Intelligent method for temporarily culturing marine products

Abstract

The present invention discloses an intelligent method for temporarily culturing marine products, relates to the field of aquaculture, solves the technical problem of inaccurate water quality prediction in the current aquaculture field, and the steps include: A, applying distributed sensors and Internet of Things technology to realize data collection of environmental parameters, and transmit and store them to a server; B, designing a blockchain data management architecture according to the characteristics of temporarily culturing marine products, and generating an electronic ID card for query for each marine product; C, analyzing the self-changing law of key parameters in water quality parameters and the interaction relationship between them, constructing a recurrent neural network model, and designing corresponding safety early warning technology; D, developing a system management module for user-oriented design, mainly providing user management, authority allocation and system setting functions; E, developing a set of marine product temporary culturing intelligent management cloud platform system, which displays water quality information and environmental information in real time. The present invention has the advantages of high informatization and automation, and high quality of marine products.

Description

Technical Field

[0001] The invention relates to the field of aquaculture, and in particular to an intelligent method for temporarily culturing marine products. Background Art

[0002] In recent years, with the rapid growth of my country's economy and the continuous improvement of people's living standards, people's consumption demand for seafood has gradually developed in the direction of diversification, high-end, personalization and safety. Seafood safety has become an important issue of social concern. The traditional breeding model is relatively extensive, with low breeding efficiency, low degree of automation, unstable output, uncontrollable quality, and many safety hazards. The temporary breeding method of marine delicacies has changed from extensive to information-based, automated, and intelligent. It is imperative to promote the standardization and scale development of the temporary breeding process and improve the quality of temporarily cultured marine delicacies.

[0003] At present, the research on water quality prediction at home and abroad is mainly used in the fields of water environment protection and sewage treatment, while it is relatively rarely used in the field of aquaculture. The reason is that there are many factors affecting the water quality of aquaculture water, the mechanism of action is complex, and the tolerance of aquaculture organisms to water quality changes is limited. It is still in the sporadic research stage, with few long-term continuous historical statistical data, lack of systematic research on aquaculture water quality, and rarely consider the comprehensive effects of multiple factors. Most of the predictions are based on single factors. Summary of the invention

[0004] In order to overcome the shortcomings of the background technology, the present invention provides an intelligent marine treasure temporary culture method with high information and automation levels and high marine treasure quality.

[0005] The technical solution adopted by the present invention is an intelligent method for temporarily culturing marine products, the steps of which include: A. Apply distributed sensors and Internet of Things technology to collect data on temporary storage environment parameters of marine products, and transmit and store them on the server; B. Design a blockchain data management architecture based on the characteristics of temporary marine products. All temporary marine product data is stored and managed in the cloud using a blockchain structure, and an electronic ID card is generated for each marine product. C. Intelligently process the temporary storage environment data collected in step A, analyze the changing rules of key parameters in water quality parameters and their interaction, build a recurrent neural network model suitable for monitoring abnormal temporary storage environment of marine products, and combine historical data with the data collected in step A to study the environmental anomaly detection method of multi-feature fusion, and design and implement corresponding safety warning technology; D. Develop a system management module designed for users, mainly providing user management, authority allocation and system setting functions; E. Develop a cloud platform system for intelligent management of temporary aquaculture of marine products, which can display water quality and environmental information in real time and provide managers with an interface for inputting aquaculture information to record the aquaculture process.

[0006] The key water quality parameters in step A include water temperature, pH value, dissolved oxygen and salinity.

[0007] In step B, QR code (quick response matrix code) is used as the encoding method of the electronic ID card; the management solution of the identity code and its associated data based on blockchain includes four parts: identity code distribution, identity code use, identity code recovery, and identity code resolution.

[0008] The step C comprises: C1. Data preprocessing: During the data collection process, sensor performance degradation, wireless network transmission interference, and water replacement may cause water quality data loss or abnormality. For missing water quality data with a small time interval, linear interpolation can be used to repair it; For missing data with a large time span, data with the same weather type or the same time scale of nearby dates can be used for completion; Since water quality parameters have the characteristics of time series and continuity, the adjacent monitoring data will not change sharply under normal circumstances. Through a large amount of data statistics, if the range of water quality data at a certain moment is beyond 10% of the previous and next values, the data is considered to be wrong and can be smoothed by the mean. C2. Modeling and improving the water quality prediction model. Construct a recurrent neural network to predict water quality based on changes in water quality parameters. In order to avoid the problem of long-term dependency loss of historical data, introduce an LSTM (long short-term memory network) control unit into the recurrent neural network structure, and introduce a genetic algorithm to improve the LSTM model, mainly to optimize the evolution of the connection weight of the LSTM model. C3. Water quality warning: Warning is issued through comprehensive evaluation of the predicted values ​​of water quality parameters; the water quality factors currently used as evaluation criteria include dissolved oxygen, pH value, water temperature and salinity.

[0009] The formula of the linear interpolation method in step C1 is as follows: ; (1) The formula for mean smoothing is as follows: ; (2) in, x k and x k+j They are k Moment and k + j The water quality parameters at each moment are k+ i Parameters of water quality at all times; The formula of the recurrent neural network in step C2 is as follows ; (3) ; (4) in, X t , H t , O t are the values ​​of the input layer, hidden layer, and output layer respectively; W xh , W ℎℎ , W ho are the connection weights from the input layer to the hidden layer, from the previous hidden layer to the current hidden layer, and from the hidden layer to the output layer; b h is the hidden layer bias; b o is the output layer bias; The data processing and transmission process formula of LSTM is as follows: ; (5) ; (6) ; (7) ; (8) ; (9) ; (10) in, i t , f t , o t , c t Indicates the states of the three gate structures and neural units corresponding to time t; W , U , b are model parameters, where W , U are the input loop weights for each type, b is the bias term; k is the number of LSTM neurons; σ is the sigmoid function; e Represents the operation of multiplying the elements at corresponding positions of two vectors; u It is a new feature representation formed by combining the feature representation output at the previous moment and the feature representation of the current input;c An internal neural unit that stores data information; h t That is the state output of the neural unit; In step C3, the evaluation model of each single factor water quality parameter is as follows: (11) or (12) In the formula W It is a single factor early warning index; c i is the predicted value of a water quality factor; c si is the maximum allowable standard value of a water quality factor; when the salinity factor concentration is smaller, the water quality is better, and formula (11) is used to calculate W ; When the factor concentration is higher, the water quality is better (dissolved oxygen), use formula (12) to calculate W ;when W When it is less than 1, the water quality is good. W When it is greater than 1, an alarm will be triggered; when any water quality factor reaches the alarm index, the system background alarm, the alarm device alarm and other methods will remind the manager to handle it, and the category of the water quality factor currently being alarmed will be prompted; W Determined based on my country's fishery breeding standards and actual breeding environment.

[0010] The step D comprises: D1. Establish a user information database and assign a username and password to each user. D2. Users are divided into multiple levels, and different roles have different permissions. D3. Provide functions such as user addition, modification, query, deletion, temporary environment parameter configuration, and system log management.

[0011] In step E, the water quality information is presented in the form of a chart, which can display the current real-time information and the historical change curves of different water quality parameters and environmental parameters as needed; The intelligent management cloud platform system for temporary storage of marine treasures is divided into local workstations and remote workstations. The online monitoring equipment is connected to the local workstation through the data acquisition platform, and displays and records in real time. The local workstation performs intelligent analysis and accurate prediction of sensor data based on the optimal growth environment required for temporary storage of marine treasures, and realizes functions such as abnormal alarm. Safety warnings can remind managers and on-duty personnel through various means such as e-mail, system background, and alarm devices. At the same time, the local workstation uploads the collected data and prediction results to the website to build a WEB-based remote workstation. Factory managers can monitor the temporary storage factory in real time through the remote workstation.

[0012] In the step A, multiple water quality monitoring points and a water quality monitoring and control system are set up, and each monitoring point is equipped with a set of sensors, including dissolved oxygen, pH / water temperature and salinity sensors, which can realize real-time collection of water quality data of the monitoring point and upload it to the water quality monitoring and control system; the water quality monitoring and control system uniformly integrates and transmits the uploaded water quality information, on the one hand, couples with 485 bus technology to communicate with the LED display screen of the breeding factory to realize real-time dynamic display of water quality data of multiple monitoring points; on the other hand, it uses 4G communication to communicate with the cloud server, and uploads the current data to the cloud server in real time for processing.

[0013] The beneficial effects of the present invention are as follows: the present technical solution collects and stores information of all links in the breeding process, generates a unique identity code for each temporarily-raised aquatic product and manages it with blockchain technology, so that the entire breeding process can be traced. The breeding environment parameters and breeding water quality parameter data are collected in a distributed manner, and transmitted and stored to the server. The collected water quality parameters are sorted out, and a water quality prediction model is established through corresponding algorithm training and learning to realize a water quality early warning function. An identity code is generated for each marine treasure, and the temporary breeding process information and the basic information of the marine treasure are stored in the identity code, providing system support for managers to manage the quality and safety of marine treasures through identity codes. The system management module is responsible for user management, authority setting, and system setting functions. The marine treasure temporary breeding intelligent management cloud platform system displays water quality information and environmental information in real time, and provides managers with a breeding information input interface to record the breeding process. It has the advantages of accurate water quality prediction, high degree of informatization and automation, and high quality of marine treasures. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The system block diagram of the intelligent temporary storage method for marine products according to the embodiment of the present invention.

[0015] Figure 2 This is the RNN network structure diagram.

[0016] Figure 3 This is the LSTM structure diagram.

[0017] Figure 4 Algorithm flow chart for evolving the connection weights of the LSTM model using a genetic algorithm.

[0018] Figure 5 It is the life cycle of the identity code management solution.

[0019] Figure 6 It is the process of identity code parsing and recycling.

[0020] Figure 7 This is the system framework diagram of the water quality environment online monitoring system.

[0021] Figure 8Real-time charts for water quality monitoring. DETAILED DESCRIPTION

[0022] The embodiments of the present invention are further described below in conjunction with the accompanying drawings: As shown in the figure, an intelligent method for temporarily culturing marine products comprises the following steps: A. Apply distributed sensors and Internet of Things technology to collect data on temporary storage environment parameters of marine products, and transmit and store them on the server; B. Design a blockchain data management architecture based on the characteristics of temporary marine products. All temporary marine product data is stored and managed in the cloud using a blockchain structure, and an electronic ID card is generated for each marine product. C. Intelligently process the temporary storage environment data collected in step A, analyze the changing rules of key parameters in water quality parameters and their interaction, build a recurrent neural network model suitable for monitoring abnormal temporary storage environment of marine products, and combine historical data with the data collected in step A to study the environmental anomaly detection method of multi-feature fusion, and design and implement corresponding safety warning technology; D. Develop a system management module designed for users, mainly providing user management, authority allocation and system setting functions; User management: Establish a user information database and assign a user name and password to each user; when a user logs into the system, he / she needs to provide a user name and password to identify the user and prevent unauthorized and illegal users from logging into the system and posing a security threat to the system; Permission management: Users are divided into multiple levels, and different roles have different permissions. According to different permissions, users can view different levels of content and operations to limit different levels of user permissions and avoid unauthorized system access, ensuring the security of the breeding management system and the confidentiality of data; System settings: provide functions such as user addition, modification, query, deletion, temporary environment parameter configuration, system log management, etc.; the system management module is an application program built on the server side; the application program is divided into two types: B / S (browser / server) architecture and C / S (client / browser) architecture; the B / S architecture is a structure based on the Web database. The client software of this structure is usually a browser, and there is no need to specially develop other clients, which greatly reduces the development difficulty and maintenance cost of the system; therefore, the server-side software of this embodiment is intended to be designed using the B / S architecture.

[0023] E. Develop a cloud platform system for intelligent management of temporary rearing of marine products, which can display water quality and environmental information in real time, and provide managers with an interface for inputting aquaculture information to record the aquaculture process; the real-time display module for water quality information is presented in the form of a chart, which can display both the current real-time information and the historical change curves of different water quality parameters and environmental parameters as needed, so that managers can trace back information; temporary rearing process information is used to record all information during the temporary rearing process, track the entire process of marine products from entering the temporary rearing factory to leaving, and provide a data basis for the quality and safety management of marine products; factory managers use temporary rearing batches and temporary rearing pools as units to record temporary rearing time, operators, operating locations, and feeding management, medication management, inspection management, warehousing and other information during the temporary rearing process.

[0024] This technical solution collects and stores information from all links in the breeding process, generates a unique identity code for each temporarily-raised aquatic product and manages it with blockchain technology, so that the entire breeding process can be traced. The breeding environment parameters and breeding water quality parameter data are collected in a distributed manner, and transmitted and stored to the server. The collected water quality parameters are sorted, and a water quality prediction model is established through corresponding algorithm training and learning to achieve a water quality early warning function; an identity code is generated for each marine treasure, and the temporary breeding process information and the basic information of the marine treasure are stored in the identity code, providing system support for managers to manage the quality and safety of marine treasures through identity codes. The system management module is responsible for user management, authority setting, and system setting functions. The marine treasure temporary breeding intelligent management cloud platform system displays water quality information and environmental information in real time, and provides managers with a breeding information input interface to record the breeding process; it has the advantages of accurate water quality prediction, high degree of informatization and automation, and high quality of marine treasures.

[0025] The key water quality parameters in step A include water temperature, pH value, dissolved oxygen and salinity.

[0026] Water quality detection equipment is divided into data intelligent terminals and water quality parameter sensors. Water quality parameter sensors are detection devices for detecting water quality information of water bodies. They can convert the relevant water quality parameters they sense into electrical signals or other forms of information output according to rules to meet the requirements of water quality parameter information processing, display, recording and control; the water quality parameter sensor is divided into three probes, which detect dissolved oxygen and temperature, salinity and pH value respectively; the data intelligent terminal is directly connected to the water quality parameter sensor, which can display the values ​​of each sensor in real time, save data locally and transmit the data collected by the sensor to the temporary storage management platform of marine products through a wireless network in real time to ensure the timeliness and effectiveness of the data; the data intelligent terminal and the temporary storage management platform of marine products transmit data through a wireless LAN; compared with wired networks, wireless LAN communications are easily affected by multiple factors such as multipath fading and co-frequency interference, resulting in reduced reliability of wireless transmission, large delay, high packet loss rate, and large retransmission overhead.

[0027] In step B, QR code (quick response matrix code) is used as the encoding method of the electronic ID card; the management solution of the identity code and its associated data based on blockchain includes four parts: identity code distribution, identity code use, identity code recovery, and identity code resolution.

[0028] In the temporary holding factory, each seafood in temporary holding is given a unique identity code, which is used to save the basic information of the temporary holding seafood, water quality information during the temporary holding process, environmental parameters, and related information such as feeding and medication management. Taking into account the operation mode, operating cost and safety, the quick response code (QR code) is used as the identity code encoding method, and a blockchain-based management solution for seafood identity codes and their associated data is adopted.

[0029] The entities in the network, the marine treasures information recording and collection module, the local server and the authentication server all participate in the operation of the blockchain, but they assume different roles respectively; the marine treasures information recording and collection module cannot participate in the maintenance of the blockchain due to its limited computing and storage capabilities and the need to collect information about the surrounding environment. Therefore, the marine treasures information recording and collection module only acts as an ordinary node, which can initiate transactions but cannot package, upload and propagate blocks; the local server conveniently collects information on all marine treasures in the area, and has sufficient computing and storage resources. It can be responsible for transaction packaging, block uploading and other tasks, and therefore participates in the operation of the blockchain as a miner node; the authentication server is responsible for the issuance of certificates and identity code resolution for all marine treasures and local servers. It is trusted by default. The function of the authentication server is extremely important. In order to ensure high responsiveness, the authentication server is responsible for creating the genesis block when the network is initialized, and then participates in the network as an ordinary node, and does not participate in block packaging and uploading and other tasks.

[0030] In blockchain, transaction data needs to be encrypted; here we only discuss the three modules of identity code distribution, identity code parsing and identity code recovery; 1) Identity code distribution Identity code distribution refers to the process of assigning a new identity code to marine treasures; identity code distribution can be divided into third-party entity distribution and self-distribution; third-party entity distribution means that all identity codes of marine treasures are generated by authoritative organizations such as authentication servers, and authentication servers record each identity code used by marine treasures. Third-party entity distribution requires more communication bandwidth. As the number of marine treasures increases, the efficiency drops sharply and the authentication server is burdened; in the self-distribution scheme, a third-party trusted entity distributes secret values ​​to marine treasures. Marine treasures can generate multiple identity codes based on the secret values ​​through an identity code generation algorithm to ensure that the identity codes of marine treasures are associated. Other entities in the network, including authentication servers, can verify the association between marine treasures and identity codes, but cannot know other identity codes used by marine treasures; taking into full consideration the characteristics of the two distribution methods, this embodiment uses a third-party entity to distribute identity codes, and the third-party entity is borne by the local server, reducing the single communication delay and the computing pressure of the authentication server.

[0031] In the scheme, identity code distribution only occurs when the marine treasures are initialized and the marine treasures join the network for the first time; the marine treasure temporary holding factory distributes digital IDs to the marine treasures when the marine treasures enter the temporary holding factory for authentication with the authentication server; initialization process: first, the marine treasure information generation module generates a pair of public and private keys for subsequent communication, and uses the authentication server public key to encrypt the ID and public key information and send it to the authentication server; after receiving the information, the authentication server authenticates the legitimacy of the marine treasures; after the authentication is passed, the certificate, identity code and corresponding public key data signature are distributed to the marine treasures, and the corresponding relationship between the marine treasure ID, certificate, identity code and public key is recorded in the local database, and the initialization process is completed; the marine treasures use the public key hash as the blockchain address and participate in the operation of the blockchain; during initialization, the marine treasures have not yet joined the network and do not participate in the operation of the network, and the latency requirement is relatively low, so third-party distribution can fully meet the latency requirement.

[0032] In the blockchain, identity code distribution transactions are used to represent and record the identity code distribution process. After the marine treasures are initialized, the authentication server constructs the transaction and broadcasts it.

[0033] 2) Identity code analysis and recovery Identity code analysis is required before identity code recovery, so identity code analysis and identity code recovery are explained together; in the seafood identity management module, seafood may have artificial false identity information to maximize its own interests; therefore, the authentication server needs to analyze the identity of seafood, recover the certificate and identity code of seafood, and remove seafood with false identity from the network.

[0034] Identity code resolution is defined as the ability to learn the identity code used by a specific marine product. In the solution, the local server sends the certificate and account address of a fake marine product, and attaches a certificate of fake marine product information, requesting the authentication server to resolve and recover the certificate and identity code of the fake marine product. The authentication server searches the database for information such as the marine product certificate hash and transaction value corresponding to the target identity code, and performs identity code resolution and recovery.

[0035] After the identity code is parsed, the authentication server needs to recycle the fake identity seafood, and the identity code recycling is completed through the identity code recycling list; according to the identity code parsing result, the authentication server puts the certificate and account address of the fake identity seafood into the identity code recycling list, and constructs an identity code recycling transaction to send to the local server to complete the identity code recycling; puts the seafood identity certificate and account address into the identity code recycling list; the identity code put into the identity code pool can be distributed to other seafood when other seafood requests the identity code; after the local server receives the identity code recycling transaction, it puts the transaction into the local transaction pool, puts the certificate and account address into the local identity code recycling list, and broadcasts the identity code recycling transaction in the subsequent mining process, and submits the transaction to the blockchain.

[0036] In addition, when relevant information of marine treasures needs to be queried, the identity code resolution module will also be accessed.

[0037] The adoption of a blockchain-based identity code management solution can, on the one hand, ensure the authenticity and non-tamperability of the identity information of marine treasures, while also ensuring the convenient access and use of identity information, and at the same time ensuring the authenticity of the information associated with the marine treasures with identity codes.

[0038] The step C comprises: C1. Data preprocessing: During the data collection process, sensor performance degradation, wireless network transmission interference, and water changes can cause water quality data loss or abnormality. If the original monitored data is directly used for temporary water quality prediction and early warning, it will seriously affect the accuracy and reliability of the prediction and early warning, and even cause complete misjudgment. Using appropriate data preprocessing methods to repair missing data and reduce non-real "bad data" is of great significance to improving temporary water quality prediction and early warning; For missing water quality data with a small time interval, linear interpolation can be used to repair them; For missing data with a large time span, data with the same weather type or the same time scale of nearby dates can be used for completion; Since water quality parameters have the characteristics of time series and continuity, the adjacent monitoring data will not change sharply under normal circumstances. Through a large amount of data statistics, if the range of water quality data at a certain moment is beyond 10% of the previous and next values, the data is considered to be wrong and can be smoothed by the mean. C2. Modeling and improving water quality prediction models. Constructing a recurrent neural network to predict water quality based on changes in water quality parameters. When there is a problem of long-term dependency loss in historical data, introduce an LSTM (long short-term memory network) control unit into the recurrent neural network structure and introduce a genetic algorithm to improve the LSTM model, mainly to optimize the evolution of the connection weight of the LSTM model. Recurrent Neural Network (RNN) is a feedforward neural network that stores past information in time series data by introducing state variables. The input of its hidden layer consists of two parts, including the output of the previous network layer at this moment and the output of the hidden layer at the previous moment. That is, the nodes of the hidden layer are interconnected, and the current hidden layer output is affected by the previous layer. The memory of the previous information can be achieved through the storage units of the hidden layer. Therefore, recurrent neural networks can process time series data very well. The change of water quality parameters is a time-gradient process, so the application of RNN for water quality prediction is more suitable than other neural network models; RNN can be regarded as a deep feedforward neural network in which all layers share the same weights; The basic RNN has a weak memory capacity. When the amount of calculation increases and the model training time increases, the gradient is more likely to decay or explode.

[0039] Long Short Term Memory (LSTM) is an extension of RNN. Its main purpose is to deal with the problem of long-term dependency loss in historical data. LSTM introduces an LSTM control unit into the ordinary RNN structure. The control unit contains three gate structures: input gates, forget gates, and output gates. The threshold mechanism solves the problem of error signal disappearance.

[0040] The LSTM network model uses memory neural units to control how to discard and update neural unit information, and can introduce long-distance dependency information in sequence data into the learning of the model; this embodiment uses the LSTM network model for aquaculture water quality prediction; in order to improve the accuracy and speed of the prediction model, we introduce a genetic algorithm to improve the LSTM model; the genetic algorithm is an effective, learning, and random learning process that can synchronously evolve the weights of forward and feedback nodes; its search has a global nature, so it is easy to obtain a global optimal solution or a suboptimal solution with good performance; The operation of a genetic algorithm consists of three basic operators: selection, crossover, and mutation; (1) Selection operation: selection is to eliminate some individuals in the old population and select the best individuals to form a new population, which is also called replication or reproduction. This operation will not produce new individuals, but will only replicate the best individuals, thereby improving the average fitness of the population. (2) Crossover operation: the individuals undergoing crossover are divided into parents and offspring. The individuals selected from the population are called parents. The parents are randomly paired, and the genes of the two individuals are exchanged according to a predetermined crossover method and crossover probability. The resulting individuals are called offspring. (3) Mutation operation: Mutation is to replace the gene loci of individuals in the population with their opposing genes according to the set mutation rate, so as to obtain new individuals. The genetic algorithm can optimize the following three aspects of the LSTM model: the evolution of connection weights, the evolution of network structure, and the evolution of learning rules. This embodiment uses a genetic algorithm to optimize the evolution of the connection weights of the LSTM model.

[0041] C3. Water quality warning: Early warning is issued through comprehensive evaluation of the predicted values ​​of water quality parameters. It can timely warn of water quality that may deteriorate, so that managers can take timely measures to prevent and reduce the impact of water quality deterioration on aquaculture activities. The water quality factors currently used as evaluation criteria include dissolved oxygen, pH value, water temperature and salinity.

[0042] The formula of the linear interpolation method in step C1 is as follows: ; (1) The formula for mean smoothing is as follows: ; (2) in, x k and x k+j They are k Moment and k + j The water quality parameters at each moment are k + i Parameters of water quality at all times; The formula of the recurrent neural network in step C2 is as follows ; (3) ; (4) in, X t , H t , O t are the values ​​of the input layer, hidden layer, and output layer respectively; W xh , W ℎℎ ,W ho are the connection weights from the input layer to the hidden layer, from the previous hidden layer to the current hidden layer, and from the hidden layer to the output layer; b h is the hidden layer bias; b o is the output layer bias; The data processing and transmission process formula of LSTM is as follows: ; (5) ; (6) ; (7) ; (8) ; (9) ; (10) in, i t , f t , o t , c t Indicates the states of the three gate structures and neural units corresponding to time t; W , U , b are model parameters, where W , U are the input loop weights for each type, b is the bias term; k is the number of LSTM neurons; σ is the sigmoid function; e Represents the operation of multiplying the elements at corresponding positions of two vectors; u It is a new feature representation formed by combining the feature representation output at the previous moment and the feature representation of the current input; c An internal neural unit that stores data information; h t That is the state output of the neural unit; In step C3, the evaluation model of each single factor water quality parameter is as follows: (11) or (12) In the formula W It is a single factor early warning index; c i is the predicted value of a water quality factor; c siis the maximum allowable standard value of a water quality factor; when the salinity factor concentration is smaller, the water quality is better, and formula (11) is used to calculate W ; When the factor concentration is higher, the water quality is better (dissolved oxygen), use formula (12) to calculate W ;when W When it is less than 1, the water quality is good. W When it is greater than 1, an alarm will be triggered; when any water quality factor reaches the alarm index, the system background alarm, the alarm device alarm and other methods will remind the manager to handle it, and the category of the water quality factor currently being alarmed will be prompted; W Determined based on my country's fishery breeding standards and actual breeding environment.

[0043] The step D comprises: D1. Establish a user information database and assign a username and password to each user. D2. Users are divided into multiple levels, and different roles have different permissions. D3. Provide functions such as user addition, modification, query, deletion, temporary environment parameter configuration, and system log management.

[0044] In step E, the water quality information is presented in the form of a chart, which can display the current real-time information, and can also display the historical change curves of different water quality parameters and environmental parameters as needed, so as to facilitate the manager to trace back the information; The intelligent management cloud platform system for temporary storage of marine treasures is divided into local workstations and remote workstations. The online monitoring equipment is connected to the local workstation through the data acquisition platform, and displays and records in real time. The local workstation performs intelligent analysis and accurate prediction of sensor data based on the optimal growth environment required for temporary storage of marine treasures, and realizes functions such as abnormal alarm. Safety warnings can remind managers and on-duty personnel through various means such as e-mail, system background, and alarm devices. At the same time, the local workstation uploads the collected data and prediction results to the website to build a WEB-based remote workstation. Factory managers can monitor the temporary storage factory in real time through the remote workstation.

[0045] During the temporary storage management process, a quick response QR code is assigned to each seafood product in the temporary storage management platform as an electronic ID number. The data collected by the sensor and the temporary storage process information recorded in the system by the technicians in the temporary storage factory are automatically associated and saved with the seafood product through the identity code. Consumers can scan the QR code with their mobile phones to query all the information about the temporary storage process of the seafood product, realizing the full traceability of the temporary storage process of the seafood product.

[0046] exist Figure 7 In the online monitoring system of water quality environment for king crab, it is composed of three parts: sensor monitoring system, cloud server and local / remote workstation; Sensor monitoring system: There are 50 temporary holding pools in the king crab breeding base, of which 8 are customs-supervised pools with independent recirculating aquaculture systems; the other 42 temporary holding pools are divided into 6 groups, each of which is equipped with a recirculating aquaculture system; in order to achieve full coverage of water quality monitoring for these 42 temporary holding pools, 6 water quality monitoring points and a water quality monitoring and control system are set up in this project; each monitoring point is equipped with a set of sensors, including dissolved oxygen, pH / water temperature and salinity sensors, which can realize real-time collection of water quality data at the monitoring point and upload it to the water quality monitoring and control system; the water quality monitoring and control system integrates and transmits the uploaded water quality information in a unified manner. On the one hand, it couples the 485 bus technology to communicate with the LED display screen of the breeding factory to realize real-time dynamic display of water quality data at the 6 monitoring points; on the other hand, it uses 4G communication to communicate with the cloud server and uploads the current data to the cloud server in real time for processing.

[0047] Cloud Server: The cloud server is built on Alibaba Cloud and runs on the Baota system, on which a MySQL database and a Web application based on a B / S architecture are built; the MySQL database is used to save and record water quality information; the Web application is built based on JavaScript and mainly includes a user management module, a water quality data receiving and display module, and a data storage and management module: The user management module is responsible for user management and permission management; the user management submodule establishes a user information database and assigns a user name and password to each user; when a user logs into the system, he or she needs to provide a user name and password to identify the user and prevent unauthorized and illegal users from logging into the system and posing a security threat to the system; the permission management submodule divides users into multiple levels, and different roles are given different permissions; the Web application limits user operations at different levels through permission management to avoid unauthorized system access and ensure the security of the breeding management system and the confidentiality of data.

[0048] The water quality data receiving and display module is responsible for establishing communication with the water quality monitoring and control system, receiving water quality information in real time, and displaying the water quality information in the form of charts; it can display not only the current real-time information, but also the historical change curves of different water quality sensors and different temporary holding ponds as needed, so as to facilitate managers to trace back information.

[0049] The data storage and management module is responsible for interacting with the MySQL database to realize data storage, query and management; this module encrypts and decrypts the stored water quality information to prevent malicious access and tampering of data by external programs, thereby ensuring data security and privacy.

[0050] Local Workstation / Remote Workstation: The local workstation / remote workstation accesses the Web application located on the cloud server through the registered domain name URL to view and display the water quality information of the temporary holding plant in real time; the local workstation is a local computer located in the factory management office, which is convenient for the factory administrator to view the water quality information on site in real time and promptly detect water quality abnormalities and deal with them; the remote workstation can be any device that can access the Internet, such as a mobile phone, computer, tablet, etc. The factory administrator can use the remote workstation to access the king crab water quality environment online monitoring system and view the current water quality parameters of the factory conveniently and quickly no matter where he is, regardless of location and time.

[0051] Technical personnel should note that although the present invention has been described according to the above specific implementation methods, the inventive concept of the present invention is not limited to this invention, and any modification using the inventive concept will be included in the scope of protection of this patent.

Claims

1. An intelligent method for temporarily culturing marine products, characterized in that: The steps include: A. Apply distributed sensors and Internet of Things technology to collect data on temporary storage environment parameters of marine products, and transmit and store them on the server; B. Design a blockchain data management architecture based on the characteristics of temporary marine products. All temporary marine product data is stored and managed in the cloud using a blockchain structure, and an electronic ID card is generated for each marine product. C. Intelligently process the temporary storage environment data collected in step A, analyze the changing rules of key parameters in water quality parameters and their interaction, build a recurrent neural network model suitable for monitoring abnormal temporary storage environment of marine products, and combine historical data with the data collected in step A to study the environmental anomaly detection method of multi-feature fusion, and design and implement corresponding safety warning technology; D. Develop a system management module designed for users, mainly providing user management, authority allocation and system setting functions; E. Develop a cloud platform system for intelligent management of temporary aquaculture of marine products, which can display water quality and environmental information in real time and provide managers with an input interface for aquaculture information to record the aquaculture process.

2. The intelligent method for temporarily culturing marine products according to claim 1, characterized in that: The key water quality parameters in step A include water temperature, pH value, dissolved oxygen and salinity.

3. The intelligent method for temporarily culturing marine products according to claim 1, characterized in that: In step B, QR code (quick response matrix code) is used as the encoding method of the electronic ID card; the management solution of the identity code and its associated data based on blockchain includes four parts: identity code distribution, identity code use, identity code recovery, and identity code resolution.

4. The intelligent method for temporarily culturing marine products according to claim 1, characterized in that: The step C comprises: C1. Data preprocessing: During the data collection process, sensor performance degradation, wireless network transmission interference, and water replacement may cause water quality data loss or abnormality. For missing water quality data with a small time interval, linear interpolation can be used to repair it; For missing data with a large time span, data with the same weather type or the same time scale of nearby dates can be used for completion; Since water quality parameters have the characteristics of time series and continuity, the adjacent monitoring data will not change sharply under normal circumstances. Through a large amount of data statistics, if the range of water quality data at a certain moment is beyond 10% of the previous and next values, the data is considered to be wrong and can be smoothed by the mean. C2. Modeling and improving the water quality prediction model. Construct a recurrent neural network to predict water quality based on changes in water quality parameters. In order to avoid the problem of long-term dependency loss of historical data, introduce an LSTM (long short-term memory network) control unit into the recurrent neural network structure, and introduce a genetic algorithm to improve the LSTM model, mainly to optimize the evolution of the connection weight of the LSTM model. C3. Water quality warning: Warning is issued through comprehensive evaluation of the predicted values ​​of water quality parameters; the water quality factors currently used as evaluation criteria include dissolved oxygen, pH value, water temperature and salinity.

5. The intelligent method for temporarily culturing marine products according to claim 4, characterized in that: The formula of the linear interpolation method in step C1 is as follows: ；（1） The formula for mean smoothing is as follows: ；（2） in, x k and x k+j They are k Moment and k + j The water quality parameters at each moment are k + i Parameters of water quality at all times; The formula of the recurrent neural network in step C2 is as follows ；（3） ；（4） in, X t , H t , O t are the values ​​of the input layer, hidden layer, and output layer respectively; W xh , W ℎℎ , W ho are the connection weights from the input layer to the hidden layer, from the previous hidden layer to the current hidden layer, and from the hidden layer to the output layer; b h is the hidden layer bias; b o is the output layer bias; The data processing and transmission process formula of LSTM is as follows: ；（5） ；（6） ；（7） ；（8） ；（9） ；（10） in, i t , f t , o t , c t Indicates the states of the three gate structures and neural units corresponding to time t; W , U , b are model parameters, where W , U are the input loop weights for each type, b is the bias term; k is the number of LSTM neurons; σ is the sigmoid function; e Represents the operation of multiplying the elements at corresponding positions of two vectors; u It is a new feature representation formed by combining the feature representation output at the previous moment and the feature representation of the current input; c An internal neural unit that stores data information; h t That is the state output of the neural unit; In step C3, the evaluation model of each single factor water quality parameter is as follows: （11） or (12) In the formula W It is a single factor early warning index; c i is the predicted value of a water quality factor; c si is the maximum allowable standard value of a water quality factor; when the salinity factor concentration is smaller, the water quality is better, and formula (11) is used to calculate W ; When the factor concentration is higher, the water quality is better (dissolved oxygen), use formula (12) to calculate W ;when W When it is less than 1, the water quality is good. W When it is greater than 1, an alarm will be triggered; when any water quality factor reaches the alarm index, the system background alarm, the alarm device alarm and other methods will remind the manager to handle it, and the category of the water quality factor currently being alarmed will be prompted; W Determined based on my country's fishery breeding standards and actual breeding environment.

6. The intelligent method for temporarily culturing marine products according to claim 1, characterized in that: The step D comprises: D1. Establish a user information database and assign a username and password to each user. D2. Users are divided into multiple levels, and different roles have different permissions. D3. Provide functions such as user addition, modification, query, deletion, temporary environment parameter configuration, and system log management.

7. The intelligent method for temporarily culturing marine products according to claim 1, characterized in that: In step E, the water quality information is presented in the form of a chart, which can display the current real-time information and the historical change curves of different water quality parameters and environmental parameters as needed; The intelligent management cloud platform system for temporary storage of marine treasures is divided into local workstations and remote workstations. The online monitoring equipment is connected to the local workstation through the data acquisition platform, and displays and records in real time. The local workstation performs intelligent analysis and accurate prediction of sensor data based on the optimal growth environment required for temporary storage of marine treasures, and realizes functions such as abnormal alarm. Safety warnings can remind managers and on-duty personnel through various means such as e-mail, system background, and alarm devices. At the same time, the local workstation uploads the collected data and prediction results to the website to build a WEB-based remote workstation. Factory managers can monitor the temporary storage factory in real time through the remote workstation.

8. The intelligent method for temporarily culturing marine products according to claim 1, characterized in that: In the step A, multiple water quality monitoring points and a water quality monitoring and control system are set up, and each monitoring point is equipped with a set of sensors, including dissolved oxygen, pH / water temperature and salinity sensors, which can realize real-time collection of water quality data of the monitoring point and upload it to the water quality monitoring and control system; the water quality monitoring and control system uniformly integrates and transmits the uploaded water quality information, on the one hand, couples with 485 bus technology to communicate with the LED display screen of the breeding factory to realize real-time dynamic display of water quality data of multiple monitoring points; on the other hand, it uses 4G communication to communicate with the cloud server, and uploads the current data to the cloud server in real time for processing.

Images