Trusted monitoring of cold chain supply chain and intelligent decision-making method for cold chain warehousing

Abstract

The cold chain is a special branch of the logistics industry, which has relatively stringent requirements on the warehousing and transportation links in the supply chain. The safety of pharmaceutical cold chain products represented by vaccines is directly related to people's livelihood and social stability. The present invention constructs a trusted monitoring method for the cold chain supply chain based on blockchain, cloud storage and the Internet of Things to achieve trusted traceability of cold chain products throughout their life cycle, ensure product safety, and improve the credibility of cold chain product quality. Intelligent management decisions for cold chain warehousing are an important way to reduce cold chain warehousing costs and enhance the market competitiveness of cold chain companies. Based on the high-quality, large-scale data generated by the cold chain supervision process, the present invention constructs a cold chain product demand forecasting solution based on deep learning to assist cold chain warehousing management decisions to reduce the warehousing costs of cold chain products.

Description

Technical Field

[0001] The present invention relates to the field of cold chain supply chain, and in particular to a cold chain supply chain credible monitoring and cold chain warehousing inventory intelligent decision-making method. Background Art

[0002] The cold chain is a special branch of the logistics industry, which has relatively stringent requirements for warehousing and transportation in the supply chain. The safety of pharmaceutical cold chain products represented by vaccines is directly related to people's livelihood and social stability. The refrigerated pharmaceutical products circulated in the pharmaceutical cold chain include preventive biological products, therapeutic biological products and diagnostic biological products. Representative biological products include various vaccines, antisera (immune serum), antitoxins, toxoids, immune preparations (such as thymosin, immune nucleic acid, etc.), diagnostic reagents, etc. The complete pharmaceutical cold chain includes a series of links such as production, transportation, storage, and use. The particularity of refrigerated pharmaceuticals makes the pharmaceutical cold chain have relatively stringent requirements in warehousing and transportation. The cold chain requires physical means to ensure appropriate temperature conditions along the supply chain. Specialized storage, loading and unloading facilities and refrigeration unit temperature control are required.

[0003] The pharmaceutical cold chain industry is still in its early stages of development worldwide. The coverage of pharmaceutical cold chains is very low. Most cold chain drug quality issues are related to cold chain logistics, and cold chain disconnection problems still occur from time to time. At present, the pharmaceutical cold chain logistics system is not sound. The cold chain logistics technology and information level are not high, and there is a lack of overall planning and coordination between upstream and downstream. Each entity in the cold chain is independent of each other and maintains its own business data. There is a possibility of data tampering, which will inevitably reduce the credibility of the information. In order to protect the lives of the people, it is necessary to achieve credible traceability of pharmaceutical products throughout their life cycle and improve the credibility of pharmaceutical products.

[0004] Cold chain storage and transportation costs are high, and how to reduce cold chain logistics costs is an important research topic. Medicines and biological products in the cold chain are sensitive to ambient temperature and require special materials and temperature control facilities. Compared with the cost of ordinary supply chains, the cost of cold chain storage and transportation is significantly higher. According to supply chain theory, improving cold chain storage management capabilities is an important way to reduce the cost of cold chain companies and avoid product backlogs. How to improve cold chain storage management capabilities through technical means has become realistic and urgent. Summary of the invention

[0005] The technical problem to be solved by the present invention is to propose a cold chain supply chain trusted monitoring and cold chain warehouse inventory management intelligent decision-making method in view of the shortcomings of the existing technology.

[0006] 1. Build a trusted monitoring method for the cold chain supply chain based on blockchain, cloud storage and the Internet of Things (unless otherwise specified, "Internet of Things" and "IoT" have the same meaning) to achieve trusted traceability of cold chain products throughout their life cycle, ensure product safety, and improve the credibility of cold chain product quality.

[0007] 2. Based on the high-quality, large-scale data generated by the cold chain supervision process, a deep learning-based cold chain product demand forecasting solution is constructed to assist cold chain warehousing inventory management decisions to reduce the storage costs of cold chain products.

[0008] The technical solution of the present invention is:

[0009] 1. A trusted monitoring method for cold chain supply chain. The architecture of the trusted monitoring system adopted by the method is mainly composed of business layer, entity layer and data persistence layer. The business layer mainly includes business entities such as manufacturers, wholesalers, distributors, consumers, and government regulatory agencies. The business layer is used to abstract and map with real business logic. Cold chain product entities and related information realize physical circulation and information circulation in the business interaction process between various entities in the business layer. The entity layer mainly involves multiple different entity process objects such as production process, multi-level cold chain warehousing and multi-level cold chain transportation. In order to ensure the safety of the entire life cycle of cold chain products, we introduce the Internet of Things monitoring module to monitor different entity process objects. The data persistence layer involves blockchain and cloud storage. The business data of the business layer and the Internet of Things monitoring data of the entity layer are stored in the cloud storage platform. The storage address of the data in the cloud storage platform and the hash value of the data will be uploaded to the blockchain platform to enhance the credibility of the data. It mainly includes the following processes:

[0010] (1) Cold chain product registration: Manufacturers submit relevant supporting data for cold chain product registration to the cloud storage platform; Manufacturers submit cold chain product registration requests to the blockchain platform;

[0011] (2) Cold chain supply chain business process: The cold chain supply chain process is composed of a series of cold chain supply chain sub-processes connected in series; each cold chain supply chain sub-process consists of three types of business entities: supplier, cold chain logistics transporter and consignee; the supplier, cold chain logistics transporter and consignee respectively initiate data storage requests to the cloud storage platform; the supplier, cold chain logistics transporter and consignee respectively submit the storage address of the business data on the cloud storage platform together with the hash value of the data to the blockchain platform;

[0012] (3) Cold chain supply chain monitoring process: The cold chain supply chain monitoring is implemented by the Internet of Things module; the cold chain supply chain monitoring is mainly divided into three types of monitoring: production process monitoring, cold chain storage monitoring and cold chain logistics and transportation monitoring; the Internet of Things monitoring module is responsible for monitoring data collection and storing the monitoring data on the cloud storage platform; the storage address of the monitoring data on the cloud storage platform together with the hash value of the data will be submitted to the blockchain platform;

[0013] (4) Integration of cold chain supply chain business processes and cold chain supply chain monitoring processes: When registering cold chain products, a mapping relationship between the cold chain product number (product ID. Unless otherwise specified, "number" and "ID" have the same meaning) and the production process monitoring data is established; when the cold chain storage module registers the entry of cold chain products into the warehouse, a mapping relationship between the cold chain product number and cold chain storage monitoring is established; when the cold chain transportation module starts to register the loading of cold chain products, a mapping relationship between the cold chain product number and cold chain logistics monitoring is established.

[0014] Furthermore, the specific steps for cold chain product registration in step (1) are as follows:

[0015] (1.1) The manufacturer uses its own private key to digitally sign the business support data related to cold chain product registration; then the data and the digital signature are stored together in the cloud storage platform; the cloud storage platform will feedback the storage address of the business support data to the manufacturer;

[0016] (1.2) The manufacturer initiates a product registration request to the blockchain platform and sends its own number, business support data storage address, and data hash value; after the blockchain platform verifies the data, it generates a cold chain product number for the product and writes the received data storage address, data hash value and other data together with the cold chain product number into the blockchain; the blockchain platform feedbacks the cold chain product number to the manufacturer.

[0017] Furthermore, the specific steps of step (2) cold chain supply chain business process are as follows:

[0018] (2.1) Cold chain supply chain business process: A complete cold chain supply chain consists of a series of cold chain business sub-processes connected in series; each cold chain business sub-process consists of three types of business entities: supplier, cold chain logistics transporter and consignee; the supplier of the current cold chain supply chain sub-process is the consignee of the previous cold chain supply chain sub-process; the supplier of the current cold chain supply chain sub-process is the supplier of the next cold chain supply chain sub-process;

[0019] (2.2) Supplier business data storage process: The supplier prepares the supplier number and business data, and uses its own private key to digitally sign the prepared data content; then initiates a data storage request to the cloud storage platform; after the cloud storage platform completes the data storage process, it feedbacks the data storage address to the supplier; the cloud storage platform initiates a business record request to the blockchain platform, and writes the supplier number, data storage address, data hash value and other information to the blockchain platform; the blockchain platform feedbacks the record address on the blockchain to the supplier;

[0020] (2.3) Cold chain logistics transporter business data storage process: The transporter prepares the transporter number and business data, and uses its own private key to digitally sign the prepared data content; then initiates a data storage request to the cloud storage platform; after the cloud storage platform completes the data storage process, it feedbacks the data storage address to the transporter; the cloud storage platform initiates a business record request to the blockchain platform, and writes the transporter number, data storage address, data hash value and other information to the blockchain platform; the blockchain platform feedbacks the record address on the blockchain to the transporter;

[0021] (2.4) Consignee's business data storage process: The consignee prepares the consignee number and business data, and uses its own private key to digitally sign the prepared data content; then initiates a data storage request to the cloud storage platform; after the cloud storage platform completes the data storage process, it feedbacks the data storage address to the consignee; the cloud storage platform initiates a business record request to the blockchain platform, and writes the consignee number, data storage address, data hash value and other information into the blockchain platform; the blockchain platform feedbacks the record address on the blockchain to the consignee.

[0022] Furthermore, the specific steps of the cold chain supply chain monitoring process in step (3) are as follows:

[0023] (3.1) Cold chain supply chain monitoring process: The cold chain supply chain monitoring is implemented by the Internet of Things module; the cold chain supply chain monitoring is mainly divided into three types of monitoring: production process monitoring, cold chain warehousing monitoring and cold chain logistics and transportation monitoring;

[0024] (3.2) The IoT monitoring of the production process is mainly organized according to the production process and other methods; after the IoT monitoring module completes the regular collection of the production process monitoring data, it will send a data storage request to the cloud storage platform, and send the production process program number, time, data, and related digital signatures corresponding to the IoT monitoring data to the cloud storage platform; the cloud storage platform receives the request, performs data storage operations, stores the relevant data on the cloud storage platform, and feedbacks the monitoring data storage address to the IoT monitoring module;

[0025] (3.3) IoT monitoring in cold chain warehousing is mainly organized according to the storage partition number or storage shelf number. After the IoT monitoring module completes the regular collection of cold chain warehousing monitoring data, it will send a data storage request to the cloud storage platform and send the storage location number (storage location ID), time, data, and related digital signature corresponding to the IoT monitoring data to the cloud storage platform. After receiving the request, the cloud storage platform performs data storage operations, stores the relevant data on the cloud storage platform, and feedbacks the monitoring data storage address to the IoT monitoring module.

[0026] (3.4) IoT monitoring in cold chain logistics transportation is mainly organized according to container numbers or vehicle numbers. After the IoT monitoring module completes the regular collection of cold chain logistics monitoring data, it will send a data storage request to the cloud storage platform and send the vehicle number (vehicle ID), time, data, and related digital signature corresponding to the IoT monitoring data to the cloud storage platform. After receiving the request, the cloud storage platform performs data storage operations, stores the relevant data on the cloud storage platform, and feedbacks the monitoring data storage address to the IoT monitoring layer.

[0027] Furthermore, the specific steps of step (4) integrating the cold chain supply chain business process and the cold chain supply chain monitoring process are as follows:

[0028] (4) Integration of cold chain supply chain business process and cold chain supply chain monitoring process:

[0029] (4.1) Integration of production process monitoring data and business processes: When manufacturers register the aforementioned cold chain products, the IoT monitoring data generated during the production process of the cold chain products will be included as part of the business support data; the data will be submitted to the cloud storage platform; the blockchain record includes the storage address of the data in the cloud storage; when the blockchain platform generates a cold chain product number for the product, a mapping relationship between the cold chain product number and the production process monitoring data is established; various stakeholders can obtain the blockchain platform record of the corresponding record through the cold chain product number, and then obtain the monitoring data in the cloud storage platform through the cloud storage address in the record, and the authenticity of the monitoring data can be verified through the data hash value in the record;

[0030] (4.2) Integration of cold chain storage monitoring and business processes: When the cold chain storage module registers the entry of cold chain products into the warehouse, it will allocate specific storage locations for the cold chain products in the warehouse, thereby establishing a mapping between the cold chain product number and the storage location number; such data belongs to the business support data of the pharmaceutical cold chain supply chain and needs to be uploaded to the cloud storage module in accordance with the business data storage method; the cold chain storage module uses the storage location number and the cold chain product's in-stock time interval as input data to request the IoT monitoring module to access the IoT monitoring data; the IoT monitoring module will feedback the corresponding monitoring data storage address based on the information provided by the cold chain storage module; the cold chain storage module sends the cold chain product number, IoT monitoring data storage address and data hash value to the blockchain platform and requests that it be written into the blockchain platform; after verifying the data, the blockchain writes the data into the blockchain and feedbacks the on-chain record address to the cold chain storage module; the on-chain record address of the IoT monitoring will be sent to stakeholders as part of the business data; each stakeholder can find the record based on the blockchain record address to obtain the access address and hash value of the monitoring data;

[0031] (4.3) Integration of cold chain transportation monitoring and business processes: When the cold chain transportation module starts to register the loading of cold chain products, it will register the vehicle number information and cold chain product number information in the cold chain logistics transportation business record, thereby establishing a mapping between the cold chain product number and the vehicle number; this type of data belongs to the business support data of the pharmaceutical cold chain supply chain and needs to be uploaded to the cloud storage module in accordance with the business data storage method; the cold chain transportation module uses the vehicle number and the transportation time interval of the cold chain product as input data to request the IoT monitoring module to access the IoT monitoring data; the IoT monitoring module will feedback the corresponding monitoring data storage address based on the information provided by the cold chain transportation module; the cold chain transportation module will send the cold chain product number, IoT monitoring data storage address and data hash value to the blockchain platform and request that it be written to the blockchain platform; after the blockchain verifies the data, it will write the data to the blockchain and feedback the on-chain record address to the cold chain supply chain module; the on-chain record address of the IoT monitoring will be sent to stakeholders as part of the business data; each stakeholder can find the record based on the blockchain record address to obtain the access address and hash value of the monitoring data.

[0032] 2. A cold chain warehouse inventory management method, comprising the following steps:

[0033] (1) Data preprocessing: The data preprocessing module obtains the storage addresses of cold chain business data and cold chain product consumption data in the cloud computing platform through the storage addresses in the blockchain records; and verifies the authenticity of the data stored in the cloud computing platform through the hash value in the blockchain;

[0034] (2) Cold chain warehousing demand prediction: The cold chain warehousing demand prediction model is constructed based on deep neural network technology and mainly consists of three modules: convolution operation module, long short-term memory network operation module and graph embedding operation module. After the original data is preprocessed, the obtained data is input into the convolution operation module in chronological order. At the same time, the original data will also be used to construct similar graphs after data preprocessing and input into the graph embedding operation module. The data processed by the convolution layer will be aggregated through the fully connected layer and input into the long short-term memory network operation module. The output of the graph embedding operation module will be processed by the fully connected layer, and then integrated and spliced ​​with the output of the long short-term memory network operation module, and then processed by the fully connected layer to obtain the final output.

[0035] (3) Uncertainty measurement of cold chain storage demand: The Gaussian model is used to model the distribution of storage demand uncertainty. Based on the aforementioned deep neural network prediction model, the dropout method is used to estimate the uncertainty of cold chain storage demand, and the mean and variance of the Gaussian model are obtained as follows:

[0036]

[0037]

[0038] Among them, T is the number of sampling times and τ is the model accuracy.

[0039] (4) Cold chain warehouse management decision: The optimal cold chain warehouse inventory decision is modeled as an expected cost minimization problem; the optimal decision result is obtained as

[0040]

[0041] Where φ represents the cumulative distribution function of the uncertain demand y, α is the purchase cost per unit of inventory, β is the various loss values ​​caused by insufficient inventory per unit, and γ is the various cost increases caused by excess inventory per unit;

[0042] Use the cumulative distribution function F(x; μ, σ) of the Gaussian random model to replace the aforementioned cumulative distribution function φ:

[0043]

[0044] The mean and variance of the Gaussian model obtained in step (3) are used as uncertainty measurements and assigned to μ and σ in the cumulative distribution function respectively. 2 .

[0045] Beneficial effects:

[0046] The solution adopted by the present invention realizes the trusted traceability of cold chain products throughout their life cycle by building a trusted monitoring method for the cold chain supply chain based on blockchain, cloud storage and the Internet of Things, ensures product safety and improves the credibility of cold chain product quality. Based on the high-quality, large-scale data generated during the cold chain supervision process, a deep learning-based cold chain product demand forecasting solution is built to assist in cold chain warehousing management decision-making, thereby reducing the storage cost of cold chain products. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 System architecture diagram

[0048] Figure 2 Pharmaceutical Product Registration

[0049] Figure 3 Cold chain logistics and cold chain logistics sub-processes

[0050] Figure 4 Trusted storage of cold chain business data

[0051] Figure 5 Trusted storage of cold chain warehouse monitoring data

[0052] Figure 6 Trusted storage of cold chain transportation monitoring data

[0053] Figure 7 Data preprocessing and inventory management process

[0054] Figure 8 Cold chain warehousing demand prediction based on deep learning

[0055] Fig. 9 Performance comparison with different parameters: (a) different convolutional sizes and layers (b) different sequence lengths DETAILED DESCRIPTION

[0056] The specific implementation process of the present invention is as follows:

[0057] 1. Overview of system architecture.

[0058] Typical application scenarios of the cold chain supply chain include pharmaceutical cold chain, agricultural product cold chain, etc. For the convenience of description, this manual mainly describes the implementation process based on the pharmaceutical (such as vaccine) cold chain application scenario.

[0059] The trusted monitoring system architecture of the cold chain supply chain trusted monitoring method is as follows Figure 1As shown. The system architecture mainly includes the business layer, entity layer and data persistence layer. The business layer is used to abstract and map with the actual business logic. Representative business layer entities mainly include manufacturers, wholesalers, distributors, consumers, government regulatory agencies, etc. Taking the vaccine cold chain scenario as an example, the business entities in the pharmaceutical cold chain mainly include manufacturers (such as pharmaceutical factories), distributors, terminal channels (such as hospitals or clinics), and consumers (such as vaccination subjects, patients, etc.). In addition, there are government regulatory agencies responsible for supervising the production and consumption safety of pharmaceutical products. Cold chain product entities and related information realize physical circulation and information circulation in the business interaction process between various entities in the business layer.

[0060] The following is a brief description of each entity in the supply chain.

[0061] Pharmaceutical factory (pharmaceutical production enterprise): An enterprise engaged in drug production activities needs to obtain a drug production license. Manufacturers shall review and inspect the entire drug production process and drug quality in accordance with regulations. Manufacturers shall establish a complete production quality management system, continuously strengthen deviation management, and use information technology to truthfully record all data generated during the production and inspection process to ensure that the entire production process continues to meet legal requirements.

[0062] Distributors: Distributors are usually middlemen who transfer goods from manufacturers to the next level of entities. There are usually multiple levels of distributors in the market. The existence of distributors is conducive to promoting the circulation of pharmaceutical products and expanding the sales scope of products.

[0063] Hospitals or clinics: Hospitals or clinics provide medical diagnosis services to patients, and prescribe and sell medical products to patients based on the diagnosis results. Some types of medicines can also be sold in pharmacies. The solution of the present invention can be extended to such situations.

[0064] Consumer: The patient is the individual who ultimately purchases and uses the product. For infants and young children, their legal guardians usually handle related matters on their behalf.

[0065] Government regulatory authorities: There are two types of government regulatory authorities. One is the pharmaceutical product quality regulatory authorities, which mainly include drug supervision and administration departments and their authorized product approval agencies. The other is the health and wellness administrative department and its subordinate agencies.

[0066] The physical layer mainly involves multiple different physical links (physical process objects) such as production process, multi-level cold chain warehousing and multi-level cold chain transportation. The process of vaccine products circulating from manufacturers to vaccine consumers usually passes through multiple business entities such as multi-level wholesalers and multi-level disease prevention and control centers. Corresponding to this, there will also be multi-level cold chain warehousing and multi-level cold chain transportation. Cold chain warehousing is usually used to meet the needs of vaccine storage of various business entities at all levels. Cold chain transportation is usually used to meet the transportation needs of vaccine products between two adjacent business entities (such as: between upper and lower disease prevention and control centers). Most vaccines are very sensitive to ambient temperature and need to maintain a low temperature throughout the process. To ensure the safety of cold chain products throughout their life cycle, we introduce the Internet of Things monitoring module to monitor different physical process objects. Many vaccine safety issues are related to cold chain breaks. To ensure the safety of vaccines throughout their life cycle, we introduce the Internet of Things monitoring module in the physical layer to monitor different physical layer objects.

[0067] The IoT monitoring module is an abstract interface. Different entity objects in the entity layer construct their own IoT module object instances according to specific needs to complete the corresponding monitoring tasks. The IoT monitoring module consists of a sensor layer, a communication layer, and an IoT interface layer. The IoT monitoring module collects various vaccine environment and vaccine status monitoring data through its sensor layer. The sensor type and number of sensors in the sensor layer can be configured in different types of combinations as needed. For example, in the cold chain transportation monitoring scenario, the IoT monitoring module usually needs to be equipped with a GPS module to monitor the vehicle trajectory, while in the cold chain storage monitoring scenario, the GPS module does not need to be configured. The communication layer is responsible for the communication between the various modules of the IoT layer. The interface layer of the IoT module abstracts the functions of the IoT module and provides an interactive interface to the outside world. For example, the IoT module interacts with the cloud storage module to achieve data storage.

[0068] The data persistence layer involves blockchain and cloud storage; the former ensures the credibility of data, and the latter is used for cloud storage of data. The business data of the business layer and the IoT monitoring data of the physical layer are stored in the cloud storage platform; the storage address of the data in the cloud storage platform and the hash value of the data will be uploaded to the blockchain platform to enhance the credibility of the data. Blockchain has the characteristics of decentralization and immutability, and can be used for data credibility issues in the process of pharmaceutical cold chain monitoring and traceability. The data maintenance cost on the blockchain is too high. The amount of IoT monitoring data and business support data in the circulation of pharmaceutical products is very large. Therefore, it is not suitable to be directly stored in the blockchain. For this reason, we introduce a cloud storage module to store data. Before storing, the data is signed with a private key to prevent the data from being forged. By recording the address and hash value of these data in the blockchain, the relevant data can be accessed and its integrity can be verified.

[0069] The system allocates public-private key pairs, blockchain platform accounts, and cloud storage platform accounts to business layer objects, entity layer objects, and the IoT monitoring modules associated with them as needed. When various entities communicate and transmit data, we will use the public key mechanism for identity authentication and digital signature of data to ensure the communication security of the system. For the convenience of description, such security field consensus technologies will not be mentioned one by one.

[0070] The cold chain supply chain trustworthy monitoring method mainly includes the following processes:

[0071] (1) Cold chain product registration: Manufacturers submit relevant supporting data for cold chain product registration to the cloud storage platform; Manufacturers submit cold chain product registration requests to the blockchain platform;

[0072] (2) Cold chain supply chain business process: The cold chain supply chain process is composed of a series of cold chain supply chain sub-processes connected in series; each cold chain supply chain sub-process consists of three types of business entities: supplier, cold chain logistics transporter and consignee; the supplier, cold chain logistics transporter and consignee respectively initiate data storage requests to the cloud storage platform; the supplier, cold chain logistics transporter and consignee respectively submit the storage address of the business data on the cloud storage platform together with the hash value of the data to the blockchain platform;

[0073] (3) Cold chain supply chain monitoring process: The cold chain supply chain monitoring is implemented by the Internet of Things module; the cold chain supply chain monitoring is mainly divided into three types of monitoring: production process monitoring, cold chain storage monitoring and cold chain logistics and transportation monitoring; the Internet of Things monitoring module is responsible for monitoring data collection and storing the monitoring data on the cloud storage platform; the storage address of the monitoring data on the cloud storage platform together with the hash value of the data will be submitted to the blockchain platform;

[0074] (4) Integration of cold chain supply chain business processes and cold chain supply chain monitoring processes: When registering cold chain products, a mapping relationship between the cold chain product number and the production process monitoring data is established; when the cold chain storage module registers the entry of cold chain products into the warehouse, a mapping relationship between the cold chain product number and cold chain storage monitoring is established; when the cold chain transportation module starts to register the loading of cold chain products, a mapping relationship between the cold chain product number and cold chain logistics monitoring is established.

[0075] 2. Cold chain product registration.

[0076] The manufacturer uses its own private key to digitally sign the business support data related to cold chain product registration; then the data is stored together with the digital signature in the cloud storage platform; the cloud storage platform will feedback the storage address of the business support data to the manufacturer. The manufacturer initiates a product registration request to the blockchain platform and sends its own number, business support data storage address, and data hash value; after the blockchain platform verifies the data, it generates a cold chain product number for the product, and writes the received data storage address, data hash value and other data together with the cold chain product number into the blockchain; the blockchain platform feedbacks the cold chain product number to the manufacturer.

[0077] like Figure 2 As shown in the figure, pharmaceutical product registration is initiated by pharmaceutical manufacturers. Pharmaceutical manufacturers use their own private keys to sign the business support data related to pharmaceutical product registration (such as approval data from government authorities and support data for the production process). The data is then stored together with the signature in the cloud storage module. The cloud storage module will feedback the data storage address. Private key signatures can prevent others from forging data.

[0078] The pharmaceutical manufacturer initiates a pharmaceutical product registration request to the blockchain platform and sends its own number, supporting data storage address, and data hash value. After verifying the data, the blockchain platform generates a number for the pharmaceutical product. The received data storage address, data hash value, and other data are written into the blockchain together with the product number. By writing the data hash value into the blockchain record, we can prevent pharmaceutical manufacturers from tampering with business supporting data.

[0079] 3. Cold chain supply chain business process.

[0080] Cold chain supply chain business process: A complete cold chain supply chain is composed of a series of cold chain business sub-processes (also called cold chain logistics sub-processes) connected in series; each cold chain business sub-process is composed of three types of business entities: supplier, cold chain logistics transporter and consignee; the supplier of the current cold chain supply chain sub-process is the consignee of the previous cold chain supply chain sub-process; the supplier of the current cold chain supply chain sub-process is the supplier of the next cold chain supply chain sub-process.

[0081] Products are transferred from producers to consumers through the cold chain logistics process. Generally speaking, the cold chain logistics process usually involves many different business entities. In order to achieve seamless supervision of the product cold chain flow process, we abstract the complete cold chain supply chain into multiple cold chain logistics sub-processes that are connected in series. Figure 3As shown in the figure, the cold chain logistics sub-process consists of three types of business entities: supplier, cold chain logistics transporter and consignee. The consignee of the previous sub-process will become the supplier of the next process. These three types of business entities will generate different types and quantities of business data. The process of business data storage for different business entities is basically similar. They need to provide their own numbers, business data, and use their own private keys to sign the business data content. The storage address of the business data and its hash value will be written to the blockchain platform. After the cloud storage platform completes the data storage, it will feedback the storage address of the business data. The blockchain platform is responsible for feedback of the record address on the blockchain.

[0082] The cold chain logistics process involves a large number of business processes, which will generate a large amount of business data. If these data are directly stored in the blockchain, the cost is too high. Therefore, we store all these business data in the cloud storage module. What is written in the blockchain is the storage address of the business data, which will significantly reduce the amount of data in the blockchain.

[0083] The supplier's business data storage process is as follows Figure 4 As shown. Supplier ( Figure 4 The supplier prepares the supplier number and business data, and uses its own private key to digitally sign the prepared data content; then initiates a data storage request to the cloud storage platform; after the cloud storage platform completes the data storage process, it feeds back the data storage address to the supplier; the cloud storage platform initiates a business record request to the blockchain platform, and writes the supplier number, data storage address, data hash value and other information into the blockchain platform; the blockchain platform feeds back the record address on the blockchain to the supplier.

[0084] The storage process of cold chain logistics transportation business data is as follows Figure 4 As shown. Cold chain logistics transporter ( Figure 4 The second column in the figure) prepares the transporter number and business data, and uses its own private key to digitally sign the prepared data content; then initiates a data storage request to the cloud storage platform; after the cloud storage platform completes the data storage process, it feeds back the data storage address to the transporter; the cloud storage platform initiates a business record request to the blockchain platform, and writes the transporter number, data storage address, data hash value and other information to the blockchain platform; the blockchain platform feeds back the record address on the blockchain to the transporter.

[0085] The storage process of the consignee's business data is as follows: Figure 4 As shown. Consignee ( Figure 4The third column in the figure prepares the consignee number and business data, and uses its own private key to digitally sign the prepared data content; then initiates a data storage request to the cloud storage platform; after the cloud storage platform completes the data storage process, it feeds back the data storage address to the consignee; the cloud storage platform initiates a business record request to the blockchain platform, and writes the consignee number, data storage address, data hash value and other information into the blockchain platform; the blockchain platform feeds back the record address on the blockchain to the consignee.

[0086] 4. Cold chain supply chain monitoring process

[0087] Cold chain supply chain monitoring process: The cold chain supply chain monitoring is implemented by the Internet of Things module; the cold chain supply chain monitoring is mainly divided into three types of monitoring: production process monitoring, cold chain warehousing monitoring and cold chain logistics and transportation monitoring. The main difference between the Internet of Things modules in the three types of monitoring lies in the monitoring granularity and data storage organization method. The Internet of Things monitoring of the production process is mainly organized according to production processes and other methods. The Internet of Things monitoring in cold chain warehousing is mainly organized according to storage partitions or shelves. The Internet of Things monitoring in cold chain logistics and transportation is mainly organized in the form of containers or vehicles. Sensor modules that share certain data in the same storage partition or shelf are conducive to reducing costs. For example, taking video surveillance as an example, all products placed in the same camera monitoring area can share the video surveillance data of the camera. For example, in cold chain warehousing monitoring ( Figure 5 When storing data, the storage location number (storage location ID) is used as the data index key field. For cold chain logistics transportation monitoring, when storing data, the vehicle number is used as the data index key field.

[0088] The IoT monitoring of the production process is mainly organized according to the production process and other methods; after the IoT monitoring module completes the regular collection of production process monitoring data, it will send a data storage request to the cloud storage platform, and send the production flow program number, time, data, and related digital signatures corresponding to the IoT monitoring data to the cloud storage platform; the cloud storage platform receives the request, performs data storage operations, stores the relevant data on the cloud storage platform, and feedbacks the monitoring data storage address to the IoT monitoring module.

[0089] IoT monitoring in cold chain warehousing is mainly organized according to storage partition numbers or storage shelf numbers. IoT monitoring module ( Figure 5In the second column, after the "IoT monitoring" module completes the regular collection of cold chain warehouse monitoring data, it will send a data storage request to the cloud storage platform, and send the storage location number (storage location ID), time, data, and related digital signatures corresponding to the IoT monitoring data to the cloud storage platform; the cloud storage platform receives the request, performs data storage operations, stores the relevant data on the cloud storage platform, and feeds back the monitoring data storage address to the IoT monitoring module.

[0090] IoT monitoring in cold chain logistics is mainly organized according to container numbers or vehicle numbers. IoT monitoring module ( Figure 6 In the second column, after the "IoT monitoring" module completes the regular collection of cold chain logistics monitoring data, it will send a data storage request to the cloud storage platform, and send the vehicle number, time, data, and related digital signatures corresponding to the IoT monitoring data to the cloud storage platform; the cloud storage platform receives the request, performs data storage operations, stores the relevant data on the cloud storage platform, and feeds back the monitoring data storage address to the IoT monitoring layer.

[0091] The IoT module is responsible for storing data in the cloud storage module. When the IoT module stores data, it needs to use a private key to sign the data to confirm the source of the data. The cloud storage module will feedback the storage address of the data.

[0092] 5. Integration of cold chain supply chain business processes and cold chain supply chain monitoring processes.

[0093] Integration of production process monitoring data and business processes: When manufacturers register the aforementioned cold chain products, they use the IoT monitoring data generated during the cold chain product production process as part of the business support data (e.g. Figure 2 The data will be submitted to the cloud storage platform; the blockchain record includes the storage address of the data in the cloud storage; when the blockchain platform generates a cold chain product number for the product, a mapping relationship between the cold chain product number and the production process monitoring data is established; each stakeholder can obtain the corresponding blockchain platform record through the cold chain product number, and then obtain the monitoring data in the cloud storage platform through the cloud storage address in the record, and the authenticity of the monitoring data can be verified through the data hash value in the record.

[0094] Integration of cold chain storage monitoring and business processes: When the cold chain storage module registers the entry of cold chain products ( Figure 5 In the "Warehouse Registration" in the warehousing, a specific storage location will be assigned to the cold chain product in the warehouse, thereby establishing a mapping between the cold chain product number and the storage location number ( Figure 5Such data belongs to the business support data of the pharmaceutical cold chain supply chain and needs to be uploaded to the cloud storage module in accordance with the business data storage method described in the previous article. The cold chain storage module uses the storage location number and the cold chain product's storage time interval as input data to request the IoT monitoring module to access the IoT monitoring data. The IoT monitoring module will feedback the corresponding monitoring data storage address based on the information provided by the cold chain storage module. The cold chain storage module sends the cold chain product number, IoT monitoring data storage address and data hash value to the blockchain platform and requests that it be written into the blockchain platform. After verifying the data, the blockchain writes the data into the blockchain and feedbacks the on-chain record address to the cold chain storage module. The on-chain record address of the IoT monitoring will be sent to stakeholders as part of the business data. Each stakeholder can find the record based on the blockchain record address to obtain the access address and hash value of the monitoring data. When the cold chain warehouse registers the entry of pharmaceutical products, it will allocate a specific storage location for the pharmaceutical products in the warehouse, thereby establishing a mapping between its pharmaceutical product number and the storage location number. This type of data belongs to the business data of pharmaceutical cold chain logistics and needs to be uploaded to the cloud storage module. The storage of business data has been introduced in the "Cold Chain Supply Chain Business Process".

[0095] Integration of cold chain transportation monitoring and business processes: When the cold chain transportation module starts to register the loading of cold chain products, the vehicle number information and cold chain product number information in the cold chain logistics transportation business record are registered to establish a mapping between the cold chain product number and the vehicle number ( Figure 6 Such data belongs to the business support data of the pharmaceutical cold chain supply chain and needs to be uploaded to the cloud storage module in accordance with the business data storage method; the cold chain transportation module uses the vehicle number and the transportation time interval of the cold chain product as input data to request the IoT monitoring module to access the IoT monitoring data; the IoT monitoring module will feedback the corresponding monitoring data storage address based on the information provided by the cold chain transportation module; the cold chain transportation module sends the cold chain product number, IoT monitoring data storage address and data hash value to the blockchain platform, requesting that it be written to the blockchain platform; after verifying the data, the blockchain writes the data to the blockchain and feedbacks the on-chain record address to the cold chain supply chain module; the on-chain record address of the IoT monitoring will be sent to stakeholders as part of the business data; each stakeholder can find the record based on the blockchain record address to obtain the access address and hash value of the monitoring data.

[0096] The cold chain warehouse uses the storage location number and the time interval of the pharmaceutical product in the warehouse as input data to request the IoT monitoring module for the access address of the IoT monitoring data. The IoT monitoring module will feedback the corresponding monitoring data storage address ( Figure 5 The cold chain warehouse sends the pharmaceutical product number, IoT monitoring data storage address, and data hash value to the blockchain platform, requesting that it be written into the blockchain platform. After the blockchain verifies the data, it writes the data into the blockchain and feeds back the on-chain record address. The on-chain record address of the IoT monitoring will be sent to stakeholders as part of the business data. Each stakeholder can find the record based on the blockchain record address to obtain the access address and hash value of the monitoring data. The cold chain transportation process is handled in a similar way.

[0097] 6. Cold chain warehouse inventory management.

[0098] Cold chain warehousing management is a very important link in the pharmaceutical cold chain. According to modern supply chain theory, warehousing is the core of the supply chain. Warehousing is the inventory control center in the supply chain. Warehousing is the dispatching center in logistics and supply chain. Warehousing is directly related to the efficiency and response speed of the supply chain.

[0099] The cost of cold chain warehousing is very high. On the one hand, if the inventory quantity is insufficient, it will trigger cross-regional cold chain product allocation or temporary emergency purchase, which will bring additional procurement costs and time delays, and the cold chain transportation cost will also increase sharply. Especially when the upper-level supplier also has insufficient inventory problems, the procurement cost and cycle will be greatly extended. Since consumer demand cannot be met in a timely manner, it will also reduce service quality and lose existing customers or potential customers. In severe cases, it may even have a negative impact on the health of the vaccine recipients. On the other hand, excessive purchases will increase the operating costs of cold chain warehousing. In severe cases, it may even lead to a backlog of cold chain products and a large number of expired products.

[0100] Through the feasible monitoring link based on the cold chain supply chain, each entity node in the supply chain uploads various business data in the supply chain to cloud storage. Data is signed by each entity node before uploading. The hash value and cloud storage address of the data are uploaded to the blockchain platform. The data signature mechanism makes it impossible for other nodes to forge data, and data hashing and blockchain technology make it impossible for entity nodes to tamper with data. This achieves the credibility of cloud storage data. Blockchain-based cold chain monitoring realizes seamless connection of full-process business data. This not only ensures the reliable traceability of the product throughout its life cycle, but also facilitates the extraction and preprocessing of training data. High-quality, large-volume data stored in the cloud storage platform and blockchain provide possibilities for intelligent applications.

[0101] The data preprocessing module obtains data from cloud storage and blockchain respectively. The data on the blockchain is mainly used for data verification. After data cleaning and sorting, the data is used as training data for training various prediction models.

[0102] The cold chain warehouse inventory management method includes the following steps:

[0103] (1) Cold chain warehousing demand prediction: The cold chain warehousing demand prediction model is constructed based on deep neural network technology and mainly consists of three modules: convolution operation module, long short-term memory network operation module and graph embedding operation module. After the original data is preprocessed, the obtained data is input into the convolution operation module in chronological order. At the same time, the original data will be used to construct similar graphs after data preprocessing and input into the graph embedding operation module. The data processed by the convolution layer will be aggregated through the fully connected layer and input into the long short-term memory network operation module. The output of the graph embedding operation module will be processed by the fully connected layer, and then integrated and spliced ​​with the output of the long short-term memory network operation module, and then processed by the fully connected layer to obtain the final output.

[0104] (2) Measurement of uncertainty in cold chain storage demand: The Gaussian model is used to model the distribution of storage demand uncertainty. Based on the aforementioned deep neural network prediction model, the dropout method is used to estimate the uncertainty of cold chain storage demand, and the mean and variance of the Gaussian model are obtained as follows:

[0105]

[0106]

[0107] (3) Cold chain warehouse management decision: The optimal cold chain warehouse inventory decision is modeled as an expected cost minimization problem; the optimal decision result is obtained as Where φ represents the cumulative distribution function of the uncertain demand y, α is the purchase cost per unit of inventory, β is the various loss values ​​caused by insufficient inventory per unit, and γ is the various cost increases caused by excess inventory per unit. The cumulative distribution function F(x; μ, σ) of the Gaussian random model is used to replace the aforementioned cumulative distribution function φ:

[0108]

[0109] The mean and variance of the Gaussian model obtained in the previous step are used as uncertainty measurements and assigned to μ and σ in the cumulative distribution function respectively. 2 The following is a detailed description of each link.

[0110] 7. Data preprocessing

[0111] The data preprocessing process is as follows Figure 7As shown. The data preprocessing module obtains the storage address of cold chain business data and cold chain product consumption data in the cloud computing platform through the storage address in the blockchain record; the authenticity of the data stored in the cloud computing platform is verified by the hash value in the blockchain. The entire city is divided into several non-intersecting rectangular areas, and the data in each area is integrated to obtain a data set of spatial dimension. The node in the spatial dimension is the aggregation of business data at the location of the area. It should be noted that since business data are all high-dimensional data, spatial dimension data are actually high-dimensional spatial data composed of multiple spatial dimension data slices superimposed. Next, the spatial dimension data is organized in the form of time series, and the size of the time slice is generally 1 day.

[0112] 8. Cold chain storage demand forecast

[0113] like Figure 8 As shown in Figure 1, the cold chain storage demand forecasting model consists of three modules: convolution operation module, long short-term memory network operation module and graph embedding operation module. The convolution operation module is used to learn the business data of cold chain products from the perspective of spatial correlation. The long short-term memory network operation module is used to learn the business data of cold chain products from the perspective of temporal correlation. The graph embedding operation module is used to break through the constraints of spatial distance and establish connections between nodes that are far away but have similar business models.

[0114] Convolution operation module: The data output by the data preprocessing module is input into the convolution operation module in chronological order. The convolution operation module extracts the spatial correlation features in the data through the convolution layer. The data processed by the convolution layer will be aggregated through the fully connected layer to achieve docking with the long short-term memory network operation module. After the spatial data at different times are extracted by the convolution operation module, a convolution operation feature sequence organized in chronological order can be obtained.

[0115] Long Short-Term Memory Network Operation Module: The Long Short-Term Memory Network Operation Module has good long-term memory capabilities. We use the Long Short-Term Memory Network Operation Module to extract time-related features. In the previous step, the convolution operation module extracted the spatial correlation features of the data at different time nodes. We send these convolution layer feature sequences to the Long Short-Term Memory Network Operation Module in chronological order to extract the time-related features contained in the sequence.

[0116] Graph embedding operation module: According to common sense, we can judge that the spatial correlation between nodes is not entirely determined by the distance of the geographical location. For example, readers can compare the following two situations: First, two areas that are far apart, but both belong to areas with a relatively concentrated residential area; second, two areas that are close to each other, but one is a residential area and the other is an industrial area. Obviously, we have reason to believe that the demand pattern similarity between the two areas in the first situation is higher. In this invention, we will capture this long-distance similarity through the graph embedding operation module. Specifically, the following steps are included:

[0117] First, construct a similarity graph. We construct an undirected graph with all regions as vertices and the similarity between regions as weights. According to the demand of each region at each time, we can get a demand sequence, and the similarity between different regions can be obtained through the dynamic time normalization algorithm.

[0118] Secondly, graph embedding operation. Based on the similarity graph obtained in the previous step, we transform it using graph embedding operation. This method can construct a vector for each region, which describes the spatial similarity between nodes.

[0119] Finally, the above vector is transformed through a fully connected layer to achieve integration with the output of the long short-term memory network operation module.

[0120] 9. Uncertainty measurement of cold chain storage demand

[0121] The present invention uses deep learning technology to predict cold chain storage demand. However, mainstream deep learning technology cannot express model uncertainty. These traditional neural networks usually use maximum likelihood estimation or maximum a posteriori to train and optimize parameters, ignoring uncertainty, so what is produced is often a point estimate result, and the probability distribution of the uncertainty value cannot be given. In this type of solution, the most common method for quantitatively measuring uncertainty is to apply the softmax function to obtain probability, but this often produces a large training bias. The Bayesian model has a mathematically complete framework that can derive model uncertainty. However, the uncertainty estimation algorithm based on the Bayesian neural network is very complex to calculate.

[0122] To address this challenge, the present invention uses a dropout-based method to estimate the uncertainty of cold chain storage demand. Dropout is a common regularization technique in the field of deep learning, and is often used to solve the problem of overfitting that is often encountered when training neural networks. The basic idea of ​​dropout is to stop the neurons in the hidden layer with a certain probability during the forward propagation process. This prevents the model from being too dependent on certain local features, thereby improving the generalization ability of the model. Dropout training in deep neural networks can be regarded as approximate Bayesian reasoning in a deep Gaussian process, so uncertainty can be modeled through dropout. This method can reduce computational complexity without affecting the accuracy of the estimation results.

[0123] The predicted probability based on the deep Gaussian model is described by the following expression:

[0124] p(y t |x t , X, Y) = ∫p(y t |ω)p(ω|x t ,X,Y)dω

[0125] in is a random variable associated with the L-layer deep model. X and Y are the input and output datasets, respectively.

[0126] Based on the above predicted probability distribution, the mean and variance can be calculated.

[0127]

[0128]

[0129] Among them, T is the number of sampling times and τ is the model accuracy.

[0130] 10. Cold chain inventory management decisions

[0131] In the context of uncertain demand for cold chain products, by making the best decision on cold chain warehouse inventory, we can ensure that the expected cost is minimized within a single ordering cycle. We model the optimal cold chain warehouse inventory decision as the following expected cost minimization problem:

[0132] s * =argminC

[0133] Where C is the expected cost. The expected cost C is mainly composed of fixed cost c 1 Variable costs c 2 , various losses caused by insufficient inventory resources c 3 , various cost increases caused by excess inventory 4These four parts can be described by the following formula:

[0134] C=c 1 +c 2 +c 3 +c 4

[0135] Fixed costc 1 It is mainly the basic cost to maintain the normal operation of the business. The variable cost can be expressed as c 2 =α(ss 0 ), where s 0 is the inventory at the beginning of the current decision cycle, s is the inventory at the current decision cycle, and α is the purchase cost per unit of inventory. 3 It can be expressed as c 3 =β(max(ys,0)), where β is the loss value caused by insufficient inventory per unit, and y is a random variable used to describe uncertain demand. The various costs caused by excess inventory can be expressed as c 4 =γ(max(sy,0)), where γ is the added value of various costs caused by excess inventory per unit.

[0136] The above cold chain warehouse inventory decision model is essentially the classic newsboy problem. Therefore, it can be solved by the standard first-order conditions of the classic newsboy problem. The optimal result is

[0137]

[0138] Where φ represents the cumulative distribution function of the uncertain demand y. According to the inventory management decision problem model and solution, in order to achieve the optimal decision of inventory management, it is necessary to predict and quantitatively measure the demand for cold chain warehousing and the uncertainty of demand to obtain its cumulative probability distribution function. The key to solving the above cold chain warehousing inventory decision problem lies in how to obtain the cumulative distribution function of the uncertain demand y. More specifically, under the assumption of the Gaussian model, we need to quantitatively calculate the mean and variance of the uncertain demand.

[0139] In the previous step, based on the aforementioned deep neural network prediction model, the dropout method was used to estimate the uncertainty of cold chain storage demand, and the mean and variance of the Gaussian model were obtained as follows:

[0140]

[0141]

[0142] The present invention uses a Gaussian model to model the uncertainty distribution of storage demand. Therefore, the cumulative distribution function F(x; μ, σ) of the Gaussian random model can be used to replace the aforementioned cumulative distribution function φ:

[0143]

[0144] The mean and variance are . The mean and variance of the Gaussian model obtained in the previous step are used as uncertainty measurements and assigned to the mean and variance (μ and σ) in the cumulative distribution function respectively. 2 ).

[0145] 11. Analysis of credible traceability.

[0146] Data credibility. First, the problem of data forgery and denial. All supporting data are stored on the cloud computing platform, and private key signatures are required before data storage. This can prevent malicious third parties from forging the data. On the other hand, it can also prevent the data uploader from denying the data. Second, data tampering. Malicious third parties cannot tamper with data without being discovered because they do not have private keys. Once a drug safety incident occurs, legitimate users have the motivation to tamper with data to evade responsibility. However, since the hash value of the data is stored on the blockchain, the hash value will change once the data is tampered with. Therefore, data tampering can be easily detected by users.

[0147] Traceability issues. Through cold chain product registration on the blockchain, cold chain products are assigned unique numbers. All IoT monitoring data and business data generated during the product flow are associated with this number, thus achieving product traceability. At the same time, due to the front-to-back series overlap of the cold chain business sub-processes, the entire cold chain monitoring is seamlessly connected.

[0148] 12. The feasibility of introducing deep learning models.

[0149] Challenges of deep learning models. Deep learning has the ability to solve complex nonlinear problems. There are many factors that affect cold chain warehousing, including demand, supply, logistics, changes in market share, etc. The impact of these factors is highly nonlinear in nature, and deep learning realizes nonlinear transformation from input to output, which is one of the important reasons why deep learning has made breakthroughs in many complex problems. Deep learning emphasizes end-to-end learning, that is, without artificially dividing steps or sub-problems, but completely letting the neural network directly learn the mapping from original input to expected output. Compared with the divide-and-conquer strategy, end-to-end learning has the advantage of synergy and is more likely to obtain a better solution globally. Nevertheless, the use of deep learning for demand forecasting also has challenges brought by model complexity. According to machine learning theory, complex models require more training data. The increase in the number of model variables places higher requirements on the amount of data used for training data. The increase in the demand for training data caused by the complexity of deep learning cannot be a reason for us to abandon deep learning models. The training process of machine learning uses models to fit data. However, the purpose of machine learning is not to correctly fit the training data set, but to correctly predict samples that have never appeared in the training set. There are many factors that affect cold chain warehousing and they are highly nonlinear, which means that the model used should have a certain degree of complexity, only in this way can these influencing factors be expressed. The use of complex models will inevitably require an increase in the amount of data used for training. Therefore, the increase in data demand is caused by the complexity of the problem, not by the complexity of the deep learning model.

[0150] Ideas to solve the challenge: The historical data accumulated from the daily business process of the node to be studied is most closely related to the node, so directly using this data to predict the demand of the node is the most effective. However, the amount of business data for a single node is not large, and the accumulation of historical data in a short period of time cannot meet the demand of complex models for the amount of training data. In theory, there is a certain direct or indirect relationship between businesses in different regions on the same market. For example, there are similar consumption patterns between users in different regions. At the same time, the same brand may have similar market influence in different regions. In addition, the connection between entities upstream and downstream of the same supply chain is inherently very close. Therefore, at least in theory, it is feasible to use the correlation implied in the data between different entities to increase the amount of training data.

[0151] Through the cold chain supply chain trusted monitoring method, the deep learning training process has met the requirements for data quantity and data quality. On the one hand, it is reflected in the quantity of data. The traceability system built on blockchain establishes the correlation between the data of the entire supply chain. The data of each node in the supply chain is no longer isolated, and it is more convenient to obtain data. We can increase the amount of training data by introducing data from other nodes in the supply chain. This will greatly increase the effect of training. On the other hand, it is reflected in the quality of data. The introduction of blockchain technology solves the problem of data credibility and ensures the quality of training data.

[0152] 13. Experimental analysis.

[0153] Experimental setup: In order to verify the performance of the cold chain storage demand prediction scheme, we conducted an experimental evaluation based on real data. The dataset used is the business data of Company T. Company T is mainly engaged in medical-related cold chain logistics services. The time span of the dataset is from January 1, 2016 to December 31, 2019, a total of three years. The minimum time granularity of the data is daily. The benchmark schemes include ARIMA, PSO-ELM, and XGBoost. ARIMA is a time series prediction model that combines moving average and autoregression. PSO-ELM is a hybrid machine learning model based on particle swarm optimization (PSO) and extreme learning machine (ELM). XGBoost is a scheme based on boosted trees. The experimental evaluation indicators include mean percentage error (MAPE) and root mean square error (RMSE), which are defined as follows.

[0154]

[0155] Where N is the total number of samples. and i are the predicted value and the actual value respectively.

[0156] The experimental results show that the MAPE of the three schemes of ARIMA, PSO-ELM and XGBoost are 0.2531, 0.2103 and 0.1986 respectively, and the MAPE of the scheme of the present invention is 0.1752; the RMSE of the three schemes of ARIMA, PSO-ELM and XGBoost are 12.352, 11.212 and 10.926 respectively, and the RMSE of the present invention is 9.782. According to the performance comparison results between the proposed scheme and the benchmark scheme. In this experiment, the proposed scheme has the best performance. ARIMA performs the worst among these schemes. ARIMA relies entirely on historical demand values ​​and does not consider nonlinearity and spatial correlation. Compared with ARIMA, the performance of PSO-ELM and XGBoost is significantly improved. PSO-ELM adopts heuristic and population-based optimization techniques. XGBoost considers contextual characteristics. Relatively speaking, the performance of XGBoost is better than the other two benchmark schemes. However, the performance of XGBoost is still worse than that of the scheme proposed in this experiment.

[0157] Performance under different parameters: We study the impact of different parameter settings on the performance of the solution by adjusting the convolution kernel size, the number of convolution layers, and the LSTM sequence length. First, we study the impact of the convolution kernel size and the number of convolution layers on the performance of the solution. The experimental results are shown in Figure 2. Fig. 9 (a). The experimental results show that these two parameters have a significant impact on the performance of the scheme. In this experiment, when the size of the convolution kernel is 7 times 7 and the number of convolution layers is 3, the scheme achieves the best performance. Then, we study the impact of the LSTM sequence length on the performance of the scheme. The experimental results are shown in Fig. 9 (b) As shown. The experimental results show that the length of the LSTM sequence also has a significant impact on the performance of the scheme. In this experiment, when the length is 5 days, our method achieves the best performance. If the "length" value continues to increase, the performance change is relatively small.

[0158] Performance under different structures: We compare the proposed scheme with the scheme without the graph embedding module. The MAPE and RMSE of the scheme without the graph embedding module are 0.1835 and 10.093 respectively; for the scheme with the graph embedding module, the MAPE and RMSE of the scheme are 0.1752 and 9.782 respectively. Without the embedding module, although it can still extract features from both temporal and spatial dimensions, the performance is significantly degraded. We believe that spatial correlation exists not only between adjacent regions but also between distant regions. The similarity between regions extracted by the graph embedding module goes beyond the limitation of regions. It can effectively explore such correlation in distant regions, which has a significant impact on the performance of the proposed scheme.

Claims

1. A cold chain supply chain trustworthy monitoring method, characterized in that The architecture of the trusted monitoring system adopted by this method consists of a business layer, a physical layer, and a data persistence layer; the business layer includes business entities such as manufacturers, wholesalers, distributors, consumers, and government regulatory agencies; the business layer is used to abstract and map with real business logic; cold chain product entities and related information realize physical flow and information flow in the business interaction process between various entities in the business layer; the physical layer involves production processes, multi-level cold chain warehousing, and multi-level cold chain transportation. These entity process-related objects; in order to ensure the safety of the entire life cycle of cold chain products, the Internet of Things monitoring module is introduced to monitor different entity process objects; The data persistence layer involves blockchain and cloud storage; the business data of the business layer and the IoT monitoring data of the entity layer are stored in the cloud storage platform; The storage address of the data on the cloud storage platform and the hash value of the data are uploaded to the blockchain platform; The cold chain warehouse inventory management method includes the following steps: (1) Data preprocessing: The data preprocessing module obtains the storage addresses of cold chain business data and cold chain product consumption data in the cloud computing platform through the storage addresses in the blockchain records; and verifies the authenticity of the data stored in the cloud computing platform through the hash value in the blockchain; (2) Cold chain warehousing demand prediction: The cold chain warehousing demand prediction model is constructed based on deep neural network technology and consists of three modules: convolution operation module, long short-term memory network operation module and graph embedding operation module. After the original data is preprocessed, the obtained data is input into the convolution operation module in chronological order. At the same time, the original data will be used to construct similar graphs after data preprocessing and input into the graph embedding operation module. The data processed by the convolution layer will be aggregated through the fully connected layer and input into the long short-term memory network operation module. The output of the graph embedding operation module will be processed by the fully connected layer, and then integrated and spliced ​​with the output of the long short-term memory network operation module, and then processed by the fully connected layer to obtain the final output. (3) Uncertainty measurement of cold chain storage demand: The Gaussian model is used to model the distribution of storage demand uncertainty. Based on the aforementioned deep neural network prediction model, the dropout method is used to estimate the uncertainty of cold chain storage demand, and the mean and variance of the Gaussian model are obtained as follows: Where T is the number of sampling times and τ is the model accuracy; (4) Cold chain warehouse management decision: The optimal cold chain warehouse inventory decision is modeled as an expected cost minimization problem; the optimal decision result is obtained as Where φ represents the cumulative distribution function of the uncertain demand y, α is the purchase cost per unit of inventory, β is the various loss values ​​caused by insufficient inventory per unit, and γ is the various cost increases caused by excess inventory per unit; Use the cumulative distribution function F(x; μ, σ) of the Gaussian random model to replace the aforementioned cumulative distribution function φ: The mean and variance of the Gaussian model obtained in step (3) are used as uncertainty measurements and assigned to μ and σ in the cumulative distribution function respectively. 2 .

2. The cold chain supply chain trusted monitoring method according to claim 1 is characterized in that: The cold chain supply chain trustworthy monitoring method includes the following processes: (1) Cold chain product registration: Manufacturers submit relevant supporting data for cold chain product registration to the cloud storage platform; Manufacturers submit cold chain product registration requests to the blockchain platform; (2) Cold chain supply chain business process: The cold chain supply chain process is composed of a series of cold chain supply chain sub-processes connected in series; each cold chain supply chain sub-process consists of three types of business entities: supplier, cold chain logistics transporter and consignee; the supplier, cold chain logistics transporter and consignee respectively initiate data storage requests to the cloud storage platform; the supplier, cold chain logistics transporter and consignee respectively submit the storage address of the business data on the cloud storage platform together with the hash value of the data to the blockchain platform; (3) Cold chain supply chain monitoring process: The cold chain supply chain monitoring is implemented by the Internet of Things module; Cold chain supply chain monitoring is divided into three categories: production process monitoring, cold chain storage monitoring, and cold chain logistics and transportation monitoring; the IoT monitoring module is responsible for monitoring data collection and storing the monitoring data on the cloud storage platform; The storage address of the monitoring data on the cloud storage platform together with the hash value of the data will be submitted to the blockchain platform; (4) Integration of cold chain supply chain business processes and cold chain supply chain monitoring processes: When cold chain products are registered, a mapping relationship between the cold chain product number and the production process monitoring data is established; when the cold chain storage module performs cold chain product entry registration, a mapping relationship between the cold chain product number and cold chain storage monitoring is established; when the cold chain transportation module begins to register cold chain product loading, a mapping relationship between the cold chain product number and cold chain logistics monitoring is established; The specific steps of step (1) for cold chain product registration are as follows: (1.1) The manufacturer uses its own private key to digitally sign the business support data related to cold chain product registration; then the data and the digital signature are stored together in the cloud storage platform; the cloud storage platform will feedback the storage address of the business support data to the manufacturer; (1.2) The manufacturer initiates a product registration request to the blockchain platform and sends its own number, business support data storage address, and data hash value; after the blockchain platform verifies the data, it generates a cold chain product number for the product and writes the received data storage address, data hash value data together with the cold chain product number into the blockchain; The blockchain platform will provide cold chain product numbers to manufacturers.

3. The cold chain supply chain trusted monitoring method according to claim 2 is characterized in that: The specific steps of step (2) are as follows: (2.1) Cold chain supply chain business process: A complete cold chain supply chain consists of a series of cold chain business sub-processes connected in series; each cold chain business sub-process consists of three types of business entities: supplier, cold chain logistics transporter and consignee; the supplier of the current cold chain supply chain sub-process is the consignee of the previous cold chain supply chain sub-process; the supplier of the current cold chain supply chain sub-process is the supplier of the next cold chain supply chain sub-process; (2.2) Supplier business data storage process: The supplier prepares the supplier number and business data, and uses its own private key to digitally sign the prepared data content; then initiates a data storage request to the cloud storage platform; after the cloud storage platform completes the data storage process, it feedbacks the data storage address to the supplier; the cloud storage platform initiates a business record request to the blockchain platform, and writes the supplier number, data storage address, and data hash value information to the blockchain platform; the blockchain platform feedbacks the record address on the blockchain to the supplier; (2.3) Cold chain logistics transporter business data storage process: The transporter prepares the transporter number and business data, and uses its own private key to digitally sign the prepared data content; then initiates a data storage request to the cloud storage platform; after the cloud storage platform completes the data storage process, it feedbacks the data storage address to the transporter; the cloud storage platform initiates a business record request to the blockchain platform, and writes the transporter number, data storage address, and data hash value information to the blockchain platform; the blockchain platform feedbacks the record address on the blockchain to the transporter; (2.4) Consignee's business data storage process: The consignee prepares the consignee number and business data, and uses its own private key to digitally sign the prepared data content; then initiates a data storage request to the cloud storage platform; after the cloud storage platform completes the data storage process, it feedbacks the data storage address to the consignee; the cloud storage platform initiates a business record request to the blockchain platform, and writes the consignee number, data storage address, and data hash value information to the blockchain platform; the blockchain platform feedbacks the record address on the blockchain to the consignee.

4. The cold chain supply chain trusted monitoring method according to claim 2 is characterized in that: The specific steps of step (3) are as follows: (3.1) Cold chain supply chain monitoring process: The cold chain supply chain monitoring is implemented by the Internet of Things module; the cold chain supply chain monitoring is divided into three types of monitoring: production process monitoring, cold chain warehousing monitoring and cold chain logistics and transportation monitoring; (3.2) IoT monitoring of the production process is organized according to the production process; After the IoT monitoring module completes the regular collection of production process monitoring data, it will send a data storage request to the cloud storage platform, and send the production flow program number, time, data, and related digital signatures corresponding to the IoT monitoring data to the cloud storage platform; the cloud storage platform receives the request, performs data storage operations, stores the relevant data on the cloud storage platform, and feeds back the monitoring data storage address to the IoT monitoring module; (3.3) IoT monitoring in cold chain warehousing is organized according to the storage partition number or storage shelf number. After the IoT monitoring module completes the regular collection of cold chain warehousing monitoring data, it will send a data storage request to the cloud storage platform and send the storage location number, time, data, and related digital signature corresponding to the IoT monitoring data to the cloud storage platform. After receiving the request, the cloud storage platform performs data storage operations, stores the relevant data on the cloud storage platform, and feedbacks the monitoring data storage address to the IoT monitoring module. (3.4) IoT monitoring in cold chain logistics transportation is organized according to container numbers or vehicle numbers; after the IoT monitoring module completes the regular collection of cold chain logistics monitoring data, it will send a data storage request to the cloud storage platform and send the vehicle number, time, data, and related digital signature corresponding to the IoT monitoring data to the cloud storage platform; the cloud storage platform receives the request, performs data storage operations, stores the relevant data on the cloud storage platform, and feedbacks the monitoring data storage address to the IoT monitoring layer.

5. The cold chain supply chain trusted monitoring method according to claim 2 is characterized in that: The specific steps of step (4) are as follows: (4) Integration of cold chain supply chain business process and cold chain supply chain monitoring process: (4.1) Integration of production process monitoring data and business processes: When manufacturers register the aforementioned cold chain products, the IoT monitoring data generated during the production process of the cold chain products will be included as part of the business support data; the data will be submitted to the cloud storage platform; the blockchain record includes the storage address of the data in the cloud storage; when the blockchain platform generates a cold chain product number for the product, a mapping relationship between the cold chain product number and the production process monitoring data is established; various stakeholders can obtain the blockchain platform record of the corresponding record through the cold chain product number, and then obtain the monitoring data in the cloud storage platform through the cloud storage address in the record, and the authenticity of the monitoring data can be verified through the data hash value in the record; (4.2) Integration of cold chain storage monitoring and business processes: When the cold chain storage module registers the entry of cold chain products into the warehouse, it will allocate specific storage locations for the cold chain products in the warehouse, thereby establishing a mapping between the cold chain product number and the storage location number; such data belongs to the business support data of the pharmaceutical cold chain supply chain and needs to be uploaded to the cloud storage module in accordance with the business data storage method; the cold chain storage module uses the storage location number and the cold chain product's in-stock time interval as input data to request the IoT monitoring module to access the IoT monitoring data; the IoT monitoring module will feedback the corresponding monitoring data storage address based on the information provided by the cold chain storage module; the cold chain storage module sends the cold chain product number, IoT monitoring data storage address and data hash value to the blockchain platform and requests that it be written into the blockchain platform; after verifying the data, the blockchain writes the data into the blockchain and feedbacks the on-chain record address to the cold chain storage module; the on-chain record address of the IoT monitoring will be sent to stakeholders as part of the business data; each stakeholder can find the record based on the blockchain record address to obtain the access address and hash value of the monitoring data; (4.3) Integration of cold chain transportation monitoring and business processes: When the cold chain transportation module starts to register the loading of cold chain products, it will register the vehicle number information and cold chain product number information in the cold chain logistics transportation business record, thereby establishing a mapping between the cold chain product number and the vehicle number; this type of data belongs to the business support data of the pharmaceutical cold chain supply chain and needs to be uploaded to the cloud storage module in accordance with the business data storage method; the cold chain transportation module uses the vehicle number and the transportation time interval of the cold chain product as input data to request the IoT monitoring module to access the IoT monitoring data; the IoT monitoring module will feedback the corresponding monitoring data storage address based on the information provided by the cold chain transportation module; the cold chain transportation module will send the cold chain product number, IoT monitoring data storage address and data hash value to the blockchain platform and request that it be written to the blockchain platform; after the blockchain verifies the data, it will write the data to the blockchain and feedback the on-chain record address to the cold chain supply chain module; the on-chain record address of the IoT monitoring will be sent to stakeholders as part of the business data; each stakeholder can find the record based on the blockchain record address to obtain the access address and hash value of the monitoring data.

Images