Open CIM space-time deep learning application platform based on value sharing

Abstract

The invention discloses an open type CIM space-time deep learning application platform based on value sharing, which is suitable for multiple fields of smart cities, such as city planning and traffic management. A dynamic learning and self-optimization platform is innovatively constructed by integrating multi-source heterogeneous data and utilizing a deep learning model and a block chain technology. The smart contract in the platform clearly divides ownership and earnings of data, models and algorithms, and ensures that all parties obtain fair earnings according to contributions thereof. According to the method, technologies of data acquisition and preprocessing, spatio-temporal data fusion, dynamic model training and the like are adopted, further through an open base and a value sharing mechanism, the limitation of an existing CIM system in the aspects of data fusion capability and collaborative innovation is solved, and the operation efficiency and decision support capability of a smart city are remarkably improved.

Description

Technical Field

[0001] The present application relates to the technical field of urban information model construction in smart cities, and in particular to an open CIM spatiotemporal deep learning application platform based on value sharing. Background Art

[0002] As an important foundation for the construction of smart cities, City Information Modeling (CIM) in smart cities can integrate and analyze multi-source heterogeneous data, and provide multi-dimensional and multi-scale information support for urban planning, construction, management and operation. However, most existing CIM systems are based on static data analysis, which makes it difficult to integrate real-time multi-source spatiotemporal data, especially when dealing with complex dynamic urban data such as traffic flow fluctuations, energy consumption, environmental monitoring, etc., lacking efficient data fusion and deep learning capabilities. In addition, most existing platforms are closed architectures, which restrict data sharing and multi-party collaboration, and are difficult to meet the needs of smart cities for dynamic optimization and collaborative innovation. Summary of the invention

[0003] In view of the deficiencies in the prior art, the present invention provides an open CIM spatiotemporal deep learning application platform based on value sharing, which solves the problems of the closedness and insufficient data fusion capability of the existing CIM platform and improves the management efficiency and decision-making support capability of smart cities.

[0004] The present invention is achieved through the following technical solutions: An open CIM spatiotemporal deep learning application platform based on value sharing, including: Data management module: used to store, manage and dispatch urban spatiotemporal data uploaded by different providers and data generated by the platform's own system; Model training module: Developers build, train and optimize deep learning models based on multi-source urban spatiotemporal data and application scenarios; Model application module: Apply the trained deep learning model to dynamic analysis, real-time prediction and scenario simulation of urban system application scenarios; Application interface module: provides standardized API interface and supports external system access; Value sharing and smart contract module: Smart contracts are used to realize the transaction, sharing and incentive mechanism of various urban spatiotemporal data, deep learning models and algorithms within the platform. Quantitative standards for contribution indicators are set for the data uploaded by data providers, the algorithms and models provided by developers, the operation and maintenance investment and service quality of platform operations. Smart contracts automatically calculate and distribute the benefits of contributors based on real-time contribution indicators; The platform adopts a consortium chain architecture and uses the Byzantine fault-tolerant algorithm as a consensus mechanism. Each node stores and verifies data through the consensus mechanism. The data management module transfers the acquired multi-dimensional spatiotemporal data to the model training module through a standard interface for training deep learning models. At the same time, the data management module records the use of data in real time and passes relevant data information to the value sharing and smart contract modules to ensure that data providers can obtain corresponding benefits based on their contributions. The model training module obtains multi-dimensional spatiotemporal data from the data management module and pre-processes the data to ensure data standardization and consistency. The pre-processed data is used to perform training and optimization of deep learning models. After training, the model training module uses the trained model in the model application module to perform dynamic analysis, real-time prediction or scenario simulation for different application scenarios. At the same time, the model training module feeds back the model performance data during the training process to the value sharing and smart contract module to calculate the contribution and profit distribution of the model developer. After completing the prediction task, the model application module returns the analysis results to the data management module for storage and for access and reference by other modules and external systems; The application interface module allows external systems to access data and models in the platform through a standardized API interface. External requests call the data of the data management module and the models of the model application module through the application interface module. Each call will be recorded in the smart contract and transmitted to the value sharing and smart contract module in real time to ensure that developers can obtain corresponding benefits according to their contributions. The value sharing and smart contract module receives usage data, model performance, API call information from the data management module, model training module, model application module and application interface module and processes them for automatic profit distribution.

[0005] Furthermore, the platform has built-in multiple deep learning models including convolutional neural networks, long short-term memory networks and graph neural networks, which use convolutional neural networks, long short-term memory networks and graph neural networks to extract spatiotemporal features of pre-processed data to capture the dynamic changes of various elements in the city: extract spatial features through convolutional neural networks to capture various geographical areas in the city and the relationship between them; use long short-term memory networks to model time series data to capture the time dependency and dynamic evolution in the data; use graph neural networks to process complex relational data and build a spatiotemporal feature model of the interaction and dependency between various elements in the urban system; The platform has a built-in adaptive learning rate optimization algorithm to dynamically adjust learning parameters to cope with changes in data streams. At the same time, cross-validation and early stopping techniques are used to prevent overfitting and ensure the generalization ability of the model. The model training and optimization process is accelerated with the help of a distributed computing framework, and parameter synchronization and model updates in distributed training are achieved through a parameter server architecture. Based on the extracted feature vectors and specific application scenarios, developers build a dynamic model framework based on deep learning to analyze and predict changes in urban systems; developers adjust and customize the platform's built-in algorithms, models, model training, and optimization technologies as needed.

[0006] Furthermore, the multidimensional spatiotemporal data includes one or more of remote sensing image data, urban sensor data, social media data, traffic monitoring data, meteorological data, environmental monitoring data, transportation network data, and power network data.

[0007] Furthermore, the preprocessing of multidimensional spatiotemporal data includes: Perform radiation correction and geometric correction on remote sensing image data; Use Kalman filters to clean and detect anomalies in urban sensor data; synchronize the data in time and correct the spatial coordinates to ensure the compatibility of different sensor data; Use natural language processing technology to perform sentiment analysis and geolocation extraction on social media data; use natural language processing technology to perform sentiment analysis, topic mining and geolocation extraction on social media texts, and use entity recognition technology to extract location, time and person information related to urban events; The density estimation algorithm is used to calculate the real-time flow distribution of traffic monitoring data; The extracted geographic location information is integrated with other data sources to enrich the urban spatiotemporal dataset.

[0008] Furthermore, the platform adopts a multi-layer encryption mechanism. First, it uses an asymmetric encryption algorithm to encrypt and exchange keys to ensure the security of data transmission. Secondly, it uses a symmetric encryption algorithm to encrypt large-scale data to ensure that data can remain efficient and secure during storage and transmission. In addition, the platform introduces zero-knowledge proof technology, allowing data providers to prove the authenticity and validity of data to the blockchain network without exposing the content of the data.

[0009] Furthermore, the data management module uses NoSQL database to store unstructured data and Elasticsearch to achieve fast data retrieval; it adopts distributed file system to achieve efficient storage and management of large-scale data, and combines in-memory computing technology to improve data access speed.

[0010] Furthermore, the model training module supports heterogeneous computing environments and enables rapid deployment and expansion of the model training environment through containerization technology.

[0011] Furthermore, the model application module feeds back the results to the model training module, triggering further optimization or retraining of the model to ensure the accuracy and timeliness of the model.

[0012] Furthermore, the data management module supports real-time updating of data and retrospective query of historical data.

[0013] Furthermore, the model training module supports both online training and offline training.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides an open CIM spatiotemporal deep learning application platform based on value sharing, which is suitable for smart city scenarios such as urban planning, traffic management, environmental monitoring and energy consumption optimization. By integrating multi-source heterogeneous data and using deep learning models and blockchain technology, an innovative dynamic learning and self-optimization platform is constructed to achieve accurate analysis and prediction of urban dynamic data. The smart contracts within the platform clearly divide the ownership and income rights of data, models and algorithms to ensure that all parties obtain fair benefits based on their contributions. The present invention adopts technologies such as data collection and preprocessing, spatiotemporal data fusion, and dynamic model training, and further solves the limitations of existing CIM systems in data fusion capabilities and collaborative innovation through an open base and value sharing mechanism, significantly improving the operational efficiency and decision-making support capabilities of smart cities, realizing open sharing of data and models, and promoting multi-party collaboration and innovation. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0016] Figure 1 A flowchart of the method for building an open CIM spatiotemporal deep learning platform based on value sharing; Figure 2 This is a schematic diagram of the architecture of an open CIM spatiotemporal deep learning platform implementation plan based on value sharing. DETAILED DESCRIPTION

[0017] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to be able to fully convey the scope of the present invention to those skilled in the art. It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0018] The purpose of the present invention is to provide an open CIM spatiotemporal deep learning application platform based on value sharing, which builds a CIM basic platform with dynamic learning, self-optimization and value sharing capabilities by integrating multi-source heterogeneous data, deep learning models and blockchain technology. The platform clearly divides the ownership, use rights and income rights of data, models, algorithms and other elements through smart contracts, and realizes deep learning modeling and multi-party collaboration of urban multidimensional data. For the contributions of different participants, such as data providers, model developers, and platform operators, dynamic quantitative standards based on dimensions such as data quality, usage frequency, and model performance are adopted to ensure that each contributor obtains corresponding benefits in real time according to his contribution. For example, data providers can obtain more rewards based on the update frequency and accuracy of data; model developers can obtain continuous benefits based on the accuracy, stability and performance of their models in real-time data processing; platform operators are responsible for building and maintaining the platform, ensuring its stable operation and data security, and coordinating cooperation among all parties. Its income comes from the commission of the income of data providers and model developers, and the commission ratio depends on the platform resource investment, operating costs, etc. This mechanism is automatically implemented through smart contracts on the blockchain, ensuring transparent, traceable and fair value distribution, and improving the operational efficiency and decision-making support capabilities of smart cities. Data providers, model developers and platform operators can obtain corresponding benefits through smart contracts based on their contributions, ensuring that participants can benefit fairly.

[0019] The main technical solutions of the present invention cover the following aspects: Data collection and preprocessing: Obtain multi-dimensional spatiotemporal data of the city from multiple data sources such as remote sensing images, sensor networks, social media, traffic monitoring, etc. Perform preprocessing such as format conversion, coordinate calibration, and time synchronization on different types of data to ensure data standardization and consistency.

[0020] Spatiotemporal data fusion and feature extraction: Use deep learning algorithms such as convolutional neural network (CNN), long short-term memory network (LSTM) and graph neural network (GNN) built into the platform to extract spatiotemporal features from multi-source heterogeneous data. CNN convolution operations are used to capture spatial features, LSTM is used to capture temporal features, and GNN is used to process complex relational data to achieve accurate modeling of urban dynamic changes.

[0021] Dynamic model training and optimization: Developers build a dynamic model framework based on deep learning according to the extracted spatiotemporal data features and specific application scenarios. The platform has a built-in adaptive learning rate optimization algorithm to dynamically adjust learning parameters when the data flow changes, improving the self-optimization ability of the model. Cross-validation and early stopping techniques are used to prevent overfitting, and distributed computing frameworks such as Hadoop or Spark are used for efficient training.

[0022] Construction of an open deep learning application platform: Based on open source platforms such as TensorFlow and PyTorch, the CIM spatiotemporal deep learning application platform is built, which includes a data management module, a model training module, a model application module, and an application interface module. The data management module supports data storage, query, and update; the model training module integrates a variety of deep learning algorithms, such as CNN, LSTM, and GNN, to cope with different types of data processing tasks; the model application module supports real-time and batch data analysis; the application interface module provides a standardized API to support the access and expansion of third-party developers.

[0023] Implementation of value contribution mechanism: Blockchain technology is used to ensure the transparency and traceability of data, models and algorithms, and further promote the implementation of value sharing. The platform adopts a consortium chain architecture, and realizes the transaction, sharing, distribution and profit distribution mechanism of data, models and algorithms (including deep learning models, traditional machine learning algorithms, etc.) through smart contracts. Each participant obtains corresponding rewards through the platform's incentive mechanism based on the value of the contribution, such as data quality, model accuracy, algorithm optimization effect, etc. Blockchain technology ensures that the interests of all participants are protected, and data circulation is transparent, secure and reliable. This mechanism can not only stimulate the opening and sharing of more data and models, but also encourage all parties to jointly improve the performance and accuracy of smart city-related applications.

[0024] Application scenarios and system integration: Based on this open platform, a variety of application scenarios are developed, including intelligent traffic dispatching, environmental monitoring and early warning, energy management optimization, etc. Through standardized interfaces, seamless integration with existing urban information systems is achieved, providing real-time data support and intelligent decision-making capabilities. As a technology supplier, the platform obtains fees paid by the application party or system integrator as overall benefits, and all participants within the platform automatically distribute benefits according to smart contracts and their respective contribution quantitative indicators.

[0025] In order to make the technical solutions and advantages of the embodiments clearer, the technical solutions of the embodiments of the present application will be fully described below in conjunction with the accompanying drawings. Figure 1 The following is a flowchart of building an open CIM spatiotemporal deep learning application platform based on value sharing provided by an embodiment of the present application. The specific process is as follows: S100: Data collection and preprocessing Obtain city-related multidimensional spatiotemporal data from a variety of data sources, including remote sensing images, urban sensor data, social media data, traffic monitoring data, meteorological data, and environmental monitoring data. Among these data, traffic network and power grid data belong to graph data types. GNN is particularly suitable for processing this type of data because it can efficiently capture the complex relationships and interactions between nodes. Specifically, traffic network data uses traffic facilities (such as intersections, road sections, etc.) as nodes, and road connections or traffic flows between nodes constitute edges. Power grid data uses power facilities (such as generators, transformers, etc.) as nodes, and power transmission relationships between devices constitute edges. GNN uses this graph structure to model and predict the dynamic changes of the system, such as traffic flow prediction and power grid load optimization. The processing method for each data type is as follows: Remote sensing image data processing: Use image preprocessing technology to perform radiation correction and geometric correction on remote sensing images. Use Fourier transform and wavelet transform methods to perform image enhancement and noise filtering, and use segmentation algorithms such as K-means clustering or region growing method to identify urban elements in the image, such as buildings, roads, and green spaces.

[0026] Urban sensor data processing: Collect data from sensor networks, such as air quality, temperature, humidity, and noise level. Use Kalman filters for data cleaning and anomaly detection to remove noise and unreliable data points. Perform time synchronization and spatial coordinate correction on the data to ensure compatibility of different sensor data.

[0027] Social media data processing: Use natural language processing technology to perform sentiment analysis, topic mining, and geographic location extraction on social media texts. Use entity recognition technology to extract location, time, and person information related to urban events. Fuse the extracted geographic location information with other data sources to enrich the urban spatiotemporal dataset.

[0028] Traffic monitoring data processing: Collect traffic flow data from traffic cameras, vehicle GPS, smart traffic lights, etc. Use density estimation algorithms such as kernel density estimation to calculate the real-time traffic flow distribution of roads.

[0029] S200: Spatiotemporal data fusion and feature extraction The fused data is extracted for spatiotemporal features through built-in deep learning models such as CNN, LSTM and GNN to capture the dynamic changes of various elements in the city. These models are mainly used for feature extraction rather than direct model training, aiming to convert raw data into feature representations suitable for subsequent training.

[0030] Spatial feature extraction: Use CNN to extract spatial features from remote sensing images and sensor data. The formula is as follows:

[0031] in, is a pixel in the output feature map, is the pixel of the input image, is the weight of the convolution kernel, is the size of the convolution kernel.

[0032] Temporal feature extraction: LSTM is used to model time series data such as traffic flow and environmental monitoring data to capture the temporal dependency in the data. The formula is as follows:

[0033] in, is the time step The hidden state of the model (i.e., the output of the model), is the input data of the current time step, Is the input data The weight matrix of is the weight matrix of the previous time step, is a bias term that helps adjust the output of the computation process, and σ is a sigmoid activation function that outputs a value between 0 and 1 to control the degree of information flow.

[0034] Complex relationship modeling: Use GNN to process complex relationship data such as transportation networks and power grids. Build graph structure ,in is a node set, is an edge set. GNN updates node features through the following formula:

[0035] in, Is a node In the The characteristics of the layer, Is a node The set of neighbor nodes of and Is a node and The degree, It is The weight matrix of the layer, Is a node In the Characteristics of the layer.

[0036] S300: Dynamic Model Training and Optimization According to the extracted feature vectors and specific application scenarios, developers build a dynamic model framework based on deep learning. This framework may include the use of deep learning models such as CNN, LSTM or GNN, depending on the type of input data. The framework can make adaptive adjustments based on changes in data streams, and continuously improve the accuracy and adaptability of the model through training and optimization.

[0037] Adaptive learning rate optimization: The model has built-in adaptive learning rate optimization algorithm function modules, such as the Adam optimizer, which dynamically adjusts learning parameters when the data stream changes. The formula is as follows:

[0038] in, and are momentum and acceleration, are the current model parameters, is the learning rate, is a constant that prevents the denominator from being zero, is the momentum decay rate constant, is the acceleration decay rate constant, is the loss function, which represents the current error of the model. is the gradient of the loss function with respect to the model parameters.

[0039] Model training and evaluation: Use cross-validation and early stopping techniques to prevent overfitting. During model training, calculate the loss function of the training set and the validation set, and stop training when the validation set loss no longer decreases.

[0040] Distributed training: Use distributed computing frameworks such as Apache Spark to efficiently process large-scale data and train models. Use the parameter server architecture to achieve parameter synchronization and model updates in distributed training.

[0041] S400: Construction of an open deep learning application platform The CIM spatiotemporal deep learning application platform is constructed based on open source platforms such as TensorFlow and PyTorch. The structure of the open CIM spatiotemporal deep learning application platform based on value sharing constructed in this embodiment is as follows: Figure 2 As shown, the platform includes a data management module, a model training module, a model application module, an application interface module, and a value sharing and smart contract module.

[0042] Data management module: The data management module is used to store, manage and schedule urban spatiotemporal data uploaded from different data sources. The data update interface is used to support real-time data updates, and the data management submodule is used for retrospective queries of historical data. NoSQL databases such as MongoDB are used to store unstructured data, and Elasticsearch is used to achieve fast data retrieval. Distributed file systems are used to achieve efficient storage and management of large-scale data, and in-memory computing technology is combined to improve data access speed.

[0043] Model training module: This module is used to train and optimize deep learning models based on multi-source data, and supports online and offline training. It integrates multiple deep learning frameworks including CNN, LSTM, and GNN, and supports the construction and training of different types of models. The module also supports heterogeneous computing environments such as CPU, GPU, and TPU to accelerate the model training process. In addition, through containerization technologies such as Docker and Kubernetes, this module enables rapid deployment and expansion of the model training environment. The data preprocessing submodule is used to preprocess urban spatiotemporal data, the deep learning training submodule is used to train and optimize the deep learning model framework, and the online / offline training switching submodule is used to switch the training mode.

[0044] Model application module: used to apply trained models to dynamically predict and analyze urban systems, supporting a variety of application scenarios. The model prediction submodule is used for real-time prediction, the dynamic analysis submodule is used for dynamic analysis, and the scenario simulation submodule is used for simulation of design scenarios. It combines big data technology to realize streaming data processing, supporting real-time analysis of high-frequency data and batch analysis of low-frequency data.

[0045] Application interface module: The standardized API interface submodule provides a standardized API interface to support access from external systems and developers. The application interface module provides a standardized RESTful API interface to support access from external systems, third-party developers, and partner organizations to promote value sharing. The developer tool submodule is used by developers to select various functions and algorithms built into the platform. This module not only allows external calls to data query, model training, prediction calls, and other functions, but also supports revenue distribution based on value contribution. Through smart contracts and API interfaces, external developers or data providers can obtain platform points or revenue based on their data, algorithm, and model contributions, maximizing incentives for open sharing and collaborative innovation of data and technical resources.

[0046] Value Sharing and Smart Contract Module: This module is the core module of the platform and is responsible for the value calculation, reward distribution and contract execution of the entire platform system. This module uses blockchain technology to ensure that all participants (data providers, model developers, application developers, etc.) receive fair economic incentives based on their contributions. Smart contracts automatically handle contribution evaluation, reward distribution and equity records to ensure transparency and security. Through interface integration with other modules, usage data and contribution indicators are collected to execute contract terms in an automated manner, including but not limited to data usage frequency, model effect evaluation and API call statistics.

[0047] The modules interact as follows: Interaction between the data management module and the model training module: The data management module is responsible for acquiring multi-dimensional spatiotemporal data from multi-source data collectors (such as urban sensor networks, remote sensing data, traffic monitoring, etc.). The multi-dimensional spatiotemporal data is passed to the model training module through a standard interface. The data is first pre-processed (such as format conversion, time synchronization, and coordinate calibration) for use in training deep learning models. At the same time, the data management module records the use of data in real time (such as data update frequency, accuracy, and completeness), and passes relevant data information to the value sharing and smart contract modules to ensure that data providers can obtain corresponding benefits based on their contributions.

[0048] Interaction between the model training module and the model application module: The model training module obtains the original multi-dimensional spatiotemporal data from the data management module, performs preprocessing, and then performs training and optimization of the deep learning model. After the training is completed, the model training module passes the trained model parameters to the model application module for subsequent analysis and prediction tasks. The model application module uses these trained models to perform dynamic analysis, real-time prediction, or scenario simulation on various scenarios in the urban system (such as traffic flow prediction and energy consumption optimization). At the same time, the model training module feeds back the model performance data (such as accuracy and convergence speed) during the training process to the value sharing and smart contract module for calculating the contribution and profit distribution of the model developer.

[0049] Interaction between the model application module and the data management module: After completing the prediction task, the model application module returns the analysis results to the data management module for storage and access and reference by other modules and systems. These analysis results can also be fed back to the model training module to trigger further optimization or retraining of the model to ensure the accuracy and timeliness of the model. Through this feedback mechanism, the system can continuously iterate and optimize key application scenarios in urban management.

[0050] Interaction between the application interface module and other modules: The application interface module allows external systems, third-party developers, and partners to access the data and models in the platform through a standardized API interface. External requests call the data of the data management module and the model of the model application module through the application interface module. Each call will be recorded in the smart contract and passed to the value sharing and smart contract module in real time to ensure that developers can obtain corresponding economic returns and rewards based on their contributions. In addition, external developers can expand system functions or develop new application scenarios through the API interface, and obtain data and models in the platform through interaction with other modules of the system.

[0051] Interaction between the value sharing and smart contract module and the platform system: The value sharing and smart contract module is the core coordinator of the entire platform system. It receives key data such as usage data, model performance, API call information, etc. from the data management module, model training module, model application module, and application interface module. These data are processed by the value sharing module and used to automatically execute revenue distribution. Through smart contracts, all participants (data providers, model developers, platform operators, etc.) can obtain corresponding benefits based on their contributions, and all transaction and distribution records are stored transparently and tamper-proof on the blockchain to ensure a fair, just and transparent value sharing mechanism.

[0052] S500: Implementation of value sharing mechanism The present invention combines blockchain technology with smart contracts to achieve transparency, traceability and security of data, models and algorithms, ensuring that all participants in the platform share benefits fairly based on their contributions. This mechanism encourages multiple parties to continuously optimize and innovate in data, algorithms, models, etc. through dynamic incentives and long-term benefit mechanisms, thereby achieving value sharing in smart cities.

[0053] Blockchain network architecture: The platform adopts a consortium chain architecture, and nodes are jointly maintained by trusted institutions such as urban management departments, scientific research institutions, and enterprises to ensure the security and distributed governance of the system. Each node stores and verifies data through a consensus mechanism to ensure the transparency and immutability of all transaction records. The system uses the Byzantine Fault Tolerance (PBFT) algorithm as a consensus mechanism to ensure that the system can still operate stably when some nodes in the network fail or behave maliciously.

[0054] Through the multi-party participation model of the alliance chain, the platform ensures the decentralized management and transparency of data governance. Each node participates in the storage and verification of data, which enhances the system's risk resistance. At the same time, the PBFT algorithm can improve the consensus efficiency of the blockchain and ensure the rapid recording and confirmation of transaction data.

[0055] Data encryption and privacy protection: To ensure the privacy and security of data, this platform adopts a multi-layer encryption mechanism. First, an asymmetric encryption algorithm is used to encrypt and exchange keys to ensure the security of data transmission. Secondly, a symmetric encryption algorithm is used to encrypt large-scale data to ensure that data can remain efficient and secure during storage and transmission.

[0056] In addition, the platform introduces zero-knowledge proof technology (ZKP), which allows data providers to prove the authenticity and validity of data to the blockchain network without exposing the data content. Through ZKP, data providers can participate in transactions and share revenue without worrying about data privacy leakage, ensuring transparency and credibility in the data transaction process.

[0057] Smart contract design: Smart contracts are the core tool for the platform to achieve value sharing, and can automatically execute the sharing, trading and profit distribution of data, models and algorithms. Smart contracts include the following functional modules: Data registration and authentication module: When the data provider uploads data or models, the contract will automatically register and authenticate the data to verify the authenticity and quality of its source. Ownership information will be permanently recorded on the blockchain to ensure traceability and immutability.

[0058] Revenue distribution module: Smart contracts automatically calculate the benefits of contributors and distribute them to data providers and model developers based on quantitative indicators set such as frequency of use, quality, and timeliness of data or models, ensuring the fairness and transparency of the platform.

[0059] Dynamic incentive mechanism: When the frequency of use of data or models on the platform changes, the smart contract dynamically adjusts the participants' benefits based on the real-time contribution data. Contributors of resources with higher frequency of use or higher precision will receive more rewards, which will motivate them to provide more high-quality resources.

[0060] Long-term profit mechanism: Smart contracts set long-term profit rules for high-quality data, models, and algorithms. When these resources are used multiple times, resource contributors can continue to receive profits to encourage their continued innovation and contribution.

[0061] Smart contracts ensure that all transaction processes on the platform are transparent and traceable by automating transaction execution and profit distribution, while reducing human intervention and improving operational efficiency.

[0062] Contribution quantification standards: The platform uses a standardized evaluation mechanism to ensure that contributions from all parties are quantified and benefits are distributed fairly. Specific standards include: Quantification of data contribution: Scoring is based on factors such as data type (such as traffic, environment, etc.), quantity, accuracy, and real-time performance. Contributors who provide high-quality, real-time data will receive higher scores and thus receive more benefits.

[0063] Quantification of model contribution: Evaluate the contribution of the model based on its accuracy, applicability, degree of optimization, etc. Model developers with high accuracy and strong versatility can obtain continuous benefits through the long-term use of the model.

[0064] Quantification of algorithm innovation: The platform conducts quantitative evaluation based on the innovation, computing efficiency and overall performance improvement of algorithms (including deep learning model algorithms, optimization algorithms and calculation methods for specific tasks). The better the algorithm performance, the higher the profit distribution weight the contributor will receive.

[0065] Quantification of operation services: The platform conducts quantitative evaluation based on the operation and maintenance investment and service quality of the operator. The platform operator is responsible for building and maintaining the platform, ensuring its stable operation and data security, and coordinating cooperation among all parties. Its income comes from the commission of the income of data providers and model developers. The commission ratio is determined by various factors such as platform resource investment and operating costs.

[0066] The contribution quantification standard ensures the fairness of the platform, enables the contributions of all participants to be accurately evaluated, and distributes corresponding benefits through smart contracts, further encouraging the openness and sharing of data and technology.

[0067] Application scenarios of value sharing: Through blockchain and smart contracts, the platform realizes value sharing in smart cities, ensuring that all participants can continue to gain benefits based on their contributions. The following are several typical application scenarios: Intelligent traffic dispatch system: Providers of real-time traffic data earn revenue based on the frequency and accuracy of their data usage. Developers of traffic light optimization algorithms earn long-term revenue by improving traffic efficiency, and the platform dynamically adjusts revenue distribution based on the optimization effect.

[0068] Environmental Monitoring System: Environmental data providers (such as operators of air quality monitoring sensors) receive revenue based on data quality and real-time availability, and developers of prediction models receive additional rewards based on prediction accuracy, ensuring that data and model contributors can benefit from the continued use of the platform.

[0069] The above mechanism ensures the open sharing and value maximization of various data, models and algorithms in smart cities. Through the application of blockchain technology, the contributions of participants are quantified in real time and receive fair profit distribution.

[0070] S600: Application scenarios and system integration Developers develop applications that meet the functions of various application scenarios based on the urban spatiotemporal data, models and algorithms of the open CIM platform, such as urban planning, traffic management, environmental monitoring and other different scenario requirements, or realize integration with government affairs and enterprise systems.

[0071] Specific applications such as intelligent traffic dispatching systems, environmental quality monitoring and early warning systems, and urban energy consumption optimization management systems all adopt a value sharing mechanism. Data providers, algorithm optimizers, system maintainers and other participants in each application scenario obtain corresponding income distribution through smart contracts in real time according to their value contribution within the system. At the same time, through the platform's standardized interface, application developers can flexibly access and expand new functional modules to further improve the platform's collaborative efficiency and innovation capabilities.

[0072] Intelligent traffic dispatch system: Use deep reinforcement learning algorithms to optimize the timing of traffic lights and maximize the efficiency of traffic flow. The value iteration formula of reinforcement learning is as follows:

[0073] in, is a state-value or state-action-value, It's a reward. is the discount factor, controlling for the impact of future rewards, is the state transition probability, describing the transition from state arrive possibility of transfer.

[0074] Environmental monitoring and early warning system: Use time series analysis and anomaly detection algorithms to monitor environmental factors such as air quality and noise pollution. Use time series prediction models such as ARIMA or LSTM to predict the air quality index. The formula is as follows:

[0075] in, is the predicted value, is the base value of the model, and are model parameters, Is with the past time The relevant input data, Is with the past time The relevant error term, is the error term.

[0076] Urban energy consumption optimization management system: Optimize energy dispatch and allocation based on real-time energy consumption data and weather forecasts. Use deep reinforcement learning models to model and predict urban energy consumption and optimize energy management strategies.

[0077] The present invention has been described in detail above through embodiments, but the contents described are only exemplary embodiments of the present invention and cannot be considered to limit the scope of implementation of the present invention. The protection scope of the present invention is defined by the claims. Anyone who utilizes the technical solution described in the present invention, or a technician in the field, inspired by the technical solution of the present invention, designs a similar technical solution within the essence and protection scope of the present invention to achieve the above technical effects, or makes equal changes and improvements to the scope of application, shall still belong to the scope of protection covered by the patent of the present invention. It should be noted that in order to make a clear statement, the description of some components and processes that have no direct and obvious connection with the scope of protection of the present invention but are known to those skilled in the art are omitted in the description of the present invention.

Claims

1. An open CIM spatiotemporal deep learning application platform based on value sharing, characterized in that: include: Data management module: used to store, manage and dispatch urban spatiotemporal data uploaded by different providers and data generated by the platform's own system; Model training module: Developers build, train and optimize deep learning models based on multi-source urban spatiotemporal data and application scenarios; Model application module: Apply the trained deep learning model to dynamic analysis, real-time prediction and scenario simulation of urban system application scenarios; Application interface module: provides standardized API interface and supports external system access; Value sharing and smart contract module: Smart contracts are used to realize the transaction, sharing and incentive mechanism of various urban spatiotemporal data, deep learning models and algorithms within the platform. Quantitative standards for contribution indicators are set for the data uploaded by data providers, the algorithms and models provided by developers, the operation and maintenance investment and service quality of platform operations. Smart contracts automatically calculate and distribute the benefits of contributors based on real-time contribution indicators; The platform adopts a consortium chain architecture and uses the Byzantine fault-tolerant algorithm as a consensus mechanism. Each node stores and verifies data through the consensus mechanism. The data management module transfers the acquired multi-dimensional spatiotemporal data to the model training module through a standard interface for training deep learning models. At the same time, the data management module records the use of data in real time and passes relevant data information to the value sharing and smart contract modules to ensure that data providers can obtain corresponding benefits based on their contributions. The model training module obtains multidimensional spatiotemporal data from the data management module and preprocesses the data to ensure data standardization and consistency. The preprocessed data is used to perform training and optimization of deep learning models. After the training is completed, the model training module uses the trained model in the model application module to perform dynamic analysis, real-time prediction or scenario simulation for different application scenarios; At the same time, the model training module feeds back the model performance data during the training process to the value sharing and smart contract module to calculate the contribution and profit distribution of the model developer; After completing the prediction task, the model application module returns the analysis results to the data management module for storage and for access and reference by other modules and external systems; The application interface module allows external systems to access data and models in the platform through a standardized API interface. External requests call the data of the data management module and the models of the model application module through the application interface module. Each call will be recorded in the smart contract and transmitted to the value sharing and smart contract module in real time to ensure that developers can obtain corresponding benefits according to their contributions. The value sharing and smart contract module receives usage data, model performance, API call information from the data management module, model training module, model application module and application interface module and processes them for automatic profit distribution.

2. The open CIM spatiotemporal deep learning application platform based on value sharing according to claim 1 is characterized in that: The platform has built-in deep learning models including convolutional neural networks, long short-term memory networks and graph neural networks. It uses convolutional neural networks, long short-term memory networks and graph neural networks to extract spatiotemporal features from pre-processed data to capture the dynamic changes of various elements in the city: extract spatial features through convolutional neural networks to capture various geographical areas in the city and their relationships; Use long short-term memory networks to model time series data and capture the temporal dependency and dynamic evolution in the data; use graph neural networks to process complex relational data and build a spatiotemporal feature model of the interaction and dependency between various elements in the urban system; The platform has a built-in adaptive learning rate optimization algorithm to dynamically adjust learning parameters to cope with changes in data streams. At the same time, cross-validation and early stopping techniques are used to prevent overfitting and ensure the generalization ability of the model. The model training and optimization process is accelerated with the help of a distributed computing framework, and parameter synchronization and model updates in distributed training are achieved through a parameter server architecture. Based on the extracted feature vectors and specific application scenarios, developers build a dynamic model framework based on deep learning to analyze and predict changes in urban systems; developers adjust and customize the platform's built-in algorithms, models, model training, and optimization technologies as needed.

3. The open CIM spatiotemporal deep learning application platform based on value sharing according to claim 1 is characterized in that: Multidimensional spatiotemporal data includes one or more of remote sensing image data, urban sensor data, social media data, traffic monitoring data, meteorological data, environmental monitoring data, transportation network data, and power network data.

4. The open CIM spatiotemporal deep learning application platform based on value sharing according to claim 3 is characterized in that: Preprocessing of multidimensional spatiotemporal data includes: Perform radiation correction and geometric correction on remote sensing image data; Use Kalman filters to clean and detect anomalies in urban sensor data; synchronize the data in time and correct the spatial coordinates to ensure the compatibility of different sensor data; Use natural language processing technology to perform sentiment analysis and geolocation extraction on social media data; use natural language processing technology to perform sentiment analysis, topic mining and geolocation extraction on social media texts, and use entity recognition technology to extract location, time and person information related to urban events; The density estimation algorithm is used to calculate the real-time flow distribution of traffic monitoring data; The extracted geographic location information is integrated with other data sources to enrich the urban spatiotemporal dataset.

5. The open CIM spatiotemporal deep learning application platform based on value sharing according to claim 1 is characterized in that: The platform adopts a multi-layer encryption mechanism. First, it uses an asymmetric encryption algorithm to encrypt and exchange keys to ensure the security of data transmission. Second, it uses a symmetric encryption algorithm to encrypt large-scale data to ensure that data can remain efficient and secure during storage and transmission. In addition, the platform introduces zero-knowledge proof technology, allowing data providers to prove the authenticity and validity of data to the blockchain network without exposing the content of the data.

6. The open CIM spatiotemporal deep learning application platform based on value sharing according to claim 1 is characterized in that: The data management module uses NoSQL database to store unstructured data and Elasticsearch to achieve fast data retrieval; it adopts distributed file system to achieve efficient storage and management of large-scale data and combines memory computing technology to improve data access speed.

7. The open CIM spatiotemporal deep learning application platform based on value sharing according to claim 1 is characterized in that: The model training module supports heterogeneous computing environments and uses containerization technology to achieve rapid deployment and expansion of the model training environment.

8. The open CIM spatiotemporal deep learning application platform based on value sharing according to claim 1 is characterized in that: The model application module feeds back the results to the model training module, triggering further optimization or retraining of the model to ensure the accuracy and timeliness of the model.

9. The open CIM spatiotemporal deep learning application platform based on value sharing according to claim 1 is characterized in that: The data management module supports real-time data updates and retrospective queries of historical data.

10. The open CIM spatiotemporal deep learning application platform based on value sharing according to claim 1, characterized in that: The model training module supports online training and offline training.

Images