5G smart grid user side data secure transmission method

Abstract

The invention relates to a 5G smart power grid user side data secure transmission method, which comprises the following steps: step 1, constructing a smart power grid user side data acquisition and secure transmission system which comprises a 5G smart power grid user side data secure terminal and realizes signal adaptive selection; step 2, establishing an intelligent power grid data transmission credible service management platform: providing a cloud management service for a data acquisition and secure transmission system, and realizing credible service management, remote security upgrade management and key life cycle management of a 5G intelligent power grid data terminal; 3, intelligent power grid user side data intelligent analysis based on machine learning is adopted, and power grid load prediction analysis, daily load feature clustering analysis, demand side flexibility quantitative evaluation and load regression prediction are achieved and improved; the method has the advantages of being high in data collection accuracy, improving the data collection range and improving the power grid data safety.

Description

Technical Field

[0001] The present invention belongs to the technical field of smart grids, and specifically relates to a method for securely transmitting data on the user side of a 5G smart grid. Background Art

[0002] In the digital age, smart grids have become an indispensable part of modern society. Smart grids combine traditional power systems with advanced communication and information technologies to achieve efficient coordination and flexible management of power supply and demand, providing users with more reliable, sustainable and cost-effective power services. As an important part of smart grids, smart grid user data collection systems play a vital role. Through high-precision data collection, real-time monitoring and remote control, power operators can better understand users' electricity consumption, power load demand and the operation status of power networks. In addition, with the development of the scale of power grids, the amount of power grid data has increased massively. However, the current power grid data security transmission system has poor processing effect on massive data due to the backward data encryption technology used, resulting in low anti-attack ability. Therefore, it is very necessary to provide a 5G smart grid user-side data security transmission method with high data collection accuracy, improved data collection breadth and improved power grid data security. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art and to provide a 5G smart grid user-side data security transmission method with high data collection accuracy, improved data collection breadth, and improved grid data security.

[0004] The object of the present invention is achieved by: a method for securely transmitting data on a 5G smart grid user side, the method comprising the following steps:

[0005] Step 1: Build a smart grid user-side data collection and secure transmission system: including a 5G smart grid user-side data security terminal to achieve adaptive signal selection;

[0006] Step 2: Establish a trusted service management platform for smart grid data transmission: Provide cloud management services for data collection and secure transmission systems, and implement trusted service management, remote security upgrade management, and key lifecycle management for 5G smart grid data terminals;

[0007] Step 3: Adopt intelligent analysis of smart grid user-side data based on machine learning: implement and improve grid load forecasting analysis, cluster analysis of daily load characteristics, flexible quantitative evaluation of demand side, and load regression prediction.

[0008] The smart grid user-side data acquisition and secure transmission system in step 1 is a grid data secure transmission system based on chaotic encryption, including a chaotic encryption module and a network attack detection module; wherein the chaotic encryption module specifically uses four-dimensional Chen discrete processing technology and chaotic dynamics theory to encrypt the data transmitted by the grid, and its dynamic system function can be expressed as: In the formula, a represents the encryption coefficient; h represents the power grid data sequence; in order to complete the data encryption process in the digital circuit, it is necessary to set the corresponding derivative in the calculation process to ensure the rationality of the calculation, then: Among them, T represents the calculated derivative, but when the value of this data is low, the above formula can be approximately expressed as:

[0009] The network attack detection module specifically uses weighted residuals to detect each data in the process of grid transmission data intersection, and performs hypothesis tests on it to determine whether there are unsafe factors in the transmitted data. The specific calculation process is set as follows: Among them, W i Represents the weighted residual sensitivity matrix, then e i →N, perform hypothesis test on it: For the detection process using the above algorithm, the operation process can be integrated as follows: Q[e] = W q = 0, Cov[e] = θ = Q[ee T ]=WE, then: e i →N(0,θ), Performing hypothesis test on the above formula, we have: According to this formula, the value range of the threshold β is determined to ensure the detection rate of the power grid transmission data. At the same time, the false detection and missed detection in the calculation process are controlled. Since some abnormal attack data are mostly disguised, in order to ensure the monitoring effect of abnormal data, the above formula is optimized, and then:

[0010] The 5G smart grid user-side data security terminal in step 1 is a smart grid user-side data collection encryption data terminal based on NB-IoT, including an NB-IoT security module, an eUICC multi-operator management module, and a multi-application management module.

[0011] The trusted service management in step 2 adopts data security exchange based on the trust of user behavior, including client driver, unified security exchange protocol of transmission process, protocol analysis, policy filtering, user behavior trust analysis and log audit of isolated exchange center; wherein the security exchange protocol is specifically: the protocol format P is described as: P= <h p ,A,B D ,E(B D)>, where h p is the protocol header; A is the data attribute; B D is the data block; E(B D ) is the random encrypted information of the blocks.

[0012] The power grid load forecasting analysis in step 3 is a power grid load forecasting analysis compatible with demand-side resources, and specifically includes the following steps:

[0013] Step a1: Solving the maximum load of a certain type of power user: The maximum load prediction model can be expressed as: Among them, P max is the forecast result of the maximum load under the action of demand-side resources; Q0 is the initial power consumption; α EE is the consumption reduction rate of this type of user under the action of energy efficiency resources; β LD is the load rate of this type of user under the action of load resources; t is the forecast time period;

[0014] Step a2: Solving the maximum load of multiple types of power users on the same feeder: Assume that the maximum load of the jth type of power user on the ith transmission line under the action of demand-side resources is expressed as P max,ij , then the benefits of all users on the i-th transmission line can be expressed as: Where m is the number of power users on this feeder; P max,ij is the maximum load of the area covered by the i-th transmission line; It is the load simultaneity rate of multiple power users on the same feeder.

[0015] The cluster analysis of daily load characteristics in step 3 is a power load analysis and demand-side response analysis based on a clustering algorithm, and specifically includes the following steps:

[0016] Step b1: data preprocessing;

[0017] Step b2: Entropy analysis: For any user n, the cluster center C of its daily average load curve is i The frequency matching its daily average load curve is P n (C i ), then the entropy value of the user is: If all cluster centers have equal matching probabilities, the entropy value is the highest; if all load curves are classified into one cluster, the entropy value is the lowest;

[0018] Step b3: Data analysis: Starting from a set of initial clusters generated by the standard K-means algorithm, the number of initial clusters K = k0, when the variance of the load curve S(t) in the data to its typical load curve exceeds the critical value, the adaptive K-means algorithm will add new cluster centers: Here, θ is the choice of critical value.

[0019] The quantitative evaluation of demand-side flexibility in step 3 is the evaluation of demand-side response potential based on the prediction of massive load data, specifically: the demand response potential value P drp It is defined as the product of the average peak power consumption that is higher than 80% of the daily maximum load and the average load reduction rate of the industry demand response: in, is the load value corresponding to user i at peak time t1, t2, ..., tM; λ i is the average load reduction rate under the demand response measures of user i; and the actual demand response peak reduction potential should be less than the system peak time. The value of the actual load of demand response minus the minimum load is the maximum value of the demand side response potential. The difference between the user's peak time and the daily minimum load is obtained from the observation of the i-type load curve data: Where h represents the peak time of the system; p imin is the minimum daily load in the typical load curve of user i; therefore, the final demand response peak load reduction potential of type i is:

[0020]

[0021] The load regression prediction in step 3 adopts a "clustering-regression" large power user short-term load prediction model based on cluster load characteristic curves. The model adopts a distributed TensorFlow framework and includes the following steps:

[0022] Step c1: Data preprocessing: The attribute power and peak power L considered when clustering the power consumption characteristics of large users peak , level charge L flat , valley charge L valley And the maximum power P max The original data is normalized and processed to obtain four attributes that reflect the characteristics of electricity consumption: peak-to-total ratio r1, flat-to-valley ratio r3, valley-to-total ratio r2, and load factor η; Where, L overall is the total power;

[0023] Step c2: Clustering of large power users: clustering users according to four attributes;

[0024] Step c3: Cluster load characteristic curve: Users with similar power consumption characteristics are grouped into one cluster. i n i Users, C cluster load curves are obtained according to the following formula: Where Load(i) is the cluster C i Cluster load curve vector; l j Cluster C i The average load of the included users at the jth moment;

[0025] Step c4: regression prediction.

[0026] Beneficial effects of the present invention: The present invention is a method for secure transmission of user-side data of a 5G smart grid. During use, the present invention adopts a power grid data security transmission system based on chaotic encryption, constructs a power grid data encryption module with the help of four-dimensional Chen discrete processing and chaotic dynamics theory, and adopts weighted residual detection data intersection process to construct a power grid data network attack detection module; the present invention provides cloud management services for data acquisition and secure transmission systems through a smart grid data transmission trusted service management platform, and realizes trusted service management, remote security upgrade management and key life cycle management of 5G smart grid data terminals; the present invention provides data support for area planning, energy storage configuration, demand response, orderly electricity consumption, and new energy consumption through intelligent analysis of smart grid user-side data based on machine learning, thereby realizing low-carbon and efficient operation of the distribution network; the present invention has the advantages of high data acquisition accuracy, improved data acquisition breadth, and improved power grid data security. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a structural diagram of the smart grid user data collection system of the present invention.

[0028] Figure 2 It is a schematic diagram of the remote communication architecture of the present invention.

[0029] Figure 3 This is a framework diagram of the power grid data security transmission system based on chaotic encryption of the present invention.

[0030] Figure 4 FIG. 4 is a schematic diagram of an embedded controller of the present invention.

[0031] Figure 5 This is a schematic diagram of a power grid data information storage circuit of the present invention.

[0032] Figure 6 This is a diagram of the user behavior trusted data security exchange framework of the present invention.

[0033] Figure 7 It is a driving process flow chart of the present invention.

[0034] Figure 8 This is a flowchart of the isolation switching center processing of the present invention.

[0035] Fig. 9 It is a schematic diagram of the strategy filtering of the present invention.

[0036] Fig.10 This is a schematic diagram of the user behavior trustworthy detection of the present invention.

[0037] Fig.11It is a data cleaning flow chart of the present invention.

[0038] Fig.12 This is a flow chart of the “clustering-regression” load forecasting model of the present invention.

[0039] Fig.13 This is a block diagram of the RNN network structure of the present invention.

[0040] Fig.14 This is a block diagram of the LSTM network structure of the present invention. DETAILED DESCRIPTION

[0041] The present invention will be further described below in conjunction with the accompanying drawings.

[0042] Example 1

[0043] like Figure 1-14 As shown, a 5G smart grid user-side data security transmission method comprises the following steps:

[0044] Step 1: Build a smart grid user-side data collection and secure transmission system: including a 5G smart grid user-side data security terminal to achieve adaptive signal selection;

[0045] In this embodiment, the framework of the power grid data security transmission system based on chaos encryption is as follows: Figure 3 As shown; the embedded controller uses ARM32-bit microprocessor as the core chip, such as Figure 4 As shown; the power grid data information storage circuit uses AT24C08 chip as the data storage chip, such as Figure 5 As shown; the power grid data security transmission system based on chaos encryption includes a chaos encryption module and a network attack detection module; wherein ① the chaos encryption module is specifically: since the current power grid data contains a large amount of image information, in order to realize the transmission of this part of information and ensure the security of the transmitted information, the present invention uses four-dimensional Chen discrete processing technology and chaos dynamics theory to encrypt the data transmitted by the power grid, and its dynamic system function can be expressed as: In the formula, a represents the encryption coefficient; h represents the power grid data sequence; in order to complete the data encryption process in the digital circuit, it is necessary to set the corresponding derivative in the calculation process to ensure the rationality of the calculation, then: Among them, T represents the calculated derivative, but when the value of this data is low, the above formula can be approximately expressed as: According to the above formula, the encryption function of the power grid transmission data is set: According to historical data, if the T value is set to 0.05, we have: The above formula is used as the chaotic encryption function of power grid data, and this function is used to complete the encryption of power grid data and serve as the basis for data transmission;

[0046] ② Network attack detection module: In the process of power grid data transmission, it may be attacked by network information. It is far from enough to rely solely on encryption algorithms to encrypt power grid data. Therefore, a network attack detection module will be added to improve data security. Weighted residuals are used to detect each data in the process of grid transmission data intersection, and hypothesis tests are performed to determine whether there are unsafe factors in the transmitted data. The specific calculation process is set as follows: Among them, W i Represents the weighted residual sensitivity matrix, then e i →N, perform hypothesis test on it: For the detection process using the above algorithm, the operation process can be integrated as follows: Q[e] = W q = 0, Cov[e] = θ = Q[ee T ]=WE, then: e i →N(0,θ), Performing hypothesis test on the above formula, we have: According to this formula, the value range of the threshold β is determined to ensure the detection rate of the power grid transmission data. At the same time, the false detection and missed detection in the calculation process are controlled. Since some abnormal attack data are mostly disguised, in order to ensure the monitoring effect of abnormal data, the above formula is optimized, and then: Assuming that there are abnormal data in G0 and G1, it is impossible to process them with high precision using formula (1). Formula (2) can be used to analyze their data status and determine whether the data is aggressive, thereby ensuring the security of data transmission.

[0047] The 5G smart grid user-side data security terminal is an encrypted data terminal for data collection on the user side of the smart grid based on NB-IoT, including an NB-IoT security module, an eUICC multi-operator management module (by storing different operator Profile files in different ISD-P security domains inside the eUICC, and ensuring mutual isolation between Profiles in accordance with the GlobalPlatform specification), and a multi-application management module (using multi-application management technology, developing a series of components to achieve hardware / manufacturer-independent application and module interfaces, and defining application security domains containing specific keys and security management in accordance with the GlobalPlatform specification, while ensuring the portability of the application, achieving complete isolation of keys between module manufacturers and other security domain providers).

[0048] The functions of the 5G smart grid user-side data security terminal are as follows: ① Data collection function: Through sensors and smart meters and other equipment, the collection terminal obtains the power consumption data of power users in real time, including data information such as power consumption, power, voltage and current. The collection terminal needs to ensure the high precision and real-time performance of the data to ensure the accuracy and reliability of the data; ② Data transmission function: The collected power data needs to be uploaded to the system's central processing unit or data management platform for further processing and analysis. The collection terminal transmits the data to the designated target node through the communication network (Ethernet and wireless network, etc.) to ensure the security and reliable transmission of the data; The data collection system structure diagram is shown in the figure below Figure 1 As shown in the figure, (1) the master station layer is responsible for managing and controlling the entire data acquisition system. It collects the user's electricity consumption data in real time by communicating with the monitoring equipment at the collection point, and ensures the accuracy and timeliness of the data. The master station layer is also responsible for data transmission and management, and transmits the collected data to the central processing unit or data management platform for further processing and analysis. (2) The data acquisition layer collects the user's power data in real time through different collection point monitoring equipment. The data acquisition layer uploads the collected data to the master station layer for processing and management through remote communication channels and local communication channels. (3) The monitoring equipment at the collection point includes reactive power compensation equipment, smart meters and user distribution switches, etc., which monitor and control the user's power equipment, and can communicate with the master station layer and the data acquisition layer to provide real-time electricity consumption data and receive instructions from the master station layer to perform corresponding control operations.

[0049] Remote communication architecture Figure 2 As shown, user data collection (including the collection mode of large dedicated transformer users, the collection mode of small and medium-sized dedicated transformer users, and the single-phase or three-phase low-voltage industrial and commercial user collection mode): the data collection equipment in the collection mode includes dedicated transformer concentrators and dedicated transformer data terminals, etc., which are connected to the power equipment of dedicated transformer users to collect users' electricity consumption data in real time, including important information such as electricity consumption, power, voltage and current; the dedicated communication solution for dedicated transformer users is based on a combination of wired communication networks (Ethernet and power line carrier communication, etc.) or wireless communication networks (general packet radio service, 4G, 5G, etc.), and the collected data is transmitted to the system's master station layer or data management platform to realize remote transmission and centralized management of data.

[0050] Step 2: Establish a trusted service management platform for smart grid data transmission: Provide cloud management services for data collection and secure transmission systems, and implement trusted service management, remote security upgrade management, and key lifecycle management for 5G smart grid data terminals;

[0051] Step 3: Adopt intelligent analysis of smart grid user-side data based on machine learning: implement and improve grid load forecasting analysis, cluster analysis of daily load characteristics, flexible quantitative evaluation of demand side, and load regression prediction.

[0052] The present invention is a 5G smart grid user-side data security transmission method. During use, the present invention adopts a power grid data security transmission system based on chaotic encryption, optimizes the embedded controller and the information storage circuit, constructs a power grid data encryption module with the help of four-dimensional Chen discrete processing and chaotic dynamics theory, and uses weighted residual detection data intersection process to construct a power grid data network attack detection module; the present invention has the advantages of high data acquisition accuracy, improved data acquisition breadth, and improved power grid data security.

[0053] Example 2

[0054] like Figure 1-14 As shown, a 5G smart grid user-side data security transmission method comprises the following steps:

[0055] Step 1: Build a smart grid user-side data collection and secure transmission system: including a 5G smart grid user-side data security terminal to achieve adaptive signal selection;

[0056] Step 2: Establish a trusted service management platform for smart grid data transmission: Provide cloud management services for data collection and secure transmission systems, and implement trusted service management, remote security upgrade management, and key lifecycle management for 5G smart grid data terminals;

[0057] In this embodiment, the trusted service management adopts the secure exchange of data based on the trustworthiness of user behavior, such as Figure 6 As shown, it includes the client driver, the unified security exchange protocol of the transmission process, the protocol analysis of the isolated exchange center, policy filtering, user behavior trust analysis and log auditing;

[0058] The client driver includes a data attribute driver and an exchange protocol driver, so as to filter data and check the credibility of user behavior through attributes in the isolated exchange center; the exchange protocol driver encapsulates the data and its attributes together as a secure exchange protocol, and the secure exchange protocol recombines the data and data attributes in a block-by-block random encryption manner to improve the security of data transmission and exchange and the unified processing of data.

[0059] The isolation exchange center first performs protocol analysis and attribute analysis to recover the data and its attributes, then verifies the user's identity through identity authentication to ensure the user is legitimate, and then filters the data based on data attributes and policy rules to filter out data that does not meet the policy rules. Finally, the user's trust value and user behavior are used to perform a trustworthy check and abnormal behavior check on the user's behavior.

[0060] Policy filtering records data attribute information, data user information, and information generated by data during transmission. Under the policy rules of the isolation exchange center, data that does not meet the rules cannot pass, thereby filtering the data.

[0061] ① Data attributes: Data attributes are used for policy-based filtering, user behavior trust check and user abnormal behavior analysis in the isolation exchange center. Policy-based filtering filters data by extracting basic data attribute values ​​and user-added information. User behavior trust check and abnormal behavior analysis make trustworthy judgments on user history and current behavior through user information in data attributes and user operations on data. Basic data attributes are direct information related to data, including data type, data size, data creator, data modification time, etc. User-added attributes are user descriptions of data, including data security level, whether data needs to be encrypted, data keywords, user identity of operating data, etc. User operations are user operations on data, including access destination, access protocol, access action, access time, access source, etc. The data attribute A is described in the form of: A = <u,A d (k,v),A o (k,v),e f ,E,t a >, where u is user information; A d is the basic attribute of the data; A o For user operations on data; f is the attribute encryption identifier; E is the attribute encryption key information; t a The attribute generation time; in order to reduce the introduction of symbols in subsequent descriptions, the description of symbols will not distinguish whether it is in an encrypted state; under the physical isolation conditions of the isolated exchange center, data attributes can be carried from one secure network to another as raw data along with the data without discarding the user's behavior in the previous network.

[0062] ②Secure Exchange Protocol: The secure exchange protocol divides the data into blocks and adopts random encryption for the blocks. Block encryption encrypts the header and tail of formatted documents such as PDF and Word, and randomly selects and encrypts the blocks in the middle of the document to improve the encryption and decryption rate without losing security. The protocol format P is described as: P = <h p ,A,B D ,E(B D )>, where h p The protocol header contains information such as protocol generation time, protocol content length, data information encryption bit, protocol checksum, etc.; A is the data attribute; B D is the data block; E(B D ) is the block-based random encryption information, including randomly selected encryption blocks and encryption keys.

[0063] ③Driver: The data with attributes is encapsulated into unified secure exchange data by the attribute driver and protocol driver on the client side, and then transmitted to the isolated exchange center through the SSL tunnel. The operation is as follows: Figure 7 As shown, the following steps are included:

[0064] Step 3.1: Perform key-value pair processing on the basic attributes of the data, the attributes added by the user, and the attributes operated by the user to obtain A d (k,v) and A o (k,v);

[0065] Step 3.2: Encrypt the data and encapsulate it to obtain the encrypted data attribute A;

[0066] Step 3.3: Randomly divide the data into blocks and obtain B D ;

[0067] Step 3.4: Randomly select data blocks and then encrypt them to obtain the encrypted B D and E(B D );

[0068] Step 3.5: Encapsulate the data attributes and data to form a unified secure exchange format data P.

[0069] ④ Isolation exchange center: Protocol analysis and attribute analysis respectively parse the data encapsulated by the client's protocol driver and attribute driver to restore the data and its attributes; identity authentication is used to authenticate the user's identity; policy filtering uses policy rules to filter data based on data attributes to filter data that does not comply with policy rules; user behavior trustworthiness check and abnormal behavior analysis use trusted operation set matching to check whether user behavior is trustworthy based on data attributes, and analyze whether user behavior is abnormal through association analysis in combination with abnormal behavior sets under the condition that user behavior is untrustworthy;

[0070] When the client data arrives at the isolation exchange center, the data will be processed as follows: Figure 8 As shown, the following steps are included:

[0071] Step 4.1: Security exchange protocol analysis, recovery of data and its attributes B D and A;

[0072] Step 4.2: In the user authentication phase, user u is extracted from A for identity authentication. In the non-authentication phase, the user directly proceeds to the next step;

[0073] Step 4.3: Data attribute analysis extracts data attribute information A;

[0074] Step 4.4: Extract basic data attributes and user-added attributes A d(k,v), filter the data by policy rules;

[0075] Step 4.5: Extract user u and user's operation behavior A on the data o (k,v), determine the credibility of the user's behavior. If the user's behavior is credible, the process ends and returns the result. If the user's behavior is credible, the process goes to the next step.

[0076] Step 4.6: Based on the user's operation behavior on the data in the data attributes, perform correlation analysis on the user behavior to determine whether the user behavior is abnormal; if the behavior is abnormal, the user's trust value will be reduced; end (the entire operation process includes a log recording function).

[0077] ⑤ Policy-based filtering: Policy-based filtering uses policy rules to filter data based on the data attributes provided by the client; the policy filtering process is as follows: Fig. 9 As shown, the following steps are included:

[0078] Step 5.1: Extract data attribute information A d (k,v);

[0079] Step 5.2: traverse the policy rule expression exp and extract the variable var in the expression;

[0080] Step 5.3: Set var as key from A d Extract the value v (integer, floating point number and string) corresponding to var from (k,v);

[0081] Step 5.4: Replace var in the policy rule with v and calculate the expression;

[0082] Step 5.5: Determine whether the data is filtered based on the calculation results and record the log.

[0083] ⑥ User behavior trust check: User behavior trust check determines whether the user is allowed to access resources by checking whether the user's behavior is trustworthy; User behavior trust check consists of two parts: user trusted behavior set and user behavior trust check rules. The user trusted behavior set is pre-set, and the behavior trust check rules are checked based on the trusted set; Before defining the trusted behavior set, it is necessary to first define the subject set, object set and behavior set: Subject set: S = {s1, s2, ..., s m}, s in S i (1≤i≤m) is a subject, also represented as a user, and the trust value tv is set on the subject i , the trust value is greater than 0, indicating that the user is allowed to perform abnormal operation steps. When the trust value is lower than 0, the user operation will be prohibited. After the problem is solved, the trust value will be reset; Object set: O = {o1, o2, ..., o n}, o in O i (1≤i≤n) is an object, which also represents the object of operation; behavior set: B={b1,b2,...,b l}, b in B i (1≤i≤l) represents an operation of a subject on a client (operation of a user on an object); Trusted behavior set: T B ={t1,t2,...,t k}, where the trusted behavior t i Described as: t i is the sequence of operations performed by the subject on the object; user behavior trustworthiness check such as Fig.10 As shown,

[0084] Step 6.1: Extract user u and user operation A from data attribute A o (k,v), and converted to access subject s i , access object o i , access operation b i (User behavior), etc.

[0085] Step 6.2: Check s i Trust value tv i , and compare it with 0 (threshold). If it is not less than 0, go to the next step. If it is less than 0, log it and issue a warning.

[0086] Step 6.3: Load the user behavior trusted set T B ;

[0087] Step 6.4: Traverse T B ;

[0088] Step 6.5: Extract t i , t i The s, o, and b in the i , o i , b i Perform a match. If a match is found, the information is credible. Otherwise, a log is recorded and the abnormal behavior analysis is performed.

[0089] ⑦ User abnormal behavior association analysis: The association based on user behavior is based on the user abnormal behavior pattern as the prototype to monitor the operator's behavior pattern; the user's abnormal behavior pattern is stored in the abnormal behavior rule library. If the user behavior conforms to the behavior pattern defined in the abnormal behavior rule library, it is checked as abnormal behavior; the abnormal behavior rule library is composed of multiple tree-structured rules, and the rule tree is composed of multiple rules; the rules record the user's abnormal operation behavior, and the association analysis extracts the behavior of the visiting users through the preset rule library, and performs association analysis based on the rule library to determine whether the user behavior is abnormal; when the user behavior is abnormal, the user's trust value will be affected, and the log will be recorded for future analysis; the steps of user behavior association analysis are as follows:

[0090] Step 7.1: Extract user access action information from data attributes;

[0091] Step 7.2: Traverse the abnormal behavior rule library, extract abnormal operation rules, and compare them with the user operation behavior. If they match, obtain the corresponding subject from the subject set and reduce the trust value of the subject by 1, issue a warning and record the log.

[0092] Step 3: Adopt intelligent analysis of smart grid user-side data based on machine learning: implement and improve grid load forecasting analysis, cluster analysis of daily load characteristics, flexible quantitative evaluation of demand side, and load regression prediction.

[0093] The present invention is a method for secure data transmission on the user side of a 5G smart grid. During use, the present invention provides cloud management services for data acquisition and secure transmission systems through a smart grid data transmission trusted service management platform, thereby realizing trusted service management, remote security upgrade management, and key lifecycle management of 5G smart grid data terminals. The present invention has the advantages of high data acquisition accuracy, improved data acquisition breadth, and improved grid data security.

[0094] Example 3

[0095] like Figure 1-14 As shown, a 5G smart grid user-side data security transmission method comprises the following steps:

[0096] Step 1: Build a smart grid user-side data collection and secure transmission system: including a 5G smart grid user-side data security terminal to achieve adaptive signal selection;

[0097] Step 2: Establish a trusted service management platform for smart grid data transmission: Provide cloud management services for data collection and secure transmission systems, and implement trusted service management, remote security upgrade management, and key lifecycle management for 5G smart grid data terminals;

[0098] Step 3: Adopt intelligent analysis of smart grid user-side data based on machine learning: implement and improve grid load forecasting analysis, cluster analysis of daily load characteristics, flexible quantitative evaluation of demand side, and load regression prediction.

[0099] In this embodiment, ① the power grid load forecasting analysis is a power grid load forecasting analysis compatible with demand-side resources, which specifically includes the following steps:

[0100] Step a1: Solving the maximum load of a certain type of power user: For a certain type of power user, the load forecast considering various types of demand-side resources is calculated in two steps: the first step is to consider the impact of energy efficiency resources on power consumption and thus the average power load; the second step is to consider the impact of different load resources on the load rate and thus the maximum load of the user on the basis of the first step;

[0101] (1) Maximum load prediction considering a single demand-side resource: For a certain type of user, only a single demand-side resource is considered, that is, the user has only one electricity consumption link, and there is an energy efficiency resource and a load resource at the same time. The maximum load prediction model can be expressed as: Among them, P max is the forecast result of the maximum load under the action of demand-side resources; Q0 is the initial power consumption, that is, the power consumption forecast value obtained by conventional forecasting methods without considering demand-side management resources; α EE is the energy saving rate of this type of user under the action of energy efficiency resources, which is equal to the product of natural power saving rate and user penetration rate; β LD is the load rate of this type of user under the action of load resources; t is the forecast time period;

[0102] (2) Maximum load prediction considering multiple demand-side resources: a. Analysis of multiple energy efficiency resources: First, consider the change in total power consumption under the action of multiple energy efficiency resources. Assume that ΔQ is the power saving, ΔQ i is the amount of electricity saved in the i-th electricity consumption link; ΔQ i,0 is the initial power consumption of the ith power consumption link; α EE,i is the consumption reduction rate of the i-th electricity consumption link of this type of user under the action of energy efficiency resources; i is the state coefficient indicating whether energy efficiency resources exist, λ i When the value is 0, it means that the user does not have such resources. i When it is 1, it means that this resource is owned;

[0103] The amount of electricity saved by the i-th electricity consumption link of this type of user under the action of energy efficiency resources is: ΔQ i =Q i,0 λ i α EE,i, under the action of various energy efficiency resources, the total power saving is the sum of the power saving of each power consumption link: ΔQ=ΔQ1+ΔQ2+...+ΔQ i +...+ΔQ k , where k is the number of power consumption links of the power user; therefore, the power consumption under the action of energy efficiency resources is Q EE For: Q EE =Q0-ΔQ;

[0104] b. Analysis of multiple load resources: After considering multiple energy efficiency resources, the effects of multiple load resources are further considered. The calculation formula for the maximum load reduction under the effect of a single load resource is: Among them, ΔP max,i is the maximum load reduced under the action of the i-th load type resource; P max,0 is the maximum load under the action of energy efficiency resources; β LD,i is the load rate of the power user under the action of the i-th load type resource; the maximum load reduction under the action of multiple load type resources is: Where w is the number of load resources owned by the user; is the load simultaneous rate of various load resources; therefore, the prediction result of the maximum load under the action of demand-side resources is: P max =P max,0 -ΔP max ;

[0105] Step a2: Solving the maximum load of multiple types of power users on the same feeder: Assume that the maximum load of the jth type of power user on the ith transmission line under the action of demand-side resources is expressed as P max,ij , then the benefits of all users on the i-th transmission line can be expressed as: Where m is the number of power users on this feeder; P max,ij is the maximum load of the area covered by the i-th transmission line; It is the load concurrency rate of various power users on the same feeder;

[0106] Step a3: Solving the comprehensive maximum load: Since the entire substation covers multiple power lines, when solving the comprehensive maximum load, it is necessary to consider the power consumption of each line and then perform power superposition. The superposition effect is: Among them, P Umax is the comprehensive maximum load; is the load simultaneity rate of different feeders; n is the number of lines in the substation.

[0107] ② Cluster analysis of daily load characteristics is a power load analysis and demand-side response analysis based on clustering algorithms, which specifically includes the following steps:

[0108] Step b1: data preprocessing;

[0109] Step b2: Entropy analysis: For any user n, the cluster center C of its daily average load curve is i The frequency matching its daily average load curve is P n (C i ), then the entropy value of the user is: If all cluster centers have equal matching probabilities, then their entropy value is the highest (p n (C i )=1K); if all load curves are classified into one cluster, the entropy value is the lowest; the cluster center C of the daily average load curve can also be calculated separately i The entropy value is calculated by matching the frequency of the daily average load curve for each day of the week;

[0110] Step b3: Data analysis: Starting from a set of initial clusters generated by the standard K-means algorithm, the number of initial clusters K = k0, when the variance of the load curve S(t) in the data to its typical load curve exceeds the critical value, the adaptive K-means algorithm will add new cluster centers: Among them, θ is the choice of critical value. Usually, θ takes the value in [0, 2]. Dynamically checking the critical value can separate the clusters that do not meet the conditions. In addition, the load curve is normalized, which makes this algorithm more robust. The specific algorithm is described as follows:

[0111] Known: Daily load curves of all users {s n (t)}, where n represents the date, and the minimum and maximum number of clusters are min.k and max.k;

[0112] Set K = min.k

[0113] while1do

[0114] Execute K-means (based on initial cluster centers)

[0115] forall cluster centers do

[0116] For n All load curves in (t)} use the formula to check the critical value, and record the number of cluster centers that exceed the critical value as N v ; A hierarchical clustering algorithm is used to merge cluster centers that are too close to each other, reduce the size of the load curve dictionary to the target size T, and generate a new cluster center after weighted averaging the two closest cluster centers. Hierarchical clustering is used to filter and unify these small clusters, so that a more effective, smaller and more stable load curve dictionary can be generated. The specific algorithm of hierarchical clustering is described as follows:

[0117] Known: Adaptive K-means calculation results (C i is the cluster center (i=1,2,...,K), n i is the size of the ith cluster (the number of points assigned to each cluster center)

[0118] Set the target dictionary size, T(<K)

[0119] whileK>Tdo

[0120] Find the two closest cluster centers C i and C j

[0121] Setting C i =(n i C i +n j C j )(n i +n j ) and then delete C j

[0122] K=K-1.

[0123] ③ The quantitative evaluation of demand-side flexibility is the evaluation of demand-side response potential based on the prediction of massive load data. Specifically, it is to build a high-precision convolution-long short-term memory hybrid neural network CNN-LSTM model to predict the load change of the next day, and build demand response evaluation indicators based on the predicted load data and load characteristics; using the predicted daily load data, according to the characteristics of the load curve, define the daily maximum load P max (for the predicted maximum daily power consumption), maximum response load P maxr (the difference between the predicted daily maximum load and the daily minimum load), minimum response load P minr (the difference between the minimum peak load and the daily minimum load), the median load of the confidence interval P mod (the average value of the maximum response load and the minimum response load), the demand response potential value ratio f i (ratio of demand response potential value to daily maximum load), demand response potential value P drp And the demand response period T (a time period with demand potential); the demand response potential value P drp It is defined as the product of the average peak power consumption that is higher than 80% of the daily maximum load and the average load reduction rate of the industry demand response: in, is the load value corresponding to user i at peak time (the time when the power consumption is higher than 80% of the daily maximum load) t1, t2, ..., tM; λ iis the average load reduction rate under the demand response measures of user i; and the actual demand response peak reduction potential should be less than the system peak time. The value of the actual load of demand response minus the minimum load is the maximum value of the demand side response potential. The difference between the user's peak time and the daily minimum load is obtained from the observation of the i-type load curve data: In the formula, h represents the peak time of the system, which is 1-M; p imin is the daily minimum load in the typical load curve of user i; therefore, the final demand response peak load reduction potential of type i is:

[0124] ④ Load regression prediction adopts the "clustering-regression" short-term load forecasting model for large power users based on cluster load characteristic curves. The model uses a distributed TensorFlow framework and includes the following steps:

[0125] Step c1: Data preprocessing: The original data of the attributes of power, peak power, average power, valley power and maximum power, which are considered when clustering the power consumption characteristics of large users, are normalized to obtain four attributes that reflect the power consumption characteristics: peak-to-total ratio r1, average-to-valley ratio r3, valley-to-total ratio r2, and load factor η; Where, L overall is the total power; L peak is the peak charge; L valley is the valley charge; L flat P is the average charge; max is the maximum power;

[0126] Step c2: Clustering of large power users: Cluster users according to four attributes. The initial cluster center can be achieved by random selection or manual assignment. The data is converted into a constant tensor through the tf.constant() function. The tf.random_shuffle() function randomizes the data. The tf.slice() function can extract some dimensions from the original data. The tf.Variable() function converts the initial cluster center from a constant to a variable. Next, the distance from the sample to the cluster center needs to be calculated. The distance measurement is generally:

[0127] (1) The Euclidean distance between two points x and y in n-dimensional space is expressed as follows: In the formula, x k ,y k are the k-th attribute (k-th component) of x and y respectively;

[0128] (2) By generalizing the Euclidean distance, we can obtain the more general Minkowski distance: In the formula, r is a parameter. By taking different values ​​of r, we can get distances in different norms: the distance when r = 1 is called Manhattan distance or Hamming distance, and the distance at this time is expressed as the L1 norm between x and y: When r = 2, it is the mathematical L2 norm, which is the same as the Euclidean distance; when r = ∞, it is the infinite norm distance (L ∞ -distance), which is the maximum distance between sample attributes: Taking Euclidean distance as an example, in TensorFlow, the tf.expand_dims() function is used to expand the variable dimension, the tf.sub() function is used to implement variable subtraction, and the tf.square() function is used to implement the square of the component; the tf.reduce_sum() function calculates the sum of the distances from the sample point to the cluster center, and the t.argmin() function returns the number of the cluster center with the smallest distance value;

[0129] The goal of clustering is that the cluster center does not change, or changes very little, which can be measured by SSE. By calculating the Euclidean distance from each data point to the cluster center and iteratively calculating the sum of squares of the errors, the smaller the SSE, the better the clustering effect. The SSE calculation formula is as follows: In the formula, x is the sample; C i is the i-th cluster; c i Cluster C i The center of mass; m i is the number of samples in the ith cluster; K is the number of clusters; dist(c i ,x) is the corresponding cluster center c i The distance from the sample x in the corresponding cluster; for cluster C i , use the t.equal() function to mark whether the sample point belongs to the cluster, the tf.gather() function extracts the samples belonging to the cluster according to the mark, and the tf.reduce_mean() function calculates the cluster C i The new cluster center c i , keep iterating until the convergence condition is met;

[0130] Step c3: Cluster load characteristic curve: By clustering large power users, a cluster load curve is obtained to characterize the power consumption level of users in the cluster, in preparation for the next step of building a load forecasting model. Users with similar power consumption characteristics are clustered into one cluster. For cluster C i n i Users, C cluster load curves are obtained according to the following formula: Where Load(i) is the cluster C i The cluster load curve vector is 96-dimensional; the cluster load curve vector element is l j , l j Cluster C iThe average load of the included users at the jth moment; the cluster load curve represents the load level of a class of users with similar electricity consumption characteristics;

[0131] Step c4: Regression prediction: Use the long short-term memory network LSTM under the TensorFlow framework for regression prediction. Specifically, the long short-term memory network is a type of recurrent neural network RNN, such as Fig.13 As shown, the recurrent neural network has an input x at every moment t , combined with the current state A of the recurrent neural network t , and get the output h t , and the current state A t is the state A at the previous moment t-1 and the current input x t The LSTM improves the classic recurrent neural network by introducing a "forget gate" to allow information to selectively affect the state of the RNN at each moment. Fig.14 As shown, the forget gate passes the current input x t and the output h at the previous moment t-1 , for the state c at the previous moment t-1 The elements in the set weight, the weight value range is from 0 to 1: f1 = σ (W f ·[h t-1 ,x t ]+b f ), where W f is the weight matrix of the forget gate; b f is the bias term; [h t-1 ,x t ] is the input vector of the forget gate; σ is the sigmoid function; the new state is supplemented by the "input gate", which is based on h t-1 Decision t-1 Which parts of the state are written into the current state c t Go to: i t =σ(W i ·[h t-1 ,x t ]+b i ), similarly, W i is the weight matrix of the input gate; b i is the bias term of the input gate; at the same time, the candidate value of the new state Produced by the tanh function, Possibly added to the new state: Then the current state c t It is the state c at the previous moment. t-1 Multiply by the forget gate f t , and then use the new state candidate value Multiply by input gate i t , and then sum the two products: Similarly, the output gate is represented as: o t =σ(W o ·[h t-1 ,x t ]+b o ), and the final output of LSTM is h t It is determined by the output gate and the unit state: h t =o t ·tanh(c t ), the function tf.nn.rn() can be directly called in TensorFlow to implement LSTM; the cluster load characteristic curve obtained in the previous step is used as the attribute factor of the actual load to train the model and then perform load prediction.

[0132] The present invention is a method for securely transmitting data on the user side of a 5G smart grid. During use, the present invention adopts machine learning theory, performs pre-processing operations such as PQ data separation, data sampling, data standardization, and data cleaning on massive data on the user side of the smart grid, and determines the user load pattern through cluster analysis of the user side data; on this basis, a flexible quantitative evaluation of the demand side is performed, and load regression prediction is performed based on the evaluation results; ultimately, data support is provided for substation planning, energy storage configuration, demand response, orderly electricity consumption, and new energy consumption, so as to realize low-carbon and efficient operation of the distribution network; the present invention has the advantages of high data collection accuracy, improved data collection breadth, and improved grid data security.

Claims

1. 5G smart grid user-side data security transmission method, characterized by: The method comprises the following steps: Step 1: Build a smart grid user-side data collection and secure transmission system: including a 5G smart grid user-side data security terminal to achieve adaptive signal selection; Step 2: Establish a trusted service management platform for smart grid data transmission: Provide cloud management services for data collection and secure transmission systems, and implement trusted service management, remote security upgrade management, and key lifecycle management for 5G smart grid data terminals; Step 3: Adopt intelligent analysis of smart grid user-side data based on machine learning: implement and improve grid load forecasting analysis, cluster analysis of daily load characteristics, flexible quantitative evaluation of demand side, and load regression prediction.

2. The 5G smart grid user-side data security transmission method according to claim 1, characterized in that: The smart grid user-side data acquisition and secure transmission system in step 1 is a grid data secure transmission system based on chaotic encryption, including a chaotic encryption module and a network attack detection module; wherein the chaotic encryption module specifically uses four-dimensional Chen discrete processing technology and chaotic dynamics theory to encrypt the data transmitted by the grid, and its dynamic system function can be expressed as: In the formula, a represents the encryption coefficient; h represents the power grid data sequence; in order to complete the data encryption process in the digital circuit, it is necessary to set the corresponding derivative in the calculation process to ensure the rationality of the calculation, then: Among them, T represents the calculated derivative, but when the value of this data is low, the above formula can be approximately expressed as:

3. The 5G smart grid user-side data security transmission method according to claim 2, characterized in that: The network attack detection module specifically uses weighted residuals to detect each data in the process of grid transmission data intersection, and performs hypothesis tests on it to determine whether there are unsafe factors in the transmitted data. The specific calculation process is set as follows: Among them, W i Represents the weighted residual sensitivity matrix, then e i →N, perform hypothesis test on it: For the detection process using the above algorithm, the operation process can be integrated as follows: Q[e] = W q = 0, Cov[e] = θ = Q[ee T ]=WE, then: e i →N(0,θ), Performing hypothesis test on the above formula, we have: According to this formula, the value range of the threshold β is determined to ensure the detection rate of the power grid transmission data. At the same time, the false detection and missed detection in the calculation process are controlled. Since some abnormal attack data are mostly disguised, in order to ensure the monitoring effect of abnormal data, the above formula is optimized, and then:

4. The 5G smart grid user-side data security transmission method according to claim 2, characterized in that: The 5G smart grid user-side data security terminal in step 1 is a smart grid user-side data collection encryption data terminal based on NB-IoT, including an NB-IoT security module, an eUICC multi-operator management module, and a multi-application management module.

5. The 5G smart grid user-side data security transmission method according to claim 1, characterized in that: The trusted service management in step 2 adopts data security exchange based on the trust of user behavior, including client-side driver, unified security exchange protocol in the transmission process, protocol analysis, policy filtering, user behavior trust analysis and log audit in the isolation exchange center; The specific security exchange protocol is: The protocol format P is described as: P = <h p ,A,B D ,E(B D )>, where h p is the protocol header; A is the data attribute; B D is the data block; E(B D ) is the random encrypted information of the blocks.

6. The 5G smart grid user-side data security transmission method according to claim 1, characterized in that: The power grid load forecasting analysis in step 3 is a power grid load forecasting analysis compatible with demand-side resources, and specifically includes the following steps: Step a1: Solving the maximum load of a certain type of power user: The maximum load prediction model can be expressed as: Among them, P max is the forecast result of the maximum load under the action of demand-side resources; Q0 is the initial power consumption; α EE is the consumption reduction rate of this type of user under the action of energy efficiency resources; β LD is the load rate of this type of user under the action of load resources; t is the forecast time period; Step a2: Solving the maximum load of multiple types of power users on the same feeder: Assume that the maximum load of the jth type of power user on the ith transmission line under the action of demand-side resources is expressed as P max,ij , then the benefits of all users on the i-th transmission line can be expressed as: Where m is the number of power users on this feeder; P max,ij is the maximum load in the area covered by the ith transmission line; ζ2 is the simultaneous load rate of multiple power users on the same feeder.

7. The 5G smart grid user-side data security transmission method according to claim 6, characterized in that: The cluster analysis of daily load characteristics in step 3 is a power load analysis and demand-side response analysis based on a clustering algorithm, and specifically includes the following steps: Step b1: data preprocessing; Step b2: Entropy analysis: For any user n, the cluster center C of its daily average load curve is i The frequency matching its daily average load curve is P n (C i ), then the entropy value of the user is: If all cluster centers have equal matching probabilities, the entropy value is the highest; if all load curves are classified into one cluster, the entropy value is the lowest; Step b3: Data analysis: Starting from a set of initial clusters generated by the standard K-means algorithm, the number of initial clusters K = k0, when the variance of the load curve S(t) in the data to its typical load curve exceeds the critical value, the adaptive K-means algorithm will add new cluster centers: Here, θ is the choice of critical value.

8. The 5G smart grid user-side data security transmission method according to claim 7, characterized in that: The quantitative evaluation of demand-side flexibility in step 3 is the evaluation of demand-side response potential based on the prediction of massive load data, specifically: the demand response potential value P drp It is defined as the product of the average peak power consumption that is higher than 80% of the daily maximum load and the average load reduction rate of the industry demand response: in, is the load value corresponding to user i at peak time t1, t2, ..., tM; λ i is the average load reduction rate under the demand response measures of user i; and the actual demand response peak reduction potential should be less than the system peak time. The value of the actual load of demand response minus the minimum load is the maximum value of the demand side response potential. The difference between the user's peak time and the daily minimum load is obtained from the observation of the i-type load curve data: Where h represents the peak time of the system; p imin is the minimum daily load in the typical load curve of user i; therefore, the final demand response peak load reduction potential of type i is:

9. The 5G smart grid user-side data security transmission method according to claim 8, characterized in that: The load regression prediction in step 3 adopts a "clustering-regression" large power user short-term load prediction model based on cluster load characteristic curves. The model adopts a distributed TensorFlow framework and includes the following steps: Step c1: Data preprocessing: The attribute power and peak power L considered when clustering the power consumption characteristics of large users peak , level charge L flat , valley charge L valley And the maximum power P max The original data is normalized and processed to obtain four attributes that reflect the characteristics of electricity consumption: peak-to-total ratio r1, flat-to-valley ratio r3, valley-to-total ratio r2, and load factor η; Where, L overall is the total power; Step c2: Clustering of large power users: clustering users according to four attributes; Step c3: Cluster load characteristic curve: Users with similar power consumption characteristics are grouped into one cluster. i n i Users, C cluster load curves are obtained according to the following formula: Where Load(i) is the cluster C i Cluster load curve vector; l j Cluster C i The average load of the included users at the jth moment; Step c4: regression prediction.

Images