached Figure Description

[0044] Figure 1 A flowchart of a privacy protection status estimation method for smart meter data collection;

[0045] Figure 2 A flowchart illustrating a privacy-preserving status estimation method for smart meter data collection;

[0046] Figure 3 This is a system architecture diagram of a privacy-preserving state estimation system for smart meter data acquisition. Detailed Implementation

[0047] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] Example 1:

[0049] A method for estimating the privacy protection status of smart meter data collection, such as Figure 1 and Figure 2 As shown, it includes the following steps:

[0050] S1: Establish a discrete-time dynamic system model of the distribution network, including state equations and measurement equations;

[0051] First, a discrete-time dynamic system model is established for the distribution network, which forms the basis for subsequent state estimation. The model includes state equations and measurement equations.

[0052] Equations of state:

[0053] x(k+1)=Ax(k)+ω(k);

[0054] Where: x(k)∈Rn represents the state vector of the system at time k, which usually includes the voltage magnitude and phase angle of each node, i.e., x(k)=[V1(k),θ1(k),V2(k),θ2(k),...,V nb (k),θ nb (k)]T,n b Let n be the number of nodes, n = 2n b .

[0055] A∈R n×n It is the state transition matrix, which describes the law of system state evolution over time. In the distribution network, if the state change between adjacent time points is assumed to be gradual, it can usually be approximated as A=I (identity matrix), or modeled according to the physical characteristics of the system (such as on-load tap-changing transformers, capacitor bank switching, etc.).

[0056] ω(k)∈R n It is the process noise vector, assumed to be zero-mean Gaussian white noise, with a covariance matrix of Q(k), i.e., ω(k) ~ N(0,Q(k)), used to model the uncertainty of state prediction.

[0057] Measurement equation: y i (k)=C i x(k)+v i (k);

[0058] Where: y i (k)∈R miThis represents the measurement vector (such as node injected power, branch power, voltage amplitude, etc.) collected by the i-th smart meter at time k.

[0059] C i ∈R mi×n This is the measurement function matrix corresponding to the i-th smart meter. Its elements are determined by the grid admittance matrix and network topology, establishing a nonlinear relationship between the system state and local measurement values. In specific calculations, its Jacobian matrix is ​​typically used. Perform linearization, where h i (·) is C i The corresponding nonlinear function form.

[0060] v i (k)∈R mi Let R be the measurement noise vector of the i-th smart meter, assumed to be zero-mean Gaussian white noise, and the noise of different meters is independent of each other, with its covariance matrix being R. i That is, v i (k)~N(0,R i ).

[0061] S2: In each estimation period, each smart meter performs local state prediction based on the global state estimation of the previous period, and calculates the local information matrix and information vector based on the local measurement data.

[0062] In each estimation period k, each smart meter i performs the following operations in parallel:

[0063] Local state prediction: Based on the global state estimate x^(k-1) broadcast from the control center at the end of the previous cycle (k-1), and local network topology coupling information (such as the influence of the states of adjacent nodes), smart meter i makes a one-step prediction of the system state in the current cycle k.

[0064] ;

[0065] in, This is the local prediction of the state of meter i at time k. (NetworkCoupling) i The (k-1) term reflects the coupling effects of the network topology and can be modeled based on the historical state estimates of neighboring nodes. A simplified form is to ignore strong coupling and directly use... .

[0066] Calculating the local information matrix and information vector: This is the core of distributed information filtering. Meter i utilizes its local measurement value z. i (k) (i.e., y) i (k) Actual reading, predicted status And the measurement model, calculating its "contribution" to the global state estimation.

[0067] Local information matrix (an approximation of the local Fisher information matrix):

[0068] Ω i (k)=H i T(k)R i-1 H i (k) where H i (k) is the measurement function at... The Jacobian matrix at Ω. i (k) is an n×n matrix that reflects the contribution of local measurements to the accuracy of state estimation.

[0069] Local information vector:

[0070] q i (k)=H i (k)⊤R i−1 [z i (k)−h i (x i −(k))+H i (k)x i−(k) ]

[0071] Among them, z i (k)−h i (x^i-(k)) is the measurement residual (innovation). This formula linearizes the nonlinear measurement equation at the current prediction point through a first-order Taylor expansion, making q i (k) has an approximately linear relationship with state x(k).

[0072] S3: Apply differential privacy perturbation to the information vector to generate a privacy-protected information vector, and perform synchronous perturbation on the information matrix;

[0073] To protect user privacy and prevent eavesdroppers from the control center or communication links from accessing the q i (k) and Ω i The original measurement value z is derived from (k). i (k) The present invention applies noise to both of them to satisfy the differential privacy mechanism.

[0074] Determine privacy budget and sensitivity:

[0075] Privacy budget (ϵ): Set by the system to control the strength of privacy protection. The smaller ϵ is, the more noise is added, the stronger the privacy, but the greater the loss of estimation accuracy.

[0076] Information vector sensitivity (Δq): Calculate the information vector q i (k) The maximum possible norm of change for any single measurement value. Since q i (k) is zi The sensitivity Δq of the linear function (k) after linearization i According to H i T(k)R i -1 and the upper limit of the measurement range are calculated analytically or conservatively estimated.

[0077] Generate and add noise:

[0078] Generate a with q i (k) Laplace noise vectors η of the same dimension i (k), whose components are independently and identically distributed in the Laplace distribution Lap(0,Δq). i / ϵ).

[0079] Calculate the privacy-preserving information vector:

[0080] Q i (k)=q i (k)+η i (k);

[0081] To ensure consistency between the information matrix and the information vector perturbation (and avoid introducing mathematical contradictions), the information matrix Ω is... i (k) also adds a carefully designed perturbation matrix Ξ i (k), to obtain the perturbed information matrix Ω i (k). The construction of the perturbation matrix Ξi(k) must ensure that its expectation is zero, and that its "strength" is related to η. i The statistical properties of (k) match. One feasible solution is to add a symmetric random matrix that satisfies a specific distribution.

[0082] At this point, the original local measurement information of the smart meter is hidden in the noise.

[0083] S4: Each smart meter uploads the disturbed information matrix and information vector to the control center;

[0084] Each smart meter transmits the calculated Ω through its communication module. i (k) and q i (k) After encryption, upload to the control center.

[0085] S5: The control center fuses all uploaded information matrices and information vectors to calculate the global state estimate;

[0086] After receiving data from all (assuming a total of N) smart meters, the control center performs a data aggregation calculation. Information aggregation:

[0087] ;

[0088] According to the principle of information filtering, the aggregated Ω(k) and q(k) contain the measurement information of all meters (although disturbed).

[0089] Global state estimation solution:

[0090] ;

[0091] Solving this system of linear equations will yield the global state estimate for the current period k. Since the added Laplace noise has a zero mean, the impact of noise on the estimation results will be averaged out when a large amount of meter data is aggregated. Thus, while ensuring differential privacy, a statistically significant high-precision state estimate can still be obtained.

[0092] S6: The control center broadcasts the global state estimate to each smart meter for local state prediction in the next cycle.

[0093] The control center will calculate the global state estimate. (Typically, the voltage amplitude and phase angle of each node are extracted) and broadcast to all smart meters in the network. Each meter receives and stores the result, using it as the initial value for local state prediction in step S2 for the next estimation period (k+1).

[0094] At this point, a complete estimation cycle ends, and the system enters the next cycle k+1, repeating S2 to S6.

[0095] Example 2:

[0096] A smart meter data acquisition and privacy protection state estimation system for implementing the above method, such as Figure 3 As shown, it includes:

[0097] Multiple smart meters are used to perform local state prediction, information calculation, and privacy perturbation;

[0098] The control center is used to receive information after disturbances and to perform global state fusion and broadcasting.

[0099] This system mainly includes two types of entities: multiple smart meters (deployed on the user side or distribution nodes) and a control center (usually located on the distribution network operator side).

[0100] Smart meters have the following functional modules:

[0101] Measurement module: Real-time acquisition of electrical measurement values ​​such as voltage, current, and power.

[0102] Computation module: Built-in processor, capable of performing algorithms such as local state prediction, information matrix and information vector calculation, and differential privacy perturbation.

[0103] Communication module: Supports two-way communication with the control center (such as using power line carrier communication, wireless cellular network, fiber optic, etc.), used to upload data after disturbance and receive global state estimation results.

[0104] Storage module: Used to temporarily store parameters required for local calculations, global state estimates from the previous cycle, etc.

[0105] The control center has the following functional modules:

[0106] Communication aggregation module: Receives disturbance information uploaded from all smart meters.

[0107] Data fusion and estimation module: High-performance server or computing cluster, used to execute information fusion and global state estimation algorithms.

[0108] Broadcast module: Broadcasts the calculated global state estimate to all smart meters.

[0109] Database: Stores historical state estimation results, system parameters, network topology information, etc.

[0110] The system operates in a discrete time frame, with each estimation period having a fixed duration (e.g., 1 minute, 5 minutes, etc.), denoted as time index k.

[0111] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for estimating the privacy protection status of smart meter data collection, characterized in that, Includes the following steps: S1: Establish a discrete-time dynamic system model of the distribution network, including state equations and measurement equations; S2: In each estimation period, each smart meter performs local state prediction based on the global state estimation of the previous period, and calculates the local information matrix and information vector based on the local measurement data. S3: Apply differential privacy perturbation to the information vector to generate a privacy-protected information vector, and perform synchronous perturbation on the information matrix; S4: Each smart meter uploads the disturbed information matrix and information vector to the control center; S5: The control center fuses all uploaded information matrices and information vectors to calculate the global state estimate; S6: The control center broadcasts the global state estimate to each smart meter for local state prediction in the next cycle.

2. The method for estimating the privacy protection status of smart meter data collection according to claim 1, characterized in that, The discrete-time dynamic system model of the distribution network in S1 is as follows: Equations of state: x(k+1)=Ax(k)+ω(k); Measurement equation: y i (k)=C i x(k)+v i (k); Where x(k) is the system state vector, A is the state transition matrix, ω(k) is the process noise, and y i (k) represents the measured value of smart meter i, C i For measurement function, v i (k) represents the measurement noise.

3. The method for estimating the privacy protection status of smart meter data collection according to claim 1, characterized in that, The local state prediction formula in S2 is: ； in, For the local predicted state, κ ii Let N be the estimator gain matrix. i (k) represents the network topology coupling information item.

4. The method for estimating the privacy protection status of smart meter data collection according to claim 1, characterized in that, The formula for calculating the local information matrix in S2 is as follows: Ω i (k)=H i (k)⊤R i−1 H i (k); The formula for calculating the local information vector is: q i (k)=H i (k)⊤R i−1 [z i (k)−h i (x i −(k))+H i (k)x i−(k) ]; Among them, H i (k) is the Jacobian matrix of the measurement function, R i To measure noise v i The covariance matrix of (k), z i (k) represents the actual measured value.

5. The method for estimating the privacy protection status of smart meter data collection according to claim 1, characterized in that, The differential privacy perturbation in S3 is specifically as follows: Calculate the information vector q i The sensitivity Δq of (k); Generate Laplace noise vector η i (k), where ε is the privacy budget; Calculate the information vector after perturbation: ； And the information matrix Ω i (k) Applying the same perturbation yields Ω i (k).

6. The method for estimating the privacy protection status of smart meter data collection according to claim 1, characterized in that, The global state estimation calculation formula in S5 is as follows: ； 。 7. The method for estimating the privacy protection status of smart meter data collection according to claim 1, characterized in that, The global state estimate broadcast in S6 includes the voltage magnitude and phase angle of each node.

8. A privacy-preserving status estimation system for smart meter data acquisition using the method described in any one of claims 1-7, characterized in that, include: Multiple smart meters are used to perform local state prediction, information calculation, and privacy perturbation; The control center is used to receive information after disturbances and to perform global state fusion and broadcasting.

Images