A power supply test data management method and system based on machine learning

Abstract

The present invention belongs to the technical field of data management, and discloses a power supply test data management method and system based on machine learning. The method comprises the following steps: based on a trusted institution, key generation and identity registration are performed; based on a data analysis center, a machine learning model and a blockchain storage network are constructed; based on a power supply test device, real-time power supply test data is collected, encrypted and signed, and uploaded to a data analysis center; based on a data analysis center, signature verification and decryption are performed; power supply test data analysis is performed; power supply operation optimization strategy generation is performed; distributed storage is performed; real-time power supply operation optimization strategy is returned to the corresponding power supply test device; and the real-time power supply operation optimization strategy after decryption is executed. The present invention solves the problems of data security, low data processing efficiency, centralized storage risks, lack of intelligent analysis, and insufficient feedback of analysis results existing in the prior art.

Description

Technical Field

[0001] The present invention belongs to the technical field of data management, and in particular relates to a power supply test data management method and system based on machine learning. Background Art

[0002] With the rapid development of science and technology, power supply systems are increasingly used in various fields. Their stability and reliability are crucial to the normal operation of the entire system. Therefore, regular testing and monitoring of power supply systems to ensure their performance and safety has become an industry consensus. However, existing power supply test data management methods have the following limitations:

[0003] 1) Data security issues: Traditional power supply test data management methods often lack effective data encryption and signature mechanisms, making data vulnerable to security threats such as leakage and tampering during transmission and storage.

[0004] 2) Low data processing efficiency: When faced with large amounts of power supply test data, traditional data processing methods are often inefficient and cannot meet the needs of real-time analysis and optimization.

[0005] 3) Centralized storage risk: Existing power supply test data management mostly adopts centralized storage. Once the data center fails or is attacked, it will lead to data loss and affect the normal operation of the entire system.

[0006] 4) Lack of intelligent analysis: Traditional power supply test data analysis mainly relies on manual experience and lacks intelligent data analysis models. It is difficult to mine deep information in the data, which limits the further improvement of power supply system performance.

[0007] 5) Insufficient feedback on analysis results: Existing power supply test data management methods lack an effective feedback mechanism and are unable to dynamically adjust the operating status of the power supply system based on real-time data analysis results to achieve energy consumption reduction and performance improvement. Summary of the Invention

[0008] In order to solve the problems of data security, low data processing efficiency, centralized storage risks, lack of intelligent analysis and insufficient feedback of analysis results in the existing technology, the purpose of the present invention is to provide a power supply test data management method and system based on machine learning.

[0009] The technical solution adopted in the present invention is:

[0010] A power supply test data management method based on machine learning includes the following steps:

[0011] Based on the trusted institution, key generation and identity registration are performed for all power test devices connected to the data analysis center, and the public-private key pair and signature information of each power test device are obtained;

[0012] Based on the data analysis center, using machine learning algorithms, we build a power test data analysis model and a power operation optimization model, and build a blockchain storage network based on several server nodes;

[0013] Based on the power supply test device, the real-time power supply test data of the power supply is collected, and the real-time power supply test data is encrypted and signed according to the private key and signature information, and the encrypted real-time power supply test data and real-time signature data are obtained and uploaded to the data analysis center;

[0014] Based on the data analysis center, the trusted organization is called to perform signature verification on the real-time signature data. After the signature verification is passed, the encrypted real-time power test data is decrypted according to the public key of the corresponding power test device to obtain the decrypted real-time power test data;

[0015] Based on the data analysis center, according to the decrypted real-time power test data, the power test data analysis model is used to analyze the power test data and obtain the real-time power test data analysis results;

[0016] Based on the data analysis center, according to the real-time power test data analysis results, the power operation optimization model is used to generate the power operation optimization strategy and obtain the real-time power operation optimization strategy;

[0017] Based on the data analysis center, the blockchain storage network is used to distribute the decrypted real-time power test data, the real-time power test data analysis results, and the real-time power operation optimization strategy;

[0018] Based on the data analysis center, the real-time power supply operation optimization strategy is encrypted according to the public key of the corresponding power supply test device, the encrypted real-time power supply operation optimization strategy is obtained, and returned to the corresponding power supply test device;

[0019] Based on the power supply test device, the encrypted real-time power supply operation optimization strategy is decrypted according to the private key to obtain the decrypted real-time power supply operation optimization strategy, and the decrypted real-time power supply operation optimization strategy is executed.

[0020] Furthermore, based on the trusted institution, key generation and identity registration are performed for all power supply test devices connected to the data analysis center to obtain the public-private key pair and signature information of each power supply test device, including the following steps:

[0021] Based on the data analysis center, collect attribute information and device IDs of all power supply test devices connected to the data analysis center, and send the attribute information and device IDs to a trusted organization;

[0022] Based on the trusted organization, key generation is performed to obtain an initial key and public parameters, and key generation is performed based on the initial key, public parameters and attribute information of the power supply test device to obtain a public-private key pair of the power supply test device;

[0023] Based on the trusted organization, the identity of the power test device is registered according to the public-private key pair and the device ID, and the signature information of the power test device is obtained;

[0024] The private key and signature information in the public-private key pair are returned to the corresponding power supply test device, and the public key in the public-private key pair is published to the data analysis center.

[0025] Furthermore, based on the data analysis center, a machine learning algorithm is used to build a power test data analysis model and a power operation optimization model. Based on several server nodes, a blockchain storage network is built, including the following steps:

[0026] Based on the data analysis center, a number of historical power supply test data are collected and pre-processed to obtain a number of pre-processed historical power supply test data;

[0027] Based on a number of pre-processed historical power supply test data, a deep learning algorithm is used to build a power supply test data analysis model and generate a number of historical power supply test data analysis results;

[0028] Based on the analysis results of several historical power supply test data, a reinforcement learning algorithm is used to build a power supply operation optimization model, and several historical power supply operation optimization strategies and several historical power supply operation optimization experiences are generated;

[0029] Several server nodes in the data analysis center are distributedly connected to obtain a blockchain storage network.

[0030] Furthermore, the power test data analysis model is constructed based on the RF-Attention-BiLSTM algorithm, and the power test data analysis model includes a key feature extraction module constructed based on the RF algorithm, an attention weight module constructed based on the Attention mechanism, and a power test data analysis module constructed based on the BiLSTM algorithm;

[0031] The power operation optimization model is built based on the DQN algorithm, and the power operation optimization model includes a deep Q network, an intelligent agent, and an experience replay pool.

[0032] Furthermore, based on the data analysis center, according to the decrypted real-time power test data, a power test data analysis model is used to analyze the power test data to obtain the real-time power test data analysis results, including the following steps:

[0033] Based on the data analysis center, the decrypted real-time power supply test data is input into the power supply test data analysis model;

[0034] Using the key feature extraction module of the power test data analysis model, extract N real-time key features of the decrypted real-time power test data, where N is the total number of key features;

[0035] According to the attention weight value preset by the attention weight module of the power test data analysis model, feature splicing is performed on N real-time key features to obtain real-time splicing features;

[0036] According to the real-time splicing feature, the power test data analysis module of the power test data analysis model is used to perform power test data analysis to obtain real-time power test data analysis results.

[0037] Furthermore, based on the data analysis center, according to the analysis results of the real-time power supply test data, the power supply operation optimization model is used to generate a power supply operation optimization strategy, and the real-time power supply operation optimization strategy is obtained, which includes the following steps:

[0038] Based on the analysis results of real-time power test data, the data analysis center searches for several historical power operation optimization experiences in the experience playback pool of the power operation optimization model to obtain several matching historical power operation optimization experiences.

[0039] According to a number of matching historical power supply operation optimization experiences, the action space of the power supply operation optimization model is updated to obtain an updated action space;

[0040] According to the analysis results of the real-time power supply test data, the state space of the power supply operation optimization model is updated to obtain an updated state space;

[0041] Input the updated state space into the deep Q network of the power operation optimization model, and use the deep Q network to generate the Q value of each possible action in the updated action space based on the updated state space;

[0042] Using the agent, according to the reward function, obtain the reward value of each possible action in the updated action space, and update the Q value of the possible action according to the reward value to obtain the updated Q value of each possible action;

[0043] Repeat the above Q-value update steps until the number of iterations reaches the iteration threshold, and use the greedy strategy to select the possible action with the highest updated Q-value in each iteration as the execution action;

[0044] Using intelligent agents, several execution actions are integrated to obtain real-time power operation optimization strategies.

[0045] Furthermore, based on the data analysis center, a blockchain storage network is used to distribute the decrypted real-time power supply test data, the real-time power supply test data analysis results, and the real-time power supply operation optimization strategy, including the following steps:

[0046] Based on the data analysis center, according to the layering mechanism, several server nodes in the blockchain storage network are divided into the consensus layer, and the remaining server nodes are divided into the storage layer;

[0047] Compress and package the decrypted real-time power supply test data, the real-time power supply test data analysis results, and the real-time power supply operation optimization strategy to obtain a real-time data packet, and slice the real-time data packet to obtain a number of real-time data slices;

[0048] Assign several real-time data shards to the storage layer, create corresponding real-time transaction data, use the consensus layer to reach consensus on the real-time transaction data of several real-time data shards, and after the consensus is successful, perform distributed storage on the assigned real-time data shards based on the storage layer.

[0049] Furthermore, based on the data analysis center, according to the layering mechanism, several server nodes in the blockchain storage network are divided into a consensus layer, and the remaining server nodes are divided into a storage layer, including the following steps:

[0050] Based on the data analysis center, all server nodes in the blockchain storage network are initialized to obtain several initialized server nodes;

[0051] Using a reward mechanism, we obtain the real-time honor values ​​of several initialized server nodes participating in the consensus process, and use the top M initialized server nodes as consensus nodes, and the remaining initialized server nodes as storage nodes, where M is the total number of consensus nodes.

[0052] Several consensus nodes constitute the consensus layer, and several storage nodes constitute the storage layer.

[0053] Furthermore, several real-time data shards are allocated to the storage layer, corresponding real-time transaction data is created, and the consensus layer is used to reach a consensus on the real-time transaction data of several real-time data shards. After the consensus is successful, the allocated real-time data shards are distributedly stored based on the storage layer, including the following steps:

[0054] According to the shard copy mechanism, a number of real-time replica data shards are generated for each real-time data shard, thereby obtaining a number of expanded real-time data shards;

[0055] Use the consistent hashing algorithm to distribute the expanded real-time data shards to the storage nodes of the storage layer, and return the real-time storage location data of each expanded real-time data shard;

[0056] Obtain the real-time data hash value of each expanded real-time data shard, and create corresponding real-time transaction data based on the corresponding real-time data hash value and real-time storage location data;

[0057] Broadcast each real-time transaction data to all consensus nodes in the consensus layer, and reach consensus on the real-time transaction data based on the consensus nodes;

[0058] Based on the storage node, the expanded real-time data shards corresponding to the real-time transaction data that has achieved successful consensus are stored locally;

[0059] Repeat the above steps until each consensus is successfully expanded and the real-time data shards are successfully distributed and stored in the corresponding storage nodes.

[0060] A power supply test data management system based on machine learning is used to implement a power supply test data management method. The system includes a data analysis center, a trusted organization, and a plurality of power supply test devices. Each power supply test device is respectively connected to the data analysis center and the trusted organization, and the trusted organization is connected to the data analysis center.

[0061] The data analysis center includes a model building unit, a signature verification and decryption unit, a data analysis unit, a power operation optimization unit, a data distributed storage unit and a data sending unit which are connected in sequence. The data analysis unit is provided with a power test data analysis model, the power operation optimization unit is provided with a power operation optimization model, the data distributed storage unit is provided with a blockchain storage network, the signature verification and decryption unit is respectively communicated with a trusted organization and several power test devices, and the data sending unit is respectively communicated with several power test devices.

[0062] The beneficial effects of the present invention are:

[0063] The present invention provides a power supply test data management method and system based on machine learning, which adopt asymmetric encryption technology and digital signature technology to ensure the confidentiality and integrity of power supply test data during transmission, effectively solving the problem of insufficient data security in traditional methods and preventing the risk of data leakage and tampering; through the distributed storage technology of blockchain, an additional security layer is provided for data storage, ensuring that data cannot be tampered with and the source of data can be traced. Compared with traditional centralized storage, distributed storage reduces the risk of single point failure and improves the reliability and robustness of data storage; using deep learning algorithms to build a data analysis model, it is possible to efficiently process a large amount of power supply test data and extract valuable information from it. Compared with traditional manual analysis methods, it significantly improves data processing efficiency and intelligence level. Through real-time analysis of power supply test data, it is possible to quickly identify potential problems of the power supply, providing technical support for real-time monitoring and early warning; using reinforcement learning algorithms to build a power supply operation optimization model, it is possible to dynamically adjust the operating status of the power supply according to the real-time data analysis results, providing an effective feedback mechanism for the analysis results, and helping to achieve energy consumption optimization and system performance improvement.

[0064] Other beneficial effects of the present invention will be further described in the specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 This is a flowchart of the power supply test data management method based on machine learning in the present invention.

[0066] Figure 2 This is a structural block diagram of the power supply test data management system based on machine learning in the present invention. DETAILED DESCRIPTION

[0067] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments.

[0068] Example 1:

[0069] like Figure 1 As shown, this embodiment provides a power supply test data management method based on machine learning, comprising the following steps:

[0070] S1: Based on a trusted institution, perform key generation and identity registration for all power supply test devices connected to the data analysis center, and obtain the public-private key pair and signature information of each power supply test device, including the following steps:

[0071] S1-1: Based on the data analysis center, collect attribute information and device IDs of all power supply test devices connected to the data analysis center, and send the attribute information and device IDs to the trusted organization;

[0072] S1-2: Based on the trusted institution, perform key generation to obtain an initial key and public parameters, and perform key generation based on the initial key, public parameters, and attribute information of the power supply test device to obtain a public-private key pair of the power supply test device;

[0073] The formula is:

[0074]

[0075] Where, is a public parameter; is the master key; is the initial key; Integer domain Random number; Both are target hash functions; Both are cyclic groups The random number generated by the generator; is a random number Bilinear mapping of ;

[0076]

[0077] Where, Power supply test equipment The private key of Integer domain Random number; Power supply test equipment The private key parameters; For public parameters The target hash function; Indicator for power supply test device; is the master key; is the initial key; Power supply test equipment 's public key; is a cyclic group The random number generated by the generator; Power supply test equipment Attribute information of

[0078] S1-3: Based on the trusted institution, perform identity registration based on the public-private key pair and device ID of the power supply test device to obtain the signature information of the power supply test device;

[0079] The formula is:

[0080]

[0081] Where, is a random number; Power supply test equipment Registration parameters; Power supply test equipment Registration ID; and the corresponding Signature information that constitutes the power supply test device ; Hash function for the target hash; Power supply test equipment Device ID; is a prime number order; It is the base point of the prime domain; To find the remainder function;

[0082] S1-4: Return the private key and signature information in the public-private key pair to the corresponding power supply test device, and publish the public key in the public-private key pair to the data analysis center;

[0083] S2: Based on the data analysis center, using machine learning algorithms, we build a power test data analysis model and a power operation optimization model. We also build a blockchain storage network based on several server nodes, including the following steps:

[0084] S2-1: Based on the data analysis center, collect some historical power supply test data and perform preprocessing to obtain some preprocessed historical power supply test data;

[0085] Power supply test data typically includes the following parameters, which help evaluate the performance, stability, and reliability of the power supply:

[0086] Output voltage: the actual output voltage value of the power supply under load;

[0087] Output current: The maximum or actual output current provided by the power supply;

[0088] Output power: The power provided by the power supply to the load, usually the product of output voltage and output current;

[0089] Efficiency: The ratio of output power to input power, usually expressed as a percentage;

[0090] Input voltage: The input voltage range required by the power supply;

[0091] Input current: The input current of the power supply at a specific input voltage;

[0092] Power factor: The phase relationship between input current and input voltage, the ideal value is 1;

[0093] Current harmonic distortion: harmonic content of input current;

[0094] Voltage harmonic distortion: harmonic content of output voltage;

[0095] Ripple and noise: The AC component in the output voltage, usually expressed as peak-to-peak or RMS value;

[0096] Transient response: The output voltage responds to rapid changes in load;

[0097] Overload protection: The protection feature of the power supply when the output is overloaded;

[0098] Short circuit protection: The protection feature of the power supply when the output is short-circuited;

[0099] Overvoltage protection: The protection feature of the power supply when the output voltage is too high;

[0100] Undervoltage protection: a protection feature of the power supply when the output voltage is too low;

[0101] Temperature range: The ambient temperature range in which the power supply operates normally;

[0102] Thermal management: The heat dissipation performance of the power supply, including temperature rise and heat dissipation design;

[0103] Electromagnetic compatibility: the power supply's anti-interference capability and external interference level;

[0104] Stability: The output stability of the power supply during long-term operation or environmental changes;

[0105] Reliability: the expected lifespan and failure rate of the power supply;

[0106] Preprocessing includes data format conversion, data cleaning and screening to remove duplicate, erroneous and noisy data, providing data support for subsequent model training;

[0107] S2-2: Based on a number of pre-processed historical power supply test data, a deep learning algorithm is used to construct a power supply test data analysis model and generate a number of historical power supply test data analysis results;

[0108] The power test data analysis model is built based on the Random Forest (RF)-Attention-Bidirectional Long Short-Term Memory (BiLSTM) algorithm. It includes a key feature extraction module based on the RF algorithm, an attention weighting module based on the Attention mechanism, and a power test data analysis module based on the BiLSTM algorithm.

[0109] The key feature extraction module uses an internal Classification and Regression Tree (CART) to screen key features of the input data and extract several key features related to the data prediction label. The attention weight module uses preset attention weights to splice key features and integrate scattered features, thereby enhancing the influence of key features on label prediction. The power test data analysis module uses a multi-layer hidden layer structure to explore the potential connection between key features and data prediction labels. It trains its own network structure with a large number of key features to predict new data features and achieve fast, efficient, and accurate prediction of data labels.

[0110] The power supply test data analysis results include the following:

[0111] Predictive maintenance recommendations: Use deep learning models to predict potential power supply failure or degradation and provide maintenance recommendations; predict the remaining service life of the power supply;

[0112] Anomaly detection: Detects unusual patterns in test data to identify possible failures or unusual behavior; anomaly scoring and confidence levels indicate the extent to which the data deviates from normal behavior;

[0113] Performance trend analysis: Analyze the performance trends of power supplies over time, such as efficiency and output voltage stability; use time series analysis to predict future performance changes;

[0114] Feature Importance Ranking: Identify which test parameters have the greatest impact on power supply performance; determine which features are most critical to the prediction model;

[0115] High-precision parameter estimation: Use deep learning models to estimate power supply parameters (such as efficiency and power factor) with high accuracy; calibrate and optimize the measurement results of test equipment;

[0116] Classification and labeling: Classify power supplies into different categories or grades based on performance and reliability; automatically label test data for subsequent analysis;

[0117] Model uncertainty assessment:

[0118] Assess the uncertainty of deep learning model predictions and provide confidence intervals for prediction results; understand the reliability and potential risks of model predictions;

[0119] S2-3: Based on the analysis results of several historical power supply test data, a reinforcement learning algorithm is used to build a power supply operation optimization model, and several historical power supply operation optimization strategies and several historical power supply operation optimization experiences are generated;

[0120] The power operation optimization model is built based on the Deep Q Network (DQN) algorithm, and includes the Deep Q Network, an intelligent agent, and an experience replay pool;

[0121] Building a power supply operation optimization model includes the following steps:

[0122] A-1: Generate a simulation environment for the DQN algorithm using a power operation optimization problem to build an intelligent agent and experience replay pool;

[0123] A-2: Based on each power supply operation state type corresponding to the historical coal purchase volume forecast results, define the state space of the DQN algorithm, and the parameters of the state space correspond to each power supply operation state;

[0124] A-3: Define the action space of the DQN algorithm based on the actions that need to be output by power operation optimization;

[0125] A-4: Based on the possible impact of each action in the action space, define the reward function of the DQN algorithm to evaluate the quality or impact of the action;

[0126] A-5: Construct the input layer, several hidden layers, and output layer of the deep Q network, connect the input layer to the state space, and connect the output layer to the action space;

[0127] A-6: Based on the state space, action space, and reward function, and according to the analysis results of several historical power supply test data, the deep Q network and intelligent agent are optimized and trained to build a power supply operation optimization model, and several historical power supply operation optimization experiences are generated;

[0128] A-7: Store several historical power supply operation optimization experiences into the experience replay pool of the power supply operation optimization model;

[0129] The power operation optimization strategy includes the following:

[0130] Dynamic load management strategy: Adjusts power output to optimize efficiency and response time based on current load conditions and historical data analysis; adjusts the on / off state of the power module in real time to adapt to load changes;

[0131] Energy allocation strategy: Develop optimal energy allocation plans to ensure power system stability and efficiency under different load conditions; in multi-power module systems, dynamically allocate loads to maximize overall system performance;

[0132] Failure prevention measures: Based on predictive maintenance recommendations, adjust operating parameters in advance to avoid potential failures; implement preventive maintenance plans to reduce unplanned downtime;

[0133] Efficiency optimization: Identify and implement operating modes that improve power efficiency; prioritize features and adjust key parameters to improve energy efficiency;

[0134] Thermal management strategy: Optimizes heat dissipation to keep the power supply within the optimal operating temperature range; dynamically adjusts the cooling system based on ambient temperature and load conditions.

[0135] Power protection strategy: Based on anomaly detection results, implement protection measures to prevent power overload, short circuit, etc.; adjust protection thresholds to balance protection effectiveness and system availability;

[0136] Output quality improvement strategy: Adjust power supply parameters to reduce ripple and noise and improve output voltage quality; based on performance trend analysis, adjust power supply design to reduce long-term performance degradation;

[0137] Startup and shutdown strategies: Optimize the power startup and shutdown sequence to reduce the impact on the system; automatically adjust the power startup delay and shutdown time according to demand and environmental conditions;

[0138] Redundancy configuration optimization: For redundant power supply systems, optimize the redundant configuration to improve system reliability and economy; adjust the redundant power supply operation strategy based on the reliability and performance level of the power supply;

[0139] S2-4: Distribute and connect several server nodes in the data analysis center to obtain a blockchain storage network;

[0140] S3: Based on the power supply test device, real-time power supply test data of the power supply is collected. The real-time power supply test data is encrypted and signed according to the private key and signature information. The encrypted real-time power supply test data and real-time signature data are obtained and uploaded to the data analysis center.

[0141] The formula is:

[0142]

[0143] Where, It is the encrypted real-time power supply test data; is an asymmetric encryption function; Real-time power supply test data; Power supply test equipment The private key of

[0144]

[0145] Where, is a random number; is a prime number order; It is the base point of the prime domain; Hash function for the target hash; Signature information Medium power supply test device Registration parameters; Signature information Medium power supply test device Registration ID; Power supply test equipment The device ID of ; Power supply test equipment Signature parameters;

[0146] S4: Based on the data analysis center, the trusted organization is called to perform signature verification on the real-time signature data. After the signature verification is passed, the encrypted real-time power test data is decrypted according to the public key of the corresponding power test device to obtain the decrypted real-time power test data;

[0147] The formula is:

[0148]

[0149] Where, Power supply test equipment Signature verification parameters; Power supply test equipment If the left formula is equal to the right formula, the signature verification is successful;

[0150]

[0151] Where, It is the real-time power supply test data after decryption; is an asymmetric decryption function; It is the encrypted real-time power supply test data; Power supply test equipment 's public key;

[0152] S5: Based on the data analysis center, the power test data is analyzed using the power test data analysis model according to the decrypted real-time power test data to obtain the real-time power test data analysis results, including the following steps:

[0153] S5-1: Based on the data analysis center, the decrypted real-time power supply test data is input into the power supply test data analysis model;

[0154] S5-2: Using the key feature extraction module of the power supply test data analysis model, extract N real-time key features of the decrypted real-time power supply test data, where N is the total number of key features, including the following steps:

[0155] S5-2-1: Use the RF structure trained in the key feature extraction module to extract the feature contribution of N candidate features in the decrypted real-time power test data;

[0156] The formula is:

[0157]

[0158] Where, For the Feature contribution of candidate features; For the The candidate features are in the random forest The feature contribution of each tree; is the CART tree indicator; is an alternative characteristic indicator; is the total number of CARTs;

[0159]

[0160] Where, CART tree node of random forest m ,node and nodes The Gini index; CART tree node m Medium Category the proportion; is the total number of categories; m 、 、 is the node indicator; is the category indicator;

[0161] S5-2-2: normalizing the feature contributions of several candidate features to obtain corresponding normalized feature contributions;

[0162] The formula is:

[0163]

[0164] Where, is the contribution of the feature after normalization; J is the total number of candidate features;

[0165] S5-2-3: Generate feature selection standard values ​​for several candidate features based on the normalized feature contribution;

[0166] The formula is:

[0167]

[0168] Where, For the Feature selection criterion values ​​for candidate features; For the The contribution of the candidate features after normalization; is an alternative characteristic indicator;

[0169] S5-2-4: Sort the candidate features in descending order according to the feature selection standard value, and select the top N candidate features as real-time key features to obtain N real-time key features;

[0170] S5-3: Based on the attention weight value preset by the attention weight module of the power test data analysis model, feature splicing is performed on N real-time key features to obtain a real-time splicing feature;

[0171] S5-4: Based on the real-time splicing features, the power test data analysis module of the power test data analysis model is used to analyze the power test data to obtain real-time power test data analysis results;

[0172] S6: Based on the data analysis center, according to the real-time power test data analysis results, the power operation optimization model is used to generate a power operation optimization strategy, and the real-time power operation optimization strategy is obtained, including the following steps:

[0173] S6-1: Based on the analysis results of the real-time power test data, the data analysis center searches for several historical power operation optimization experiences in the experience playback pool of the power operation optimization model to obtain several matching historical power operation optimization experiences.

[0174] S6-2: Update the action space of the power operation optimization model based on some matching historical power operation optimization experiences to obtain the updated action space ,in, For the updated Action value, is the action indicator, is the total number of action space dimensions;

[0175] S6-3: Update the state space of the power supply operation optimization model based on the real-time power supply test data analysis results to obtain the updated state space ,in, For the updated Status value, is the status indicator, is the total number of state space dimensions;

[0176] S6-4: Input the updated state space into the deep Q network of the power operation optimization model, and use the deep Q network to generate the Q value of each possible action in the updated action space based on the updated state space;

[0177] S6-5: Using the agent, according to the reward function, obtain the reward value of each possible action in the updated action space, and update the Q value of the possible action according to the reward value to obtain the updated Q value of each possible action;

[0178] The formula is:

[0179]

[0180] Where, The updated status value and updated action value The corresponding updated Q value; Status value and action value The corresponding Q value of the possible action; is the learning rate; is the Q value of the highest possible action; It is a comprehensive indicator; To update the parameters;

[0181] S6-6: Repeat the above Q-value update steps until the number of iterations reaches the iteration threshold, and use the greedy strategy to select the possible action with the highest updated Q-value in each iteration as the execution action;

[0182] S6-7: Use the agent to integrate several execution actions to obtain a real-time power operation optimization strategy;

[0183] S7: Based on the data analysis center, using the blockchain storage network, the decrypted real-time power test data, the real-time power test data analysis results, and the real-time power operation optimization strategy are distributedly stored, including the following steps:

[0184] S7-1: Based on the data analysis center and according to the layering mechanism, several server nodes in the blockchain storage network are divided into the consensus layer, and the remaining server nodes are divided into the storage layer, including the following steps:

[0185] S7-1-1: Based on the data analysis center, initialize all server nodes in the blockchain storage network to obtain several initialized server nodes;

[0186] S7-1-2: Use the reward mechanism to obtain the real-time honor values ​​of several initialized server nodes participating in the consensus process, and use the top M initialized server nodes in terms of real-time honor values ​​as consensus nodes, and the remaining initialized server nodes as storage nodes, where M is the total number of consensus nodes;

[0187] The formula for real-time reputation value is:

[0188]

[0189] Where, For server nodes Real-time reputation value; Indicates the amount of server nodes; is the initial reputation value of each server node; is the weight of the reward mechanism; The single reward value of the server node; is the weight of the penalty mechanism; is the single penalty value of the server node;

[0190]

[0191] Where, For server nodes Number of successful consensus participations; e is a natural constant;

[0192]

[0193] Where, For server nodes Number of consensus failures or other malicious behaviors; The threshold for the number of penalty triggers;

[0194] S7-1-3: Several consensus nodes form a consensus layer, and several storage nodes form a storage layer;

[0195] S7-2: compressing and packaging the decrypted real-time power supply test data, the real-time power supply test data analysis results, and the real-time power supply operation optimization strategy to obtain a real-time data packet, and fragmenting the real-time data packet to obtain a number of real-time data fragments;

[0196] S7-3: Allocate several real-time data shards to the storage layer, create corresponding real-time transaction data, and use the consensus layer to reach consensus on the real-time transaction data of several real-time data shards. After the consensus is successful, the allocated real-time data shards are distributedly stored based on the storage layer, including the following steps:

[0197] S7-3-1: Generate several real-time replica data shards for each real-time data shard according to the shard copy mechanism, and obtain several expanded real-time data shards;

[0198] S7-3-2: Use the consistent hashing algorithm to distribute the expanded real-time data shards to the storage nodes of the storage layer, and return the real-time storage location data of each expanded real-time data shard;

[0199] S7-3-3: Obtain the real-time data hash value of each expanded real-time data shard, and create corresponding real-time transaction data based on the corresponding real-time data hash value and real-time storage location data;

[0200] S7-3-4: Broadcast each real-time transaction data to all consensus nodes in the consensus layer, and reach consensus on the real-time transaction data based on the consensus nodes;

[0201] S7-3-5: Based on the storage node, the expanded real-time data shards corresponding to the real-time transaction data that has achieved successful consensus are locally stored;

[0202] S7-3-6: Repeat the above steps until each consensus-successful expansion of real-time data shards is successfully distributed and stored in the corresponding storage nodes;

[0203] S8: Based on the data analysis center, the real-time power supply operation optimization strategy is encrypted according to the public key of the corresponding power supply test device, the encrypted real-time power supply operation optimization strategy is obtained, and the encrypted real-time power supply operation optimization strategy is returned to the corresponding power supply test device;

[0204] The formula is:

[0205]

[0206] Where, Optimize strategies for real-time power operation after encryption; is an asymmetric encryption function; Optimize strategies for real-time power operation;

[0207] S9: Based on the power supply test device, decrypt the encrypted real-time power supply operation optimization policy according to the private key, obtain the decrypted real-time power supply operation optimization policy, and execute the decrypted real-time power supply operation optimization policy;

[0208]

[0209] Where, Optimize strategies for real-time power operation after decryption; is an asymmetric decryption function; Run optimized policies for real-time power after encryption.

[0210] Example 2:

[0211] like Figure 2 As shown, this embodiment provides a power supply test data management system based on machine learning, which is used to implement a power supply test data management method. The system includes a data analysis center, a trusted organization, and a plurality of power supply test devices. Each power supply test device is respectively connected to the data analysis center and the trusted organization, and the trusted organization is connected to the data analysis center.

[0212] A trusted authority is used to generate keys and register identities for all power supply test devices connected to the data analysis center, and obtain the public and private key pairs and signature information of each power supply test device;

[0213] A power supply test device is used to collect real-time power supply test data of the power supply, encrypt and sign the real-time power supply test data according to the private key and signature information, obtain the encrypted real-time power supply test data and real-time signature data, and upload them to the data analysis center; decrypt the encrypted real-time power supply operation optimization strategy according to the private key, obtain the decrypted real-time power supply operation optimization strategy, and execute the decrypted real-time power supply operation optimization strategy;

[0214] The data analysis center includes a model building unit, a signature verification and decryption unit, a data analysis unit, a power operation optimization unit, a data distributed storage unit, and a data sending unit, which are connected in sequence. The data analysis unit is provided with a power test data analysis model, the power operation optimization unit is provided with a power operation optimization model, the data distributed storage unit is provided with a blockchain storage network, the signature verification and decryption unit is respectively connected to a trusted institution and a plurality of power test devices in communication, and the data sending unit is respectively connected to a plurality of power test devices in communication;

[0215] A model building unit, which is used to use machine learning algorithms to build a power test data analysis model and a power operation optimization model, and to build a blockchain storage network based on a number of server nodes;

[0216] The signature verification and decryption unit is used to call a trusted institution to perform signature verification on the real-time signature data. After the signature verification is passed, the encrypted real-time power test data is decrypted according to the public key of the corresponding power test device to obtain the decrypted real-time power test data;

[0217] A data analysis unit is used to analyze the power test data according to the decrypted real-time power test data using a power test data analysis model to obtain real-time power test data analysis results;

[0218] A power operation optimization unit is used to generate a power operation optimization strategy based on the real-time power test data analysis results using a power operation optimization model to obtain a real-time power operation optimization strategy;

[0219] A data distributed storage unit is used to use a blockchain storage network to distribute and store decrypted real-time power supply test data, real-time power supply test data analysis results, and real-time power supply operation optimization strategies;

[0220] The data sending unit is used to encrypt the real-time power supply operation optimization strategy according to the public key of the corresponding power supply test device, obtain the encrypted real-time power supply operation optimization strategy, and return it to the corresponding power supply test device.

[0221] The present invention provides a power supply test data management method and system based on machine learning, which adopt asymmetric encryption technology and digital signature technology to ensure the confidentiality and integrity of power supply test data during transmission, effectively solving the problem of insufficient data security in traditional methods and preventing the risk of data leakage and tampering; through the distributed storage technology of blockchain, an additional security layer is provided for data storage, ensuring that data cannot be tampered with and the source of data can be traced. Compared with traditional centralized storage, distributed storage reduces the risk of single point failure and improves the reliability and robustness of data storage; using deep learning algorithms to build a data analysis model, it is possible to efficiently process a large amount of power supply test data and extract valuable information from it. Compared with traditional manual analysis methods, it significantly improves data processing efficiency and intelligence level. Through real-time analysis of power supply test data, it is possible to quickly identify potential problems of the power supply, providing technical support for real-time monitoring and early warning; using reinforcement learning algorithms to build a power supply operation optimization model, it is possible to dynamically adjust the operating status of the power supply according to the real-time data analysis results, providing an effective feedback mechanism for the analysis results, and helping to achieve energy consumption optimization and system performance improvement.

[0222] The present invention is not limited to the above optional embodiments. Anyone can derive various other forms of products based on the teachings of the present invention. The above specific embodiments should not be construed as limiting the scope of protection of the present invention. The scope of protection of the present invention shall be based on the scope defined in the claims, and the description can be used to interpret the claims.

Claims

1. A power supply test data management method based on machine learning, characterized in that: The steps include: Based on the trusted institution, key generation and identity registration are performed for all power supply test devices connected to the data analysis center to obtain the public-private key pair and signature information of each power supply test device, including the following steps: Based on the data analysis center, collect attribute information and device IDs of all power supply test devices connected to the data analysis center, and send the attribute information and device IDs to a trusted institution; Based on the trusted institution, key generation is performed to obtain an initial key and public parameters, and key generation is performed according to the initial key, public parameters and attribute information of the power supply test device to obtain a public-private key pair of the power supply test device; Based on the trusted institution, the identity is registered according to the public-private key pair and the device ID of the power supply test device, and the signature information of the power supply test device is obtained; Return the private key and signature information in the public-private key pair to the corresponding power supply test device, and publish the public key in the public-private key pair to the data analysis center; Based on the data analysis center, using machine learning algorithms, we build a power test data analysis model and a power operation optimization model, and build a blockchain storage network based on several server nodes, including the following steps: Based on the data analysis center, a number of historical power supply test data are collected and preprocessed to obtain a number of preprocessed historical power supply test data; Based on a number of pre-processed historical power supply test data, a deep learning algorithm is used to build a power supply test data analysis model and generate a number of historical power supply test data analysis results; Based on the analysis results of several historical power supply test data, a power supply operation optimization model is constructed using a reinforcement learning algorithm, and several historical power supply operation optimization strategies and several historical power supply operation optimization experiences are generated; Distribute and connect several server nodes in the data analysis center to obtain a blockchain storage network; Based on the power supply test device, the real-time power supply test data of the power supply is collected, and the real-time power supply test data is encrypted and signed according to the private key and signature information, and the encrypted real-time power supply test data and real-time signature data are obtained, and uploaded to the data analysis center; Based on the data analysis center, the trusted institution is called to perform signature verification on the real-time signature data. After the signature verification is passed, the encrypted real-time power test data is decrypted according to the public key of the corresponding power test device to obtain the decrypted real-time power test data; Based on the data analysis center, according to the decrypted real-time power test data, the power test data analysis model is used to analyze the power test data to obtain the real-time power test data analysis results; Based on the data analysis center, according to the real-time power test data analysis results, the power operation optimization model is used to generate the power operation optimization strategy to obtain the real-time power operation optimization strategy; Based on the data analysis center, the blockchain storage network is used to distribute the decrypted real-time power supply test data, the real-time power supply test data analysis results, and the real-time power supply operation optimization strategy; Based on the data analysis center, the real-time power supply operation optimization strategy is encrypted according to the public key of the corresponding power supply test device, the encrypted real-time power supply operation optimization strategy is obtained, and returned to the corresponding power supply test device; Based on the power supply test device, the encrypted real-time power supply operation optimization strategy is decrypted according to the private key to obtain the decrypted real-time power supply operation optimization strategy, and the decrypted real-time power supply operation optimization strategy is executed.

2. The power supply test data management method based on machine learning according to claim 1, characterized in that: The power supply test data analysis model is constructed based on the RF-Attention-BiLSTM algorithm, and the power supply test data analysis model includes a key feature extraction module constructed based on the RF algorithm, an attention weight module constructed based on the Attention mechanism, and a power supply test data analysis module constructed based on the BiLSTM algorithm; The power operation optimization model is constructed based on the DQN algorithm, and the power operation optimization model includes a deep Q network, an intelligent agent, and an experience replay pool.

3. The power supply test data management method based on machine learning according to claim 2, characterized in that: Based on the data analysis center, according to the decrypted real-time power test data, using the power test data analysis model, the power test data analysis is performed to obtain the real-time power test data analysis results, including the following steps: Based on the data analysis center, the decrypted real-time power supply test data is input into the power supply test data analysis model; Use the key feature extraction module of the power test data analysis model to extract N real-time key features of the decrypted real-time power test data, where N is the total number of key features; According to the attention weight value preset by the attention weight module of the power supply test data analysis model, feature splicing is performed on N real-time key features to obtain real-time splicing features; According to the real-time splicing feature, the power test data analysis module of the power test data analysis model is used to perform power test data analysis to obtain real-time power test data analysis results.

4. The power supply test data management method based on machine learning according to claim 2, characterized in that: Based on the data analysis center, according to the real-time power test data analysis results, the power operation optimization model is used to generate the power operation optimization strategy, and the real-time power operation optimization strategy is obtained, including the following steps: Based on the data analysis center, according to the analysis results of real-time power test data, several historical power operation optimization experiences in the experience playback pool of the power operation optimization model are searched to obtain several matching historical power operation optimization experiences; According to a number of matching historical power supply operation optimization experiences, the action space of the power supply operation optimization model is updated to obtain an updated action space; According to the real-time power supply test data analysis results, the state space of the power supply operation optimization model is updated to obtain an updated state space; Input the updated state space into the deep Q network of the power operation optimization model, and use the deep Q network to generate the Q value of each possible action in the updated action space based on the updated state space; Using the agent, according to the reward function, obtain the reward value of each possible action in the updated action space, and according to the reward value, update the Q value of the possible action to obtain the updated Q value of each possible action; Repeat the above Q-value update steps until the number of iterations reaches the iteration number threshold, and use the greedy strategy to select the possible action with the highest updated Q-value in each iteration as the execution action; Using intelligent agents, several execution actions are integrated to obtain real-time power operation optimization strategies.

5. The power supply test data management method based on machine learning according to claim 1, characterized in that: Based on the data analysis center, the blockchain storage network is used to distribute the decrypted real-time power test data, the real-time power test data analysis results, and the real-time power operation optimization strategy, including the following steps: Based on the data analysis center, according to the hierarchical mechanism, several server nodes in the blockchain storage network are divided into the consensus layer, and the remaining server nodes are divided into the storage layer; Compress and package the decrypted real-time power supply test data, the real-time power supply test data analysis results and the real-time power supply operation optimization strategy to obtain a real-time data packet, and slice the real-time data packet to obtain a number of real-time data slices; Assign several real-time data shards to the storage layer, create corresponding real-time transaction data, use the consensus layer to reach consensus on the real-time transaction data of several real-time data shards, and after the consensus is successful, perform distributed storage on the assigned real-time data shards based on the storage layer.

6. The power supply test data management method based on machine learning according to claim 5, characterized in that: Based on the data analysis center, according to the hierarchical mechanism, several server nodes in the blockchain storage network are divided into the consensus layer, and the remaining server nodes are divided into the storage layer, including the following steps: Based on the data analysis center, all server nodes in the blockchain storage network are initialized to obtain several initialized server nodes; Use the reward mechanism to obtain the real-time honor values ​​of several initialized server nodes participating in the consensus process, and use the top M initialized server nodes as consensus nodes, and the remaining initialized server nodes as storage nodes, where M is the total number of consensus nodes; Several consensus nodes constitute the consensus layer, and several storage nodes constitute the storage layer.

7. The power supply test data management method based on machine learning according to claim 5, characterized in that: Assign several real-time data shards to the storage layer, create corresponding real-time transaction data, use the consensus layer to reach consensus on the real-time transaction data of several real-time data shards, and after the consensus is successful, perform distributed storage on the assigned real-time data shards based on the storage layer, including the following steps: According to the shard copy mechanism, a number of real-time copy data shards are generated for each real-time data shard, and a number of extended real-time data shards are obtained; Use a consistent hashing algorithm to distribute a number of expanded real-time data shards to a number of storage nodes in the storage layer, and return the real-time storage location data of each expanded real-time data shard; Obtain the real-time data hash value of each expanded real-time data shard, and create corresponding real-time transaction data according to the corresponding real-time data hash value and real-time storage location data; Broadcast each real-time transaction data to all consensus nodes of the consensus layer, and reach consensus on the real-time transaction data based on the consensus nodes; Based on the storage node, the extended real-time data shards corresponding to the real-time transaction data with successful consensus are stored locally; Repeat the above steps until each consensus-successfully expanded real-time data shard is successfully distributed and stored in the corresponding storage node.

8. A power supply test data management system based on machine learning, used to implement the power supply test data management method according to any one of claims 1 to 7, characterized in that: The system includes a data analysis center, a trusted institution and a plurality of power supply test devices, each of the power supply test devices is respectively connected to the data analysis center and the trusted institution in communication, and the trusted institution is connected to the data analysis center in communication; The data analysis center includes a model building unit, a signature verification and decryption unit, a data analysis unit, a power supply operation optimization unit, a data distributed storage unit and a data sending unit which are connected in sequence. The data analysis unit is provided with a power supply test data analysis model, the power supply operation optimization unit is provided with a power supply operation optimization model, the data distributed storage unit is provided with a blockchain storage network, the signature verification and decryption unit is respectively communicated with a trusted organization and several power supply testing devices, and the data sending unit is respectively communicated with several power supply testing devices.

Images