Intelligent electric meter data security aggregation method and system for power load prediction

Abstract

The invention discloses a power load prediction-oriented intelligent electric meter data security aggregation method and system, and the method comprises the steps: configuring a node for each intelligent electric meter, and constructing a hierarchical network topology of a power wireless sensor network; each node collects data of the intelligent electric meter, and an iteration state initial value of consistency aggregation of each node is obtained; each node adds initial noise to an iteration state initial value, broadcasts noise-added information to other nodes according to a preset communication radius, receives noise-added state information of neighbor nodes, carries out consistency iteration on own nodes, stops iteration until an iteration precision requirement is met, and obtains an average value of the whole hierarchical network topology; based on the average value of the hierarchical network topology and the number of network nodes, the total power load value of the shared area is calculated, and safe aggregation of the intelligent electric meter data is completed. The convergence speed and performance of data aggregation are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric power data collection, and in particular, relates to a method and system for securely aggregating smart meter data for electric power load prediction. Background Art

[0002] Power load forecasting plays a vital role in the construction of new power systems. Accurate power load forecasting can provide technical support for efficient power management and power optimization and dispatching, but it requires a large amount of historical power load data to expand data features. Smart meters use digital technology to measure, calculate, collect, process and transmit electric energy, collect a large amount of fine-grained load data from the user side, and realize various load profile analyses, which is an important source of power data.

[0003] However, smart meters cover the privacy information of users, including their daily habits and lifestyles. In smart meter data perception for power load forecasting, how to protect user data privacy is an urgent problem to be solved.

[0004] Prior art 1 - the paper "State Grid Power Data Sharing Scheme Based on Combination of CKKS and CP-ABE Attribute Encryption" (Information Security Research, 2023, 9 (03): 262-270, Zhang Haijun et al., March 31, 2023), encrypts data through KS to achieve secure data transmission and calculation, and implements data access control and one-to-many data transmission and sharing through the CP-ABE algorithm to improve data security sharing efficiency. Prior art 2 - the paper "Privacy-preserving and efficient aggregation based on blockchain for power grid communications insmart communities" (IEEE Communications Magazine, Guan Z, Si G, Zhang X, et al., 2018, 56 (7): 82-88, July 25, 2018), divides users into several groups, and each group of users has a private blockchain to record member data, so as to ensure data security. Prior Art 3 - The paper "A Power Data Sharing Scheme Based on Blockchain and Attribute-Based Encryption" (Power System Protection and Control, Yang Xiaodong et al., 2023, 51 (13): 169-176, July 1, 2023), combines attribute-based encryption and symmetric encryption algorithms to achieve confidentiality and fine-grained access control of power data, uploads attribute sets and hash values ​​of power data to the blockchain, and ensures the unforgeability of user attributes and the correctness of decrypted ciphertext. Prior Art 4 - The paper "DPWGAN: high-quality load profiles synthesis with differential privacy guarantees" (IEEE Transactions on Smart Grid, Huang J, Huang Q, Mou G, et al. 2022, 14 (4): 3283-3295, December 20, 2022), converts real data sets into high-quality synthetic load data sets, so that research and analysis on the synthetic data sets can automatically meet user-level differential privacy guarantees.

[0005] However, the above-mentioned encryption technology, blockchain technology and data synthesis technology can provide good privacy security, but there are still problems of high computational complexity and high computational overhead. A lightweight security collaboration method needs to be studied for power data.

[0006] In the data perception of smart meters for power load forecasting, secure aggregation technology that considers data privacy is an optional solution. Prior Art 5 - The paper "Privacy-preserving consensus-based energy management in smart grids" (IEEE Transactions on Industrial informatics, ZhaoC, Chen J, He J, et al. 2018, 15 (3): 6162-6176, October 9, 2018) proposed a privacy protection algorithm based on secret functions to protect the privacy information of power parameters, and explored two cases. One case is that it is assumed that the local secret function can only be obtained by adjacent nodes, and the other case is that each node uses real information to update the state and broadcasts noisy information. The zero-sum and exponential decay noise adding communication mechanisms in the two cases are given. Prior Art 6 - The paper "Real-time privacy-preserving data release for smart meters" (IEEE Transactions on Smart Grid, Liu Y, Guo W, Fan CI, et al. 2020, 11 (6): 5174-5183, June 30, 2020) focuses on real-time privacy threats, that is, potential attackers who attempt to infer sensitive information from smart meter data in an online manner, and adopts an information-theoretic privacy measure to provide a target level of privacy protection by adding a small amount of perturbation signals to the smart meter measurements. Prior Art 7 - The paper "Consensus-based data-privacypreserving data aggregation" (IEEE Transactions on Automatic Control, He J, CaiL, Cheng P, et al. 2019, 64 (12): 5222-5229, April 11, 2019) studies data aggregation with privacy protection based on consistency theory, and proposes a secure consistency data mining algorithm that ensures accurate data aggregation while ensuring the privacy of sensitive data, and provides a proof of the exact convergence of the proposed algorithm.

[0007] However, the above-mentioned noise addition or secret function methods do not consider the impact of the power network topology on the algorithm performance, and fail to study the performance improvement of the algorithm from the perspective of topological hierarchical control. Summary of the invention

[0008] In order to solve the deficiencies in the prior art, the present invention provides a smart meter data security aggregation method for power load forecasting to solve the problems of low convergence speed and low performance of data aggregation.

[0009] The present invention adopts the following technical solution.

[0010] A first aspect of the present invention provides a method for securely aggregating smart meter data for power load forecasting, comprising: Equip each smart meter with a node to build a hierarchical network topology for the power wireless sensor network; Each node collects smart meter data and obtains the initial value of the iterative state of the consistency aggregation of each node; Each node adds initial noise to the initial value of the iteration state, broadcasts the noisy information to other nodes within a preset communication radius, receives the noisy state information from neighboring nodes, performs consistency iteration on its own node, stops iteration after the iteration accuracy requirement is met, and obtains the average value of the entire hierarchical network topology; Based on the average value of the hierarchical network topology and the number of network nodes, the total power load value of the shared area is calculated to complete the secure aggregation of smart meter data.

[0011] Optionally, a hierarchical network topology is constructed for the power wireless sensor network, including: The communication relationship of the power wireless sensor network nodes is recorded as an undirected weighted graph ,in, is a collection of nodes, N Represents the number of nodes in the network, and the edge set Indicates the communication relationship between N child nodes; If the node Receive nodes information, then ,on the contrary , and there is no self-connection in the network, ; is the adjacency weight matrix of network G, and is defined Representation Node The input neighbor node set, for the adjacency weight matrix W, if , otherwise ; Define the nodes in the undirected weighted graph G The in-degree is , the in-degree matrix of G is , The Laplacian matrix of G is defined as , and satisfies .

[0012] Optionally, the hierarchical network topology includes one or more underlying networks and an upper network, wherein the upper network is composed of relay nodes of the underlying networks, and each underlying network is assigned a relay node.

[0013] Optionally, the secure aggregation of smart meter data includes underlying consistency iteration and upper-layer consistency iteration, wherein the underlying consistency iteration process is performed between nodes of the same underlying network, and the relay node calculates the total power load value of the underlying network based on the final iteration value of the underlying consistency iteration and the number of nodes in the underlying network; the upper-layer consistency iteration process is performed between relay nodes of the upper-layer network, and any relay node calculates the total power load value of the shared area based on the final iteration value of the upper-layer consistency iteration and the number of nodes in the upper-layer network.

[0014] Optionally, the bottom-level consistency iteration and the upper-level consistency iteration are cascaded, and the initial value of the iteration state of each node in the bottom-level consistency iteration is the smart meter data collected by each node; the initial value of the iteration state of each relay node in the upper-level consistency iteration is the total power load value of the bottom-level network corresponding to each relay node.

[0015] Optionally, during the node consistency iteration process, the node noise state information and the node state information are updated according to the following formula: (1) in, For the The node at the iteration The noise state information, For the The node at the iteration Node status information, For the The node at the iteration The updated value of the node status information, For the The node at the iteration The noise information is a random number that makes the network mean unchanged in each iteration. To update the gain, , For the The node at the iteration Neighbor nodes The noise state information, For Node and nodes The adjacency weight of is the number of network nodes.

[0016] Optional, node and nodes The adjacency weight Calculated as follows: (2) in, Respectively represent nodes and The in-degree, is the edge set of the network, representing the communication relationship between nodes in the network.

[0017] Optionally, in the cascade aggregation of the underlying consistency iteration and the upper consistency iteration, at the node Receive all neighbor nodes Bidirectional transmission Then, the node is Add noise information: (3) (4) in, For Node In the The noise information added in the iteration is For the The node Designed random numbers, For the The node Additional parameters for design, For Node To Node The parameters of the transmission, For Node To Node The parameters of the transmission, is a constant and satisfies , is the neighbor node of node j.

[0018] Optionally, the method further includes: calculating the node according to the following formula For all neighbor nodes The maximum privacy leakage is analyzed: (5) in, Privacy Disclosure For Node j Initial state Node i To estimate the accuracy The estimated probability, For Node All random noise A collection of For Node i For Node j Deviation The estimated value of , For Node j Initial state The estimated value of is the estimated accuracy, for Random noise at any moment The probability density function of For Node j The set of input neighbor nodes.

[0019] Optional, for each node Setting private parameter variance , The mean is 0 and the variance is of uniform distribution.

[0020] The second aspect of the present invention provides a smart meter data security aggregation system, including a construction module, a collection module, a consistency iteration module and a calculation module. Among them, the construction module is used to equip each smart meter with a node to build a hierarchical network topology of the power wireless sensor network. The collection module is used to collect smart meter data at each node and obtain the initial value of the iterative state of the consistency aggregation of each node. The consistency iteration module is used to add initial noise to each initial value of the iterative state, broadcast the noisy information to other nodes with a preset communication radius, and receive the noisy state information of neighboring nodes, perform consistency iteration on its own nodes, and stop the iteration after meeting the iteration accuracy requirements, and obtain the average value of the entire hierarchical network topology. The calculation module is used to calculate the total power load value of the shared area based on the average value of the hierarchical network topology and the number of network nodes, and complete the secure aggregation of smart meter data.

[0021] The third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is loaded into the processor, the method for securely aggregating smart meter data for power load forecasting is implemented.

[0022] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned smart meter data security aggregation method for power load forecasting.

[0023] Compared with the prior art, the beneficial effects of the present invention include at least: Aiming at the acquisition of smart meter data for power load forecasting, the present invention proposes a cascade consistency aggregation method based on hierarchical topology, which can effectively improve the convergence speed of data aggregation and reduce the communication energy consumption of data aggregation; at the same time, different noise variances are designed for smart meters to improve the privacy protection of smart meter data. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them: Figure 1 A schematic diagram of a noise-added consistency aggregation architecture for smart meter data for load prediction provided by an embodiment of the present invention; Figure 2 A schematic diagram of a cascade consistency data aggregation block diagram provided by an embodiment of the present invention; Figure 3 A schematic diagram of an upper layer communication topology based on IEEE-14 provided in an embodiment of the present invention; FIG4 (a) is a bottom layer network of a cascade consistency data aggregation provided by an embodiment of the present invention Schematic diagram of state iteration curve; FIG4( b ) is a cascade consistency data aggregation middle and upper layer network provided by an embodiment of the present invention Schematic diagram of state iteration curve; Figure 5 A schematic diagram of MSE iteration curves of three networks provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in this application are only embodiments of a part of the present invention, rather than all embodiments. Based on the spirit of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the protection scope of the present invention.

[0026] Aiming at the scenario of smart meter data acquisition for power load forecasting, this paper studies a lightweight power data security aggregation method based on hierarchical topology, replacing real value aggregation with average value security aggregation. It mainly includes the following aspects: first, construct a hierarchical network topology of power WSNs (Wireless Sensor Networks), divide the network into several underlying networks and determine several relay nodes, all relay nodes form the upper network; second, averagely aggregate the power data based on the average consistency theory, replace the node real value with the network mean, and protect the node data privacy from the method design level; finally, in the process of smart meter data aggregation based on average consistency, design noise for each node, replace the real value transmission with the noise disturbance value transmission, and protect the user's smart meter data from being inferred.

[0027] Combination Figure 1 to Figure 2 As shown, embodiment 1 of the present invention provides a smart meter data security aggregation method for power load prediction, including the following contents: S1. Equip each smart meter with a node to build a hierarchical network topology for the power wireless sensor network.

[0028] In S1, a hierarchical network topology of the power wireless sensor network is constructed, including: The communication relationship of the power wireless sensor network nodes is recorded as an undirected weighted graph ,in, is a collection of nodes, N Represents the number of nodes in the network, and the edge set express N The communication relationship between child nodes; if the node Receive nodes information, then ,on the contrary , and there is no self-connection in the network, ; is the adjacency weight matrix of network G, and is defined Representation Node The input neighbor node set, for the adjacency weight matrix W, if , otherwise ; Define the nodes in the undirected weighted graph G The in-degree is , the in-degree matrix of G is , the Laplacian matrix of G is defined as , and satisfies .

[0029] Furthermore, the hierarchical network topology includes one or more underlying networks and an upper network, wherein the upper network is composed of relay nodes of the underlying networks, and each underlying network is assigned a relay node.

[0030] Specifically, in order to ensure that the power consumption data transmission of power users has high reliability and real-time performance in terminal-dense areas, the power data sharing network topology (i.e., the power wireless sensor network topology) is layered, and the underlying network and the upper network are maintained based on the communication distance. The upper network is composed of the relay nodes of the underlying network, and the relay nodes are dynamically selected in the underlying network based on the energy balance principle.

[0031] When constructing a hierarchical communication topology, the embodiment of the present invention organizes nodes located in the same building or the same office building into an underlying network based on geographical location and communication distance, and allocates a relay node to each underlying network. The nodes complete the consistency aggregation of electricity consumption data based on short-range wireless communication technologies, including ZigBee, etc.; the upper-layer network is composed of relay nodes, and the nodes complete the consistency aggregation of data based on long-range wireless communication technologies, such as LoRa, etc.

[0032] Specifically, in the construction of a hierarchical network topology, the nodes of the network have unique identifiers. , and will not move after deployment; the node can estimate the approximate distance between it and the sending node based on RSSI (Received Signal Strength Indicator); when sending data, the node can automatically adjust its own transmission power according to the transmission distance requirements of the sending and receiving nodes; the nodes have different initial energies, but have the same communication and data processing capabilities.

[0033] S2. Each node collects smart meter data and obtains the initial value of the iterative state of the consistent aggregation of each node.

[0034] S3. Each node adds initial noise to the initial value of the iteration state, broadcasts the noisy information to other nodes within a preset communication radius, receives the noisy state information from neighboring nodes, performs consistency iteration on its own node, and stops iterating after the iteration accuracy requirement is met, and obtains the average value of the entire hierarchical network topology.

[0035] Combination Figure 2As shown, in S3, the secure aggregation of smart meter data includes underlying consistency iteration and upper consistency iteration. The underlying consistency iteration process is performed between nodes of the same underlying network, and the relay node calculates the total power load value of the underlying network based on the final iteration value of the underlying consistency iteration and the number of nodes in the underlying network; the upper consistency iteration process is performed between relay nodes of the upper network, and any relay node calculates the total power load value of the shared area based on the final iteration value of the upper consistency iteration and the number of nodes in the upper network.

[0036] Furthermore, the bottom-level consistency iteration and the upper-level consistency iteration are cascaded, and the initial value of the iteration state of each node in the bottom-level consistency iteration is the smart meter data collected by each node; the initial value of the iteration state of each relay node in the upper-level consistency iteration is the total power load value of the bottom-level network corresponding to each relay node.

[0037] In one embodiment, the power wireless sensor network is divided into The bottom layer network and the top layer network. o The underlying networks are used Indicates that the upper network uses express.

[0038] It should be noted that in the present invention, the network is uniformly recorded as To represent the underlying network Or upper network , nodes in the network j Is a node i The output neighbor nodes of .

[0039] Specifically, in each underlying network In all nodes Get electricity usage data from smart meters , which is used as the initial value of the iteration state, that is, For the upper network , whose nodes The initial value of is the total power load value of the corresponding underlying network.

[0040] Furthermore, to ensure the initial value of the iteration Not inferred, design initial noise , the information after superimposing the noise Broadcast to other nodes. At the iteration, the node Receive the noise status information of neighboring nodes ,Then according to the average consistency theory, the nodes are updated when the ,state is inconsistent based on the noise addition.

[0041] Furthermore, in S3, during the node consistency iteration process, the node noise state information and the node state information are updated according to the following formula: (1) in, For the The node at the iteration The noise state information, For the The node at the iteration Node status information, For the The node at the iteration The updated value of the node status information, For the The node at the iteration The noise information is a random number that makes the network mean unchanged in each iteration. To update the gain, , For the The node at the iteration Neighbor nodes The noise state information, For Node and nodes The adjacency weight of is the number of network nodes.

[0042] In this embodiment, first, all underlying networks In all nodes Get electricity usage data from smart meters , which is used as the initial value of the iteration state, that is, , and perform noise consistency iteration of the state until the iteration accuracy meets the requirements; then, all underlying networks The relay nodes in the network calculate the total power load value of the network and use it as the initial value of the upper iteration, and then perform noise consistency iteration of the state until the iteration accuracy meets the requirements; finally, any relay node calculates the total power load value of the entire area, and the meter data aggregation process ends, providing the implementation load data in a certain area to support power load forecasting. This embodiment uses a noisy consistency sharing method to replace the direct sharing of power consumption data with the network mean, which protects the data privacy of the nodes in terms of sharing and improves security. In addition, the impact of the power network topology on the performance of the data aggregation algorithm is taken into account. By constructing a hierarchical network topology of the power wireless sensor network, the convergence speed of data aggregation is improved and the energy consumption of data aggregation is reduced.

[0043] Optional, node and nodes The adjacency weight Calculated as follows: (2) in, Respectively represent nodes and The in-degree, is the edge set of the network, representing the communication relationship between nodes in the network.

[0044] Optionally, in the cascade aggregation of the underlying consistency iteration and the upper consistency iteration, at the node Receive all neighbor nodes Bidirectional transmission Then, the node is Add noise information: (3) (4) in, For Node In the The noise information added in the iteration is For the The node Designed random numbers, For the The node Additional parameters for design, For Node To Node The parameters of the transmission, For Node To Node The parameters of the transmission, is a constant and satisfies , is the neighbor node of node j.

[0045] The absolute upper limit is as the number of iterations increases. exponentially decaying, and The value size represents the ability to adjust its attenuation speed.

[0046] In this embodiment, the privacy of the transmission process can be guaranteed through bidirectional transmission, thereby improving the privacy of the iteration initial value.

[0047] Furthermore, the iteration accuracy is required to be that the mean square error of the node status is less than 0.1.

[0048] It should be noted that those skilled in the art may set the iteration accuracy requirement according to actual needs, and the embodiments of the present invention do not limit the specific content of the iteration accuracy requirement.

[0049] S4. Based on the average value of the hierarchical network topology and the number of network nodes, the total power load value of the shared area is calculated to complete the secure aggregation of smart meter data.

[0050] Optionally, the method further includes: Calculate the nodes according to the following formula For Node The maximum privacy leakage is analyzed: (5) Among them, privacy leakage For Node j Initial state Node i To estimate the accuracy The estimated probability, For Node j All random noise A collection of For Node i For Node j Deviation The estimated value of , For Node j Initial state Node i Estimated value of the estimate, is the estimated accuracy, for Random noise at any moment The probability density function of For Node j The set of input neighbor nodes.

[0051] Furthermore, for each node Set the private parameter mean square error , The mean is 0 and the variance is of uniform distribution.

[0052] Specifically, The mean is 0 and the variance is When the distribution is uniform, the maximum privacy leakage degree in the embodiment of the present invention is: (11) From (11), we can see that in the estimation error Under certain circumstances, the representative node j Noise variance The larger the value, the greater the added noise intensity. j Initial value The smaller the estimated probability, the j The smaller the privacy leakage, the better the privacy of the node initial value.

[0053] In this embodiment, the added noise is designed according to a uniform distribution. At the same time, a different noise variance is designed for each smart meter, which improves the privacy protection of the initial value from the perspective of personalized variance design and further improves the privacy protection of smart meter data.

[0054] Optionally, the method further includes: Convergence analysis of the consistent iterative aggregation algorithm is performed.

[0055] Specifically, for a random sequence , if any , then the sequence is expected to converge to ; If for all Both ,and , then the mean square error of the sequence converges to ,in represents mathematical expectation.

[0056] In this embodiment, , and is the system matrix of network G, where is the N-dimensional unit matrix, For the network No. The mean value at the iteration, For the network No. The difference between the mean of the iteration and the mean of the initial value, that is, . (1) can be organized into the following matrix form: (6) In the weight matrix W When designed according to (2), the matrix is a doubly random matrix, so Therefore, (6) can be written as:

[0057] Next, we analyze the expected convergence of the consistent iterative aggregation algorithm.

[0058] From (3) and (4), we can get and . So we have: The expected convergence of the proposed algorithm is proved.

[0059] Next, the mean square error convergence of the consistency iterative aggregation algorithm is carried out.

[0060] First give Proof. For the consistent iterative aggregation algorithm (1), we can get the state The solution is:

[0061] Note that when hour, . So we have:

[0062] In the above formula, it is obvious that Established.

[0063] remember For the network k The difference between the state at the iteration and the initial state of the network. W When designed according to (2), the matrix A is a double random matrix, which can be sorted out as follows: (7) In formula (7), ,in .exist and When designed according to (3) and (4) respectively, It can be further written as:

[0064] in , Due to the added noise is random noise, so for the node , as well as , and Independent and identically distributed. Therefore, can be written as: (8) Notice and is a symmetric matrix, and Established. According to the Rayleigh-Ritz theorem, (8) can be further written as: (9) in and Respectively represent matrices A The Frobenius norm and trace of , and the last inequality holds because N dimensional symmetric matrix ,have Established. .because ,have . (9) can be written as: (10) It shows that the proposed consistency iterative aggregation algorithm converges to mean square consistency.

[0065] Embodiment 2 of the present invention provides a smart meter data security aggregation system, which runs the smart meter data security aggregation method for power load prediction as described in embodiment 1, and the system includes: a construction module, a collection module, a consistency iteration module and a calculation module. Among them, the construction module is used to equip each smart meter with a node to build a hierarchical network topology of the power wireless sensor network. The collection module is used to collect smart meter data at each node and obtain the initial value of the iterative state of the consistency aggregation of each node. The consistency iteration module is used to add initial noise to each initial value of the iterative state, broadcast the noisy information to other nodes with a preset communication radius, and receive the noisy state information of the neighboring nodes, perform consistency iteration on its own nodes, stop iteration after meeting the iteration accuracy requirements, and obtain the average value of the entire hierarchical network topology. The calculation module is used to calculate the total power load value of the shared area based on the average value of the hierarchical network topology and the number of network nodes, and complete the secure aggregation of smart meter data.

[0066] Embodiment 3 of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded into the processor, the method for secure aggregation of smart meter data for power load forecasting described in Embodiment 1 is implemented.

[0067] Embodiment 4 of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for secure aggregation of smart meter data for power load forecasting according to Embodiment 1 is implemented.

[0068] Embodiment 5 of the present invention provides a verification embodiment of a smart meter data security aggregation method for power load forecasting.

[0069] This example verifies the effectiveness of the proposed iterative aggregation method for power data consistency on a Python simulation platform. First, the network topology generation result is given, then the convergence of the proposed method is verified, and the influence of the iteration factor on the convergence speed is compared and analyzed, and then the privacy protection of the algorithm is analyzed. All curves are drawn according to the average of 1000 running results.

[0070] The experimental data comes from the Electricity Load Data (ECL), and the original data contains the hourly electricity consumption data of 321 users from 2012 to 2014, in kilowatt-hours. In the experiment, the data of 100 users at 0:00 in 2012 are used as smart meter data, and the total electricity consumption is 47,342 kWh.

[0071] The embodiment of the present invention constructs a hierarchical communication topology (i.e., the hierarchical network topology of the power wireless sensor network in S1) based on the IEEE-14 system. In IEEE-14, 9 nodes 4, 5, 7, 9, 10, 11, 12, 13, and 14 are load nodes, and the remaining 5 nodes represent generators. For each load node, it is assumed that it collects the total number of power loads in a local area. The number of underlying networks where each load node is located is randomly selected between [5, 15], and it is guaranteed that the sum of the numbers of the 9 underlying networks is 100. The communication topology between all 9 underlying network nodes is generated as an undirected connected graph with the in-degree of all nodes being 3. The communication topology of the upper network composed of 9 load nodes is as follows: Figure 3 As shown, it is a ring communication topology, and the in-degree of all nodes is 2.

[0072] Next, we conduct convergence analysis and performance comparison.

[0073] First, let’s look at the convergence simulation results. Setting the iteration gain ,variance , attenuation factor , the number of iterations is set to 100. At the same time, it is shown that the hierarchical topology proposed in the present invention can improve the convergence speed of the algorithm. According to the same parameter settings, a 100-node non-hierarchical network Run the consistency iteration algorithm and compare the performance without layering the network The in-degree of all nodes in is set to 3.

[0074] The 9th underlying network with 16 nodes , the upper network contains 9 nodes The state iteration curve is shown in Figure 4. As can be seen from Figure 4 (a) and Figure 4 (b), no matter the underlying network Or upper network ,After a certain number of iterations, the noise adding consistency iterative aggregation algorithm proposed in the present invention can converge to the corresponding mean value, proving the expected convergence of the proposed algorithm.

[0075] Next, we will verify the mean square error convergence of the algorithm. and non-hierarchical networks The MSE iteration curve is as follows Figure 5 As shown in the figure, the bottom layer network and the upper layer network under the method proposed in the present invention are iterated only 200 times, while the non-layered network is iterated 6000 times to ensure that the final value of MSE (Mean Squared Error) is similar. Figure 5 It can be seen that the MSE of the three types of networks all converge to 0, indicating that the proposed noise addition consistency algorithm has the mean square error convergence.

[0076] Next, we will conduct a performance comparison experiment. In the simulation, the iteration accuracy threshold is 0.1 and the error is ,when When the iteration accuracy meets the requirement, the iteration is stopped and the number of iterations at this time is recorded. , which is used to represent the convergence speed of the consistent iterative aggregation algorithm, and the iteration state at this time is recorded as , used to represent the final value of the state iteration. Table 1 shows the 9 underlying networks , upper network and non-hierarchical networks The number of nodes and the number of iterations and iterative mean .

[0077] As can be seen from Table 1, 1) for all underlying networks ,Apart from Except for , all networks can converge to the mean with less than 80 convergence times, and The convergence number of 105 is slightly larger than that of other sub-networks. This is because The number of network nodes is the largest, which also verifies the conclusion that the algorithm convergence speed decreases with the increase of network size; 2) Based on the underlying network The final value of the iteration, the upper network After 111 iterations, it converged to a network mean of 5260.22 kWh; 3) Non-layered network To achieve the same convergence accuracy, 3876 iterations are required, and the convergence speed is much lower than the cascade consistency iteration method under the hierarchical topology proposed in the present invention.

[0078] Table 1

[0079] Next, we will conduct privacy analysis. To analyze the privacy of the node initial value, each node Get an initial value from the dataset , and add a variance , random noise with uniform distribution , get the initial value of adding noise , and transmit it to neighboring nodes. Neighbor nodes Received , considering is the private parameter of node j, node i Estimate the variance of node j according to a certain method And generate a random number according to uniform distribution , and follow Estimation Node of .like , then it is called a node j The initial state of i In this way, the nodes can be calculated according to (5) j Privacy leakage In this simulation, The running results give the experimental results of privacy leakage.

[0080] In the simulation, node j is randomly selected from [5,15]. ,node i The variance can be estimated in two different ways , one is to randomly select , recorded as method 1; one is to assign its own variance to , denoted as method 2. For the underlying network 9 with 16 nodes, two different variances The experimental results of privacy leakage under the estimation method are shown in Table 2. The theoretical value in the figure represents the privacy leakage calculated according to (11). It can be seen from Table 2 that 1) when the variance of the node is small, the privacy leakage of the node is large. For example, the privacy leakage of node 14 with a variance of 5 under method 1 is 0.71%, while the privacy leakage of node 7 with a variance of 13 is 0.41%; 2) The privacy leakage under methods 1 and 2 are both smaller than the values ​​under method 3, which means that the node uses the variance as the local value, which can improve the privacy of the node initial value; 3) The privacy leakage of nodes under methods 1-3 is all smaller than the theoretical calculated value, which shows the correctness of (11).

[0081] Table 2

[0082] Compared with the prior art, the beneficial effects of the present invention include at least: Aiming at the acquisition of smart meter data for power load forecasting, the present invention proposes a cascade consistency aggregation method based on hierarchical topology, which can effectively improve the convergence speed of data aggregation and reduce the communication energy consumption of data aggregation; at the same time, different noise variances are designed for smart meters to improve the privacy protection of smart meter data.

[0083] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0084] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0085] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the above. More specific examples (a non-exhaustive list) of computer-readable storage media include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the above. The computer-readable storage medium used herein is not to be interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0086] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0087] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect through the Internet). In some embodiments, by using the state information of the computer-readable program instructions to personalize an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), the electronic circuit may execute the computer-readable program instructions, thereby implementing various aspects of the present disclosure.

[0088] 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 above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A smart meter data security aggregation method for power load forecasting, characterized in that: include: Equip each smart meter with a node to build a hierarchical network topology for the power wireless sensor network; Each node collects smart meter data and obtains the initial value of the iterative state of the consistency aggregation of each node; Each node adds initial noise to the initial value of the iteration state, broadcasts the noisy information to other nodes within a preset communication radius, receives the noisy state information from neighboring nodes, performs consistency iteration on its own node, stops iteration after the iteration accuracy requirement is met, and obtains the average value of the entire hierarchical network topology; Based on the average value of the hierarchical network topology and the number of network nodes, the total power load value of the shared area is calculated to complete the secure aggregation of smart meter data.

2. The method for secure aggregation of smart meter data for power load forecasting according to claim 1 is characterized in that: Construct a hierarchical network topology for the power wireless sensor network, including: The communication relationship of the power wireless sensor network nodes is recorded as an undirected weighted graph ,in, is a collection of nodes, N Represents the number of nodes in the network, and the edge set express N The communication relationship between sub-nodes; If the node j Receive nodes i information, then ,on the contrary , and there is no self-connection in the network, ; is the adjacency weight matrix of network G, and is defined Representation Node i The input neighbor node set, for the adjacency weight matrix W, if , otherwise ; Define the nodes in the undirected weighted graph G i The in-degree is , the in-degree matrix of G is , The Laplacian matrix of G is defined as , and satisfies .

3. The method for secure aggregation of smart meter data for power load forecasting according to claim 2 is characterized in that: The hierarchical network topology includes one or more underlying networks and an upper network, wherein the upper network is composed of relay nodes of the underlying networks, and each underlying network is assigned a relay node.

4. The method for secure aggregation of smart meter data for power load forecasting according to claim 3 is characterized in that: The secure aggregation of smart meter data includes underlying consistency iteration and upper consistency iteration. The underlying consistency iteration process is performed between nodes of the same underlying network, and the relay node calculates the total power load value of the underlying network based on the final iteration value of the underlying consistency iteration and the number of nodes in the underlying network. The upper consistency iteration process is performed between relay nodes of the upper network, and any relay node calculates the total power load value of the shared area based on the final iteration value of the upper consistency iteration and the number of nodes in the upper network.

5. The method for secure aggregation of smart meter data for power load forecasting according to claim 4 is characterized in that: The bottom-level consistency iteration and the upper-level consistency iteration are cascaded. The initial value of the iteration state of each node in the bottom-level consistency iteration is the smart meter data collected by each node; the initial value of the iteration state of each relay node in the upper-level consistency iteration is the total power load value of the bottom-level network corresponding to each relay node.

6. The method for secure aggregation of smart meter data for power load forecasting according to claim 5 is characterized in that: During the node consistency iteration process, the node noise state information and node state information are updated according to the following formula: （1） in, For the The node at the iteration The noise state information, For the The node at the iteration Node status information, For the The node at the iteration The updated value of the node status information, For the The node at the iteration The noise information is a random number that makes the network mean unchanged in each iteration. To update the gain, , For the The node at the iteration Neighbor nodes The noise state information, For Node and nodes The adjacency weight of is the number of network nodes.

7. The method for secure aggregation of smart meter data for power load forecasting according to claim 6 is characterized in that: node and nodes The adjacency weight Calculated as follows: （2） in, Respectively represent nodes and The in-degree, is the edge set of the network, representing the communication relationship between nodes in the network.

8. The method for secure aggregation of smart meter data for power load forecasting according to claim 7 is characterized in that: In the cascade aggregation of bottom-level consistency iteration and top-level consistency iteration, at the node Receive all neighbor nodes Bidirectional transmission Then, the nodes are adjusted according to the following formula: Add noise information: （3） （4） in, For Node In the The noise information added in the iteration is For the The node Designed random numbers, For the The node Additional parameters for design, For Node To Node The parameters of the transmission, For Node To Node The parameters of the transmission, is a constant and satisfies , is the neighbor node of node j.

9. The method for secure aggregation of smart meter data for power load forecasting according to claim 8 is characterized in that: The method further comprises: Calculate the nodes according to the following formula For all neighbor nodes The maximum privacy leakage is analyzed: （5） in, Privacy Disclosure For Node j Initial state Node i To estimate the accuracy The estimated probability, For Node All random noise A collection of For Node i For Node j Deviation The estimated value of , For Node Initial state The estimated value of is the estimated accuracy, for Random noise at any moment The probability density function of For Node j The set of input neighbor nodes.

10. The method for secure aggregation of smart meter data for power load forecasting according to claim 9 is characterized in that: For each node Setting private parameter variance , The mean is 0 and the variance is of uniform distribution.

11. A smart meter data security aggregation system using the smart meter data security aggregation method for power load prediction according to any one of claims 1 to 10, characterized in that: The system comprises: A building module for equipping each smart meter with a node and constructing a hierarchical network topology of a power wireless sensor network; The acquisition module is used to collect smart meter data at each node and obtain the initial value of the iterative state of the consistency aggregation of each node; The consistency iteration module is used to add initial noise to the initial value of each iteration state, broadcast the noisy information to other nodes with a preset communication radius, receive the noisy state information of neighboring nodes, perform consistency iteration on its own node, stop iteration after the iteration accuracy requirement is met, and obtain the average value of the entire hierarchical network topology; The calculation module is used to calculate the total power load value of the shared area based on the average value of the hierarchical network topology and the number of network nodes, and complete the secure aggregation of smart meter data.

12. An electronic device, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the steps of the smart meter data security aggregation method for power load forecasting according to any one of claims 1 to 10.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the smart meter data security aggregation method for power load forecasting as described in any one of claims 1 to 10 are implemented.

Images