CN117932694A

CN117932694A - Block chain-based energy transaction and supervision system

Info

Publication number: CN117932694A
Application number: CN202410135775.3A
Authority: CN
Inventors: 田一辉; 赖恩毅; 王刚; 邓园园
Original assignee: Fujian Zhongdian Strait Intelligent Equipment Research Institute
Current assignee: Fujian Zhongdian Strait Intelligent Equipment Research Institute
Priority date: 2024-01-31
Filing date: 2024-01-31
Publication date: 2024-04-26

Abstract

The invention relates to the technical field of energy management, in particular to an energy transaction and supervision system based on a blockchain, which comprises a data integrity verification module, a time sequence analysis module, an anomaly detection module, a relation mining module, a transaction strategy optimization module, an energy consumption prediction module, a federal learning framework module and a system integration and supervision module. In the invention, the recognition capability of abnormal energy data is improved through the application of isolated forest and neural network technology, the relation mining module reveals complex association among the energy data by utilizing cluster analysis, association rule mining and network analysis methods, deep hole finding is provided for decision making, the combination of the deep neural network and reinforcement learning algorithm enables the transaction strategy to be more flexible, the energy consumption prediction module accurately predicts future demands and analyzes consumption modes, the demands of target user groups and regions are better met, and the federal learning framework module enhances data privacy and safety and improves model performance by sharing learning results across nodes.

Description

Block chain-based energy transaction and supervision system

Technical Field

The invention relates to the technical field of energy management, in particular to an energy trading and supervision system based on a blockchain.

Background

The technical field of energy management mainly utilizes information technology, in particular to blockchain technology to optimize the distribution, transaction and supervision of energy resources. The blockchain technology is applied to ensure the non-tamperability of transaction data and improve the transparency and the security of the system. This field also relates to intelligent contracts, distributed ledger technology, etc., enabling more automated, efficient energy transactions and management.

The energy transaction and supervision system based on the blockchain is a system for realizing energy transaction and supervision by using the blockchain technology. The system aims to provide a safe, transparent and efficient platform for buying and selling and monitoring energy, realize the decentralization of energy transaction, reduce the transaction cost, improve the transaction efficiency and ensure the authenticity and the non-tamper property of the transaction record. The system aims to improve the energy use and distribution efficiency through real-time supervision and data analysis, thereby achieving the effects of energy conservation, emission reduction and resource allocation optimization, recording all transactions by using a distributed account book technology of a blockchain, and ensuring the safety and transparency of data. And the transaction protocol is automatically executed by using the intelligent contract, so that manual intervention is reduced, and the transaction efficiency is improved. And the technology of the Internet of things is combined to realize real-time monitoring and management of energy consumption, and the big data analysis is utilized to optimize energy distribution and forecast future energy demands.

Traditional energy transaction and supervision systems lack effective data integrity and security guarantee and are easy to tamper with data and threaten safety. In terms of energy trend analysis, traditional methods fail to provide comprehensive and accurate predictions, lack understanding of periodic and long-term trends, have limited anomaly detection capabilities, and have difficulty in accurately identifying and processing anomaly data, resulting in significant energy management errors. In relation mining, traditional systems have difficulty in revealing complex relations and potential links between data, limiting the depth and breadth of decisions. In transaction policy optimization, the traditional method lacks flexibility and adaptability and cannot cope with market changes. The energy demand predictions are too rough and lack detailed consideration of the target user group and region. The traditional model has loopholes in terms of data privacy and security, and the efficiency and reliability of the integration and supervision module are to be improved.

Disclosure of Invention

The invention aims to solve the defects in the prior art, and provides a block chain-based energy transaction and supervision system.

In order to achieve the above purpose, the present invention adopts the following technical scheme: the energy trading and supervising system based on the blockchain comprises a data integrity verification module, a time sequence analysis module, an anomaly detection module, a relation mining module, a trading strategy optimization module, an energy consumption prediction module, a federal learning framework module and a system integration and supervising module;

The data integrity verification module is based on a blockchain technology and is used for guaranteeing data safety transmission through a hash algorithm and a digital signature, the data safety transmission relates to transaction records from an energy transaction platform, and a consensus mechanism is utilized to enable data to keep consistent, so that verified energy data is generated;

The time sequence analysis module is used for carrying out energy periodicity and trend analysis by adopting an autoregressive integral moving average and time sequence prediction method based on the verified energy data to generate energy trend analysis;

The anomaly detection module is used for carrying out anomaly mode identification on energy data by adopting an isolated forest and neural network model based on energy trend analysis, and generating an anomaly mode detection report;

The relation mining module is used for mining abnormal data relation of the energy data by adopting cluster analysis, association rule mining and network analysis methods based on the abnormal mode detection report to generate an energy data relation graph;

the transaction strategy optimization module is used for optimizing the energy transaction strategy by adopting a deep neural network and a reinforcement learning algorithm based on the energy data relation diagram, and generating an optimized transaction strategy;

the energy consumption prediction module is used for performing energy demand prediction by adopting a graph neural network to analyze historical and real-time data based on the post-optimization transaction strategy, so as to generate an energy demand prediction report;

The federal learning framework module performs decentralized data processing and model synchronous updating by using a federal learning technology based on an energy demand prediction report, protects data privacy by using encryption and data anonymization technology, and generates a federal learning enhancement model;

The system integration and supervision module is based on a federal learning enhancement model, adopts an association rule learning method and a support vector machine algorithm to conduct data association analysis and data classification optimization, and conducts policy decision analysis to generate a system supervision report.

As a further aspect of the present invention, the verified energy data is specifically a data non-tamper-resistant and encrypted data set, the energy trend analysis includes energy periodicity, trend and future prediction, the abnormal pattern detection report is specifically energy data which does not conform to a conventional pattern, the energy data relationship graph includes energy data correlations and potential links, the post-optimization transaction policy is specifically an automatic adjustment, market energy transaction scheme, the energy demand prediction report includes prediction and consumption pattern analysis for target user groups and regional energy demands, the federal learning enhancement model is specifically node sharing learning results to ensure data privacy and security, and the system supervision report is specifically a whole energy transaction and supervision system performance evaluation and monitoring result.

As a further scheme of the invention, the data integrity verification module comprises a data encryption sub-module, a consensus machine sub-module and a data uploading sub-module;

The data encryption submodule encrypts the blockchain data by adopting a secure hash algorithm based on a blockchain technology to generate encrypted data;

The consensus machine sub-module verifies the blockchain network data by adopting a workload proof consensus mechanism based on the encrypted data to generate consensus verified data;

and the data uploading submodule carries out data signature by adopting an elliptic curve digital signature algorithm based on the data after the common recognition verification and uploads the data to the blockchain network to generate the energy data after the verification.

As a further scheme of the invention, the time sequence analysis module comprises a trend analysis sub-module, a periodicity analysis sub-module and a prediction modeling module;

The trend analysis submodule carries out trend analysis by adopting an autoregressive integral moving average algorithm based on the verified energy data, carries out data trend analysis and generates a primary trend analysis result;

the periodic analysis submodule performs periodic analysis on the energy data by adopting a Fourier transform method based on the preliminary trend analysis result, performs data periodic fluctuation mode identification, and generates a periodic analysis result;

The prediction modeling module is used for performing energy data prediction modeling by adopting a time sequence prediction method based on a periodic analysis result, and performing trend prediction analysis to generate energy trend analysis.

As a further scheme of the invention, the abnormality detection module comprises an isolated forest submodule, a first neural network submodule and a pattern recognition submodule;

the isolated forest submodule performs preliminary abnormal mode identification by adopting an isolated forest algorithm based on energy trend analysis, performs abnormal data point screening and generates a preliminary abnormal mode report;

The first neural network sub-module adopts a multi-layer perceptron neural network to conduct deepened abnormal mode analysis based on the preliminary abnormal mode report, and generates a deep abnormal mode analysis report;

The pattern recognition submodule carries out abnormal pattern recognition and classification by adopting a pattern recognition technology based on the depth abnormal pattern analysis report to generate an abnormal pattern detection report.

As a further scheme of the invention, the relation mining module comprises a cluster analysis sub-module, an association rule sub-module and a network analysis sub-module;

The clustering analysis sub-module carries out energy data classification by adopting a K-means clustering algorithm based on the abnormal mode detection report to generate a clustering analysis result;

The association rule submodule adopts an Apriori association rule mining algorithm to search potential data relations based on the clustering analysis result, and generates an association rule mining result;

The network analysis submodule adopts a social network analysis method based on the association rule mining result to generate an energy data relation graph by exploring complex relations among data.

As a further scheme of the invention, the transaction strategy optimization module comprises a strategy control sub-module, a second neural network sub-module and a reinforcement learning sub-module;

the strategy making submodule carries out energy trading strategy design based on the energy data relation diagram to generate a preliminary trading strategy;

The second neural network sub-module adopts a convolutional neural network to perform policy structure optimization based on the preliminary transaction policy, and performs policy performance enhancement to generate a neural network optimization policy;

and the reinforcement learning submodule optimizes the transaction strategy by adopting a reinforcement learning algorithm based on the neural network optimization strategy to generate an optimized transaction strategy.

As a further scheme of the invention, the energy consumption prediction module comprises a graph neural network sub-module, a historical data analysis sub-module and a real-time data processing sub-module;

the graph neural network submodule adopts a graph convolution network to conduct historical and real-time data analysis based on the optimized transaction strategy, and conducts energy data correlation analysis to generate a data correlation analysis result;

the historical data analysis submodule adopts a time sequence analysis method to carry out historical data trend mining based on the data relevance analysis result to generate a historical data analysis result;

and the real-time data processing sub-module predicts the energy demand by adopting a real-time data stream processing technology based on the historical data analysis result, and generates an energy demand prediction report.

As a further scheme of the invention, the federal learning framework module comprises a decentralized data processing sub-module, a model synchronous updating sub-module and a privacy protection sub-module;

The distributed data processing submodule carries out data fusion by adopting a federal learning distributed processing technology based on the energy demand prediction report to generate a distributed data processing result;

The model synchronous updating sub-module synchronously updates the model by adopting a model aggregation algorithm based on the scattered data processing result to generate a model synchronous updating result;

And the privacy protection submodule carries out data privacy protection by adopting a homomorphic encryption method and a data anonymization strategy based on a model synchronous updating result to generate a federal learning enhancement model.

As a further scheme of the invention, the system integration and supervision module comprises a monitoring sub-module, an evaluation sub-module and a report generation sub-module;

The monitoring submodule carries out system performance tracking by adopting a data monitoring method based on the federal learning enhancement model to generate a monitoring analysis result;

The evaluation submodule carries out data classification and relevance analysis by adopting a support vector machine algorithm based on the monitoring analysis result to generate a data classification optimization result;

And the report generation submodule carries out system performance analysis by adopting an association rule learning method based on the data classification optimization result, and carries out comprehensive evaluation strategy decision to generate a system supervision report.

Compared with the prior art, the invention respectively explains the beneficial effects brought by each module according to the module of the right 1:

According to the invention, through the combination of the autoregressive integral moving average and the time sequence prediction method, the energy trend analysis is more accurate, the comprehensive view angles of periodicity, trend and future prediction are provided, the use of isolated forests and neural network models is realized, the abnormal mode is effectively identified, and the detection capability of abnormal energy data is improved. The cluster analysis, association rule mining and network analysis method of the relation mining module helps to reveal complex relations and potential relations between energy data, and provides deeper insight for decision making. In the aspect of transaction strategy optimization, the application of the deep neural network and the reinforcement learning algorithm enables the transaction strategy to be more flexible and adaptive, and improves the efficiency of market energy transaction. The energy consumption prediction module predicts future demands and analyzes consumption modes, so that the demands of target user groups and regions can be met more accurately. The federal learning framework module is introduced, so that the privacy and safety of data are ensured and the performance of the model is improved through the mode of sharing learning results by nodes. The system integration and supervision module ensures the high-efficiency operation and reliability of the whole system through comprehensive evaluation and monitoring.

Drawings

FIG. 1 is a system flow diagram of the present invention;

FIG. 2 is a schematic diagram of a system framework of the present invention;

FIG. 3 is a flow chart of a data integrity verification module of the present invention;

FIG. 4 is a flow chart of a time series analysis module according to the present invention;

FIG. 5 is a flow chart of an anomaly detection module of the present invention;

FIG. 6 is a flow chart of a relationship mining module of the present invention;

FIG. 7 is a flow chart of a transaction policy optimization module of the present invention;

FIG. 8 is a flow chart of an energy consumption prediction module of the present invention;

FIG. 9 is a flow chart of a federal learning framework module of the present invention;

FIG. 10 is a flow chart of the system integration and supervision module according to the present invention.

Detailed Description

The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.

In the description of the present invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," and the like indicate orientations or positional relationships based on the orientation or positional relationships shown in the drawings, merely to facilitate describing the present invention and simplify the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be configured and operated in a specific orientation, and therefore should not be construed as limiting the present invention. Furthermore, in the description of the present invention, the meaning of "a plurality" is two or more, unless explicitly defined otherwise.

Example 1

Referring to fig. 1-2, the blockchain-based energy trading and supervision system includes a data integrity verification module, a time sequence analysis module, an anomaly detection module, a relationship mining module, a trading strategy optimization module, an energy consumption prediction module, a federal learning framework module, and a system integration and supervision module;

the data integrity verification module is based on a blockchain technology and is used for guaranteeing data safety transmission through a hash algorithm and a digital signature, the data safety transmission relates to transaction records from an energy transaction platform, the data is kept consistent by utilizing a consensus mechanism, and verified energy data is generated;

the transaction strategy optimization module optimizes the energy transaction strategy by adopting a deep neural network and a reinforcement learning algorithm based on the energy data relation diagram to generate an optimized transaction strategy;

The energy consumption prediction module is used for performing energy demand prediction by adopting a graph neural network to analyze historical and real-time data based on the optimized transaction strategy, so as to generate an energy demand prediction report;

the federal learning framework module performs decentralized data processing and model synchronous updating by using federal learning technology based on the energy demand prediction report, protects data privacy by using encryption and data anonymization technology, and generates a federal learning enhancement model;

The energy data after verification is a data non-tamper-resistant and encrypted data set, the energy trend analysis comprises energy periodicity, trend and future prediction, the abnormal mode detection report comprises energy data which does not accord with a conventional mode, the energy data relation diagram comprises energy data interrelationships and potential relations, the optimized transaction strategy comprises automatic adjustment and a market energy transaction scheme, the energy demand prediction report comprises data privacy and safety aiming at a target user group, regional energy demand prediction and consumption mode analysis, the federal learning enhancement model comprises node sharing learning results, and the system supervision report comprises whole energy transaction and supervision system performance assessment and monitoring results.

In the data integrity verification module, secure transmission of data is ensured by application of a blockchain technology. Specifically, the module receives transaction records from the energy transaction platform, the records are firstly processed by a hash algorithm, transaction information is converted into hash values with fixed length, the hash values are unique, any tiny data change can cause significant changes of the hash values, so that the non-falsification of the data is ensured, a digital signature technology is applied to the hash values, and the hash values are encrypted by using a private key to generate signatures. The signature is transmitted to the recipient along with the transaction record. The receiver uses the public key to decrypt the signature, verifies the authenticity and integrity of the data source, and the consensus mechanism plays a role in the module in maintaining the consistency of all node data in the network. The consensus mechanism adopted is selected according to the practical application environment, such as the Proof of Work (Proof of Work) or the Proof of stock (Proof of stare) and the like. The nodes synchronize data through a consensus algorithm, so that the data records on each node are consistent, trusted verification of transaction records is formed in the whole network, the module finally generates verified energy data, and the data provides accurate and reliable data bases for subsequent modules while ensuring the safety.

In the time series analysis module, based on the verified energy data, an autoregressive integral moving average (ARIMA) model and a time series prediction method are applied to carry out energy periodicity and trend analysis. The ARIMA model integrates three methods, autoregressive (AR), differential (I), and Moving Average (MA). The autoregressive part reflects the dependency relationship between variables, the differential part is used for converting the non-stationary time series into stationary series, the moving average part is used for smoothing random fluctuation in data, the module firstly performs proper preprocessing on the energy data, such as removing abnormal values and trend decomposition, and then determines parameters (p, d, q) of the ARIMA model, which represent the quantity of autoregressive terms, differential orders and moving average terms respectively. By applying these parameters to the historical data, the model is able to identify the periodicity and trend of energy usage, and the module outputs the energy trend analysis results, providing important information about future energy usage trends and periodicity changes for the energy market.

In the anomaly detection module, based on the analysis result of the energy trend, an isolated forest algorithm and a neural network model are adopted to identify the anomaly mode of the energy data, wherein the isolated forest is a tree-based algorithm, a plurality of isolated trees are constructed by randomly selecting data points and randomly selecting segmentation values, the normal data points can be isolated only by more steps due to the similar mode, and the anomaly points are more easily isolated. The algorithm is efficient in that it does not require any distance or density calculations and is suitable for processing large-scale data. Meanwhile, an abnormality detection model based on a neural network is also applied to the module, particularly the neural network is excellent in processing complex modes and high-dimensional data, the neural network trains a normal mode in learning data, marks data points which are obviously different from the learning mode as abnormal when the data points are identified, and finally generates an abnormality mode detection report to specify the detected abnormality mode and the reason thereof, so that important auxiliary information is provided for energy management and decision.

In the relation mining module, based on the abnormal mode detection report, clustering analysis, association rule mining and a network analysis method are applied to perform abnormal data relation mining of the energy data. Cluster analysis helps identify natural patterns or populations in the data by grouping similar data points together, revealing associations between outlier data. Association rule mining is then used to find interesting relationships between data items, such as certain abnormal patterns often appear with specific energy usage behaviors. The network analysis reveals the structural characteristics and the relation strength in the data by constructing a connection network between data points, and the module can deeply analyze the potential relation and interaction between abnormal data to generate an energy data relation graph, and the graph not only shows the relation between abnormal modes, but also reveals the potential reasons for causing the abnormality, thereby providing an important basis for improving an energy system.

In the transaction strategy optimization module, based on the energy data relation diagram, the deep neural network and the reinforcement learning algorithm are adopted to conduct energy transaction strategy optimization, and the deep neural network can process a large amount of complex data and learn deep features and modes from the deep neural network. In this module, the neural network identifies trends and potential opportunities that are beneficial to the energy trading by analyzing patterns in the energy data relationship graph, the reinforcement learning algorithm learns optimal strategies through interactions with the environment, the algorithm continuously tries different trading strategies, and receives rewards or penalties according to the effects of the strategies, gradually learns how to make optimal decisions under various market conditions, and the module finally outputs optimized trading strategies that can help the energy trading participants make more effective and profitable trading decisions in the complex and variable market environment.

In the energy consumption prediction module, historical and real-time data are analyzed through a Graph Neural Network (GNN) algorithm, and energy demand prediction is performed, the data format processed by the module is time sequence data, and the data comprises historical energy consumption, time stamps and related environmental factor data, and the key point of using the graph neural network is that the graph neural network can process irregular data structures, particularly when predicting a mode related to an energy network topological structure. In the specific implementation process, a graph structure is firstly constructed according to physical connection of an energy network, then nodes are utilized to represent energy consumption points, edges represent connection relations, and a model can learn time dependence and space dependence of energy consumption by applying a convolutional neural network to the graph. The graph neural network captures complex consumption modes by iteratively updating node states and integrating information of adjacent nodes, and after training, the model can predict energy requirements in a specific time period in the future. The generated energy demand prediction report shows the predicted energy consumption of each time point in the future in detail, which is very critical to energy management and optimization, is beneficial to adjusting energy distribution and optimizing energy use efficiency.

In the federal learning framework module, decentralized data is processed by federal learning techniques and the model is updated synchronously, the data format processed by the module comprising energy consumption data for a plurality of data sources, each data source format being different. The core of federal learning is to have multiple data sources contribute together to the training of a shared model while maintaining data privacy, in practice, each data source trains the model on local data first, and then sends model updates to a central server. The central server aggregates the updates to form a global model, which is then distributed back to the various data sources, and the process repeats until the model converges. To protect data privacy, encryption techniques and data anonymization techniques are applied to ensure data security during transmission. By the method, the generated federal learning enhancement model can improve prediction accuracy under the cooperation of different data sources, and meanwhile, the privacy of each data source is protected.

In the system integration and supervision module, data association analysis and data classification optimization are performed through an association rule learning method and a Support Vector Machine (SVM) algorithm. The data formats processed by this module include energy consumption data and predictive data, as well as other data related to energy management, association rules learning for identifying interesting patterns and associations in the data, e.g., increases or decreases in energy consumption over a particular period of time are associated with a particular event. By analyzing these associations, one can get insight into the reasons behind the energy consumption behavior, support vector machines are used for data classification, and data points are classified into different categories, e.g., normal consumption mode and abnormal consumption mode, by constructing a classification model. The SVM maximizes the interval between different categories by searching the optimal hyperplane, ensures the accuracy of classification, and the final output of the system integration and supervision module is a system supervision report which summarizes the key modes and trends of energy consumption and also provides data support for energy policy formulation and policy decision making, thereby realizing more efficient and sustainable energy management.

Referring to fig. 3, the data integrity verification module includes a data encryption sub-module, a consensus sub-module, and a data uploading sub-module;

the data uploading submodule carries out data signature by adopting an elliptic curve digital signature algorithm based on the data after the consensus verification and uploads the data to the blockchain network to generate the energy data after the verification.

In the data encryption sub-module, the source data mainly originate from various application and transaction scenes related to energy, for example, energy production data come from capacity records of wind power, solar energy, hydraulic power or traditional coal power stations, the energy consumption data record electricity consumption conditions of consumers such as factories, commercial buildings, residential areas and the like, the energy transaction records cover transaction details such as transaction time, quantity and price of buyers and sellers in an energy market, the intelligent power grid operation data comprise energy distribution and load balance information generated by sensors and intelligent meters, and the maintenance and detection data record maintenance conditions and equipment operation states of energy facilities. The source data exist in the forms of text, numbers and the like before encryption, are converted into unique hash values after encryption processing by a secure hash algorithm (such as SHA-256), so that the security and the integrity of the data in the transmission and storage processes are ensured, the data security of the encrypted data is ensured while the original information is maintained, and particularly, the key effect of ensuring the non-tamper property and the transparency of the transaction data is exerted when the blockchain technology is applied to the energy transaction records.

In the consensus machine sub-module, based on the encrypted data, a working quantity Proof (PoW) consensus mechanism is used for verifying the data in the blockchain network, a plurality of nodes in the network participate in calculation and verification, the consistency of the data and the safety of the network are ensured, the encrypted data can be received and recorded in the blockchain only after the verification of most nodes is passed, and the step is a key link of the blockchain technology for ensuring the non-falsification and transparency of the data.

In the data uploading sub-module, based on the data after the consensus verification, an Elliptic Curve Digital Signature Algorithm (ECDSA) is used for signing the data and uploading the data to a blockchain network, the digital signature of the data not only ensures the authenticity and the credibility of the data source, but also enhances the safety of the data, and the data after the signature is finished, namely the data which is regarded as the verified energy data, is stored in the blockchain for subsequent analysis and use.

Referring to fig. 4, the time sequence analysis module includes a trend analysis sub-module, a periodicity analysis sub-module, and a prediction modeling sub-module;

The periodic analysis submodule performs periodic analysis on the energy data by adopting a Fourier transform method based on the primary trend analysis result, performs data periodic fluctuation mode identification, and generates a periodic analysis result;

The prediction modeling submodule performs energy data prediction modeling by adopting a time sequence prediction method based on a periodic analysis result, performs trend prediction analysis and generates energy trend analysis.

In the trend analysis submodule, an autoregressive integral moving average (ARIMA) algorithm is applied to the verified energy data to conduct trend analysis, and the focus is on exploring the mode and rule of time-varying aspects of energy production, consumption, price and the like. For example, analysis of seasonal fluctuations in wind or solar energy production in a region reveals long-term trends in the production of non-renewable energy sources such as coal and natural gas, or explores the effects of winter home heating and industrial production peaks on energy demand, and also includes in-depth studies on energy market price changes, such as analysis of how crude oil prices are affected by supply-demand relationships, policy variations, by which key trends can be identified from historical data, providing predictions for future energy production, consumption and pricing, which provide valuable information and guidance for long-term strategic planning of energy enterprises, government energy policy formulation, and even for investors' decisions.

In the periodicity analysis sub-module, based on the primary trend analysis result, the periodicity of the energy data is analyzed by using a fourier transform method, and the importance is that the periodicity fluctuation mode in the energy data, including seasonal fluctuation or repeated occurrence modes, is identified and understood, and the periodicity analysis is helpful for better understanding the fluctuation rule of the data and provides key input for predictive modeling.

In the prediction modeling module, based on a periodical analysis result, a time sequence prediction method is adopted to conduct prediction modeling on energy data, future trends and modes are involved, energy trend analysis is aimed at being generated, prediction on future energy demands and supplies is included, and trend prediction of prices, consumption and other relevant factors is covered.

Referring to fig. 5, the abnormality detection module includes an isolated forest submodule, a first neural network submodule, and a pattern recognition submodule;

The isolated forest submodule performs preliminary abnormal pattern recognition by adopting an isolated forest algorithm based on energy trend analysis, performs abnormal data point screening, and generates a preliminary abnormal pattern report.

The first neural network sub-module adopts a multi-layer perceptron neural network to conduct deepened abnormal pattern analysis based on the preliminary abnormal pattern report, and generates a deep abnormal pattern analysis report.

The pattern recognition submodule carries out abnormal pattern recognition and classification by adopting a pattern recognition technology based on the deep abnormal pattern analysis report to generate an abnormal pattern detection report.

In the isolated forest sub-module, an isolated forest algorithm is applied to conduct abnormal pattern recognition on the energy trend analysis result, and the step is focused on screening abnormal or outliers from the data to recognize the data which do not accord with the conventional pattern. For example, in terms of energy consumption data, the anomaly data is manifested as high energy consumption of the industrial area at non-working times or low energy consumption of the residential area at usual days, which implies equipment failure or improper energy use, in energy production data, such as unexpected low energy output of the solar power plant in sunny weather, is directed to equipment damage or maintenance requirements. For energy trading markets, abnormal trading data, such as the amount of trades or prices that deviate suddenly from historical trends, reflect market manipulation or information leakage. The isolated forest algorithm isolates the abnormal points by randomly selecting data features and score values, so that abnormal or irregular modes in the data are effectively revealed, a preliminary abnormal mode report is generated, the identification of abnormal data is important for establishing an early warning system, ensuring effective operation of energy equipment and supervising energy market transactions, and the method helps to discover and solve potential problems in time.

In the first neural network sub-module, based on the preliminary abnormal mode report, the abnormal mode is deeply analyzed by using the multi-layer perceptron neural network, the abnormal mode can be deeply understood by the application of the neural network, the accuracy and depth of identifying the abnormal mode are improved by simulating the processing mode of the human brain, and the depth abnormal mode analysis report generated in the process provides detailed basic information for final mode identification and classification.

In the pattern recognition sub-module, based on the deep abnormal pattern analysis report, the abnormal patterns are carefully identified and classified by adopting a pattern recognition technology, and the key of the step is to accurately judge the specific category and characteristic of abnormal data, which is important for understanding the cause of the occurrence of the abnormal and making corresponding countermeasures, and the finally generated abnormal pattern detection report provides comprehensive abnormal data analysis and facilitates the subsequent data processing and decision making.

Referring to fig. 6, the relation mining module includes a cluster analysis sub-module, an association rule sub-module, and a network analysis sub-module;

The association rule sub-module adopts an Apriori association rule mining algorithm to search potential data relations based on the clustering analysis result, and generates an association rule mining result;

In the clustering analysis sub-module, the energy data based on the abnormal mode detection report is classified by applying a K-means clustering algorithm, and the important point in the process is that a plurality of energy data points are grouped into different categories according to the similarity of data characteristics. For example, the energy consumption data may be classified into low, medium, high consumption categories or categories such as night peak, daytime peak, uniform consumption, etc. according to the electricity consumption, consumption frequency or time, the energy production data may be classified into high-efficiency, medium and low-efficiency production categories according to the production efficiency or the equipment status, or classified into categories such as normal operation, maintenance and failure according to the equipment operation status, etc., the energy transaction data may be classified into categories such as high fluctuation, medium fluctuation and low fluctuation according to the amplitude and frequency of price fluctuation, etc., and the energy data may be also grouped according to the geographical location characteristics to identify the energy usage and demand patterns of different areas. The clustering method not only helps to identify natural modes and classifications in the energy data and provides a basis for deep relation analysis, but also provides a macroscopic view of the data for the generated clustering analysis result, reveals a potential data distribution mode and provides important information support for further data analysis and strategy formulation in the energy industry.

In the association rule sub-module, based on the clustering analysis result, an Apriori association rule mining algorithm is used for exploring potential relations among data, the key point of the step is to identify association rules among data features, such as specific consumption behaviors caused by certain energy use modes, an association rule mining result is generated, hidden relations and dependencies among the data features are revealed, and valuable insights are provided for decision making.

In the network analysis submodule, based on the association rule mining result, a social network analysis method is adopted to deeply explore complex relations among data, the relation among the data is analyzed and visualized by utilizing a network theory, the relation among the data comprises connection strength and modes among data points, an energy data relation graph can be generated, visual understanding of interaction and influence among the data is provided, and identification of key influence factors and potential risk points is assisted.

Referring to fig. 7, the transaction policy optimization module includes a policy making sub-module, a second neural network sub-module, and a reinforcement learning sub-module;

the second neural network sub-module adopts a convolutional neural network to perform strategy structure optimization based on the preliminary transaction strategy, and performs strategy performance enhancement to generate a neural network optimization strategy;

The reinforcement learning submodule optimizes the transaction strategy by adopting a reinforcement learning algorithm based on the neural network optimization strategy to generate an optimized transaction strategy.

In the policy making sub-module, a preliminary energy trading policy is designed by based on the energy data relationship diagram. The key of the process is to form a preliminary strategy framework by using deep data insight provided by the relation diagram, ensure that the strategy can reflect the current dynamic and potential trend of the energy market, and the method can generate a preliminary transaction strategy and provide a basis for further optimization.

In the second neural network sub-module, based on the preliminary transaction strategy, a Convolutional Neural Network (CNN) is used for optimizing a strategy structure, the performance of the strategy is enhanced, the neural network plays a key role in the step, the strategy structure is optimized to improve the accuracy and efficiency of decision making by simulating complex data relations and modes, and the neural network optimization strategy generated in the process is more suitable for the change and the demand of the market.

In the reinforcement learning sub-module, the transaction strategy is optimized by adopting a reinforcement learning algorithm based on the neural network optimization strategy, the self-adaptive learning capacity of reinforcement learning is utilized, the strategy is adjusted through continuous experiments and errors, the strategy is more refined and effective, and finally the generated optimized transaction strategy has high adaptability and optimization performance, and can realize effective transaction decision in a changeable energy market.

Referring to fig. 8, the energy consumption prediction module includes a neural network sub-module, a historical data analysis sub-module, and a real-time data processing sub-module;

The graph neural network submodule adopts a graph convolution network to conduct historical and real-time data analysis based on the optimized transaction strategy, conduct energy data correlation analysis and generate a data correlation analysis result;

In the graph neural network sub-module, a graph rolling network (GCN) is applied based on an optimized transaction strategy to analyze historical and real-time data, and the key point of the step is to analyze complex relevance of energy data by utilizing the strong capability of the graph neural network, including interaction and influence among different energy nodes, so as to generate a data relevance analysis result, and a foundation for deeply understanding data relations is provided for energy consumption prediction.

In the historical data analysis submodule, based on the data relevance analysis result, a time sequence analysis method is used for mining trends and modes in the historical data, the important point of the step is to identify important trends and periodic modes in the historical data, including seasonal changes and long-term consumption trends, the historical data analysis result can be generated, and an important reference basis is provided for real-time prediction.

In the real-time data processing sub-module, based on the analysis result of the historical data, the current and short-term energy demands are predicted by adopting a real-time data stream processing technology, and the core of the step is to rapidly and effectively predict the short-term energy demand change by combining the historical trend and the current data so as to generate an energy demand prediction report.

Referring to fig. 9, the federal learning framework module includes a decentralized data processing sub-module, a model synchronization update sub-module, and a privacy protection sub-module;

The privacy protection sub-module performs data privacy protection by adopting a homomorphic encryption method and a data anonymization strategy based on the model synchronous updating result to generate a federal learning enhancement model.

In the distributed data processing sub-module, based on the energy demand prediction report, the data fusion is carried out by adopting the distributed processing technology of federal learning, the core of the step is to effectively integrate the data distributed on the nodes by utilizing the federal learning capacity without intensively storing or transmitting the data, so that the data processing efficiency is improved, the risk of data leakage is reduced, the distributed data processing result is generated, and the foundation is provided for the subsequent model updating and learning.

In the model synchronous updating sub-module, based on a scattered data processing result, a model aggregation algorithm is used for synchronously updating the model, local model updating on the nodes is summarized and aggregated to form a global model, the method ensures that the model can keep the consistency and accuracy of the whole while continuously learning and adapting to new data, and the model synchronous updating result generated in the process provides strong support for realizing collaborative learning across the nodes.

In the privacy protection submodule, based on a model synchronous updating result, the data is subjected to privacy protection by adopting a homomorphic encryption method and a data anonymization strategy, the homomorphic encryption allows the encrypted data to be calculated without decryption, so that the safety of the data is ensured, the privacy of personal information is not revealed by the data anonymization strategy, and the federal learning enhancement model generated by the method is improved in performance and also remarkably enhanced in the aspect of protecting the privacy of users.

Referring to fig. 10, the system integration and supervision module includes a monitoring sub-module, an evaluation sub-module, and a report generation sub-module;

the evaluation sub-module carries out data classification and relevance analysis by adopting a support vector machine algorithm based on the monitoring analysis result to generate a data classification optimization result;

And the report generation sub-module is used for carrying out system performance analysis by adopting an association rule learning method based on the data classification optimization result, carrying out comprehensive evaluation strategy decision and generating a system supervision report.

In the monitoring sub-module, based on the federal learning enhancement model, continuous tracking of system performance is performed by adopting a data monitoring method, and the key of the step is to monitor the running state of the system in real time, including key performance indexes of data flow, processing speed and accuracy, so as to generate a monitoring analysis result and provide real-time data support for stable operation and efficiency of the system.

In the evaluation sub-module, based on the monitoring analysis result, the data is classified and associativity analyzed by using a Support Vector Machine (SVM) algorithm, the application of the SVM algorithm can accurately identify and classify various types of data, meanwhile, the potential relation among the data is analyzed, and a data classification optimization result is generated, so that an important analysis basis is provided for subsequent system performance evaluation and strategy decision.

In the report generation sub-module, based on the data classification optimization result, the system performance is comprehensively analyzed by adopting an association rule learning method, and comprehensive evaluation strategy decision is carried out by combining the overall operation condition of the system, the core of the step is to combine the analysis result and the business target, evaluate the overall efficiency and effect of the system, generate a system supervision report, provide comprehensive evaluation of the system performance and future strategy scheme, and ensure continuous optimization and effective supervision of the system.

The present invention is not limited to the above embodiments, and any equivalent embodiments which are changed or modified to equivalent changes by those skilled in the art can be applied to other fields by using the technical contents disclosed above, but any simple modification, equivalent changes and modification made to the above embodiments according to the technical matters of the present invention are still within the scope of the technical solutions of the present invention.

Claims

1. Energy transaction and supervisory systems based on block chain, its characterized in that: the system comprises a data integrity verification module, a time sequence analysis module, an abnormality detection module, a relation mining module, a transaction strategy optimization module, an energy consumption prediction module, a federal learning framework module and a system integration and supervision module;

2. The blockchain-based energy trading and supervision system of claim 1, wherein: the energy data after verification is specifically a data non-tamper-resistant and encrypted data set, the energy trend analysis comprises energy periodicity, trend and future prediction, the abnormal mode detection report specifically refers to energy data which does not accord with a conventional mode, the energy data relation diagram comprises energy data interrelationships and potential relations, the optimized transaction strategy specifically comprises automatic adjustment and market energy transaction schemes, the energy demand prediction report comprises energy demand prediction and consumption mode analysis aiming at a target user group and a region, the federal learning enhancement model specifically comprises node sharing learning achievement guaranteeing data privacy and safety, and the system supervision report specifically comprises whole energy transaction and supervision system performance evaluation and monitoring results.

3. The blockchain-based energy trading and supervision system of claim 1, wherein: the data integrity verification module comprises a data encryption sub-module, a consensus machine sub-module and a data uploading sub-module;

4. The blockchain-based energy trading and supervision system of claim 1, wherein: the time sequence analysis module comprises a trend analysis sub-module, a periodicity analysis sub-module and a prediction modeling sub-module;

5. The blockchain-based energy trading and supervision system of claim 1, wherein: the abnormality detection module comprises an isolated forest submodule, a first neural network submodule and a mode identification submodule;

6. The blockchain-based energy trading and supervision system of claim 1, wherein: the relation mining module comprises a cluster analysis sub-module, an association rule sub-module and a network analysis sub-module;

7. The blockchain-based energy trading and supervision system of claim 1, wherein: the transaction strategy optimization module comprises a strategy system submodule, a second neural network submodule and a reinforcement learning submodule;

8. The blockchain-based energy trading and supervision system of claim 1, wherein: the energy consumption prediction module comprises a graph neural network sub-module, a historical data analysis sub-module and a real-time data processing sub-module;

9. The blockchain-based energy trading and supervision system of claim 1, wherein: the federal learning framework module comprises a scattered data processing sub-module, a model synchronous updating sub-module and a privacy protecting sub-module;

10. The blockchain-based energy trading and supervision system of claim 1, wherein: the system integration and supervision module comprises a monitoring sub-module, an evaluation sub-module and a report generation sub-module;