Remote intelligent regulation and control method and system for marine fish fry breeding environment

Abstract

The invention relates to the technical field of intelligent regulation and control, and discloses a remote intelligent regulation and control method and system for a marine fish fry breeding environment, and the method comprises the steps: transmitting environment monitoring data collected by a multi-parameter sensor array in a marine fish fry breeding pool to a cloud-based remote monitoring platform through an Internet of Things channel; eliminating noisy points and abnormal values in the environment monitoring data in the cloud-based remote monitoring platform to obtain a standardized data set of the marine fish nursery pond; on the basis of the time sequence characteristic data extracted in the standardized data set, predicting the environmental parameter change trend of the marine fish nursery pond to obtain environmental change data of the marine fish nursery pond; performing multi-objective optimization on control parameters of the marine fish nursery pond according to the reservoir environment change data to obtain optimal regulation and control parameters of the marine fish nursery pond; transmitting the optimal regulation and control parameters to a local actuator of the marine fish nursery pond through a cloud-based remote monitoring platform, and adjusting the pond body environment of the marine fish nursery pond based on the optimal regulation and control parameters; the efficiency of remote regulation and control of the seedling raising environment can be improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, and in particular to a remote intelligent control method and system for marine fish fry breeding environment. Background Technology

[0002] In the process of raising marine fish larvae, precise and real-time control of key environmental parameters such as water temperature, dissolved oxygen, and salinity is a core prerequisite for ensuring the survival rate and growth quality of larvae. Current technologies for monitoring the marine fish larvae environment mostly rely on scattered, single-function sensors, resulting in limited data collection coverage and insufficient correlation. Furthermore, data transmission often employs traditional short-range communication methods, which suffer from short transmission distances, weak anti-interference capabilities, and high data loss rates. This makes it difficult to transmit multi-parameter environmental data to remote management terminals in real time and completely, leading to staff being unable to promptly grasp the dynamic environmental conditions of each larvae pond, and resulting in reactive and delayed decision-making.

[0003] Furthermore, existing technologies lack efficient preprocessing mechanisms for collected environmental data, failing to effectively eliminate noise and outliers. This easily leads to subsequent environmental trend analysis being based on distorted data, resulting in low prediction accuracy. In determining control parameters, settings often revolve around a single objective (such as maintaining stable water temperature), neglecting to consider multiple objectives such as seedling efficiency, energy costs, and water quality balance. This limits the practicality of control schemes, either failing to meet the comprehensive environmental requirements for seedling growth or causing energy waste. Simultaneously, existing control methods largely rely on local manual operation, lacking remote automated control capabilities. When facing multi-pond management scenarios in large-scale seedling bases, the control response speed is slow, operational consistency is poor, and overall control efficiency is difficult to improve. Summary of the Invention

[0004] This invention provides a remote intelligent control method and system for marine fish fry rearing environment to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a remote intelligent control method for marine fish fry rearing environment, comprising:

[0006] S1. Transmit the environmental monitoring data collected by the multi-parameter sensor array in the marine fish breeding pond to the cloud-based remote monitoring platform through the Internet of Things channel;

[0007] S2. Eliminate noise and outliers in the environmental monitoring data in the cloud-based remote monitoring platform to obtain a standardized dataset of the marine fish breeding pond.

[0008] S3. Based on the time-series feature data extracted from the standardized dataset, predict the changing trend of environmental parameters in the marine fish broodstock pond to obtain environmental change data of the marine fish broodstock pond.

[0009] S4. Based on the environmental change data described in the database, perform multi-objective optimization on the control parameters of the marine fish broodstock pond to obtain the optimal control parameters of the marine fish broodstock pond;

[0010] S5. The optimal control parameters are transmitted to the local actuator of the marine fish fry pond through the cloud-based remote monitoring platform, and the pond environment of the marine fish fry pond is adjusted based on the optimal control parameters.

[0011] In a preferred embodiment, transmitting the environmental monitoring data collected by the multi-parameter sensor array in the marine fish nursery pond to the cloud-based remote monitoring platform via an Internet of Things (IoT) channel includes:

[0012] Add data headers and verification information to the environmental monitoring data collected by the multi-parameter sensor array in the marine fish breeding pond to obtain the transmission data packet of the marine fish breeding pond;

[0013] The data packets are sent to the cloud-based remote monitoring platform via the Internet of Things (IoT) channel.

[0014] The cloud-based remote monitoring platform verifies the received data packets and restores the environmental monitoring data of the marine fish fry pond.

[0015] In a preferred embodiment, the step of eliminating noise and outliers in the environmental monitoring data on the cloud-based remote monitoring platform to obtain a standardized dataset for the marine fish fry pond includes:

[0016] Based on preset thresholds, missing and abnormal data points in the environmental monitoring data are identified.

[0017] Remove the abnormal data points and perform interpolation on the missing data points to obtain the environmental data of the marine fish broodstock pond after cleaning;

[0018] The environmental data after cleaning is scaled to a preset numerical range to obtain a standardized dataset for the marine fish fry breeding pond.

[0019] In a preferred embodiment, the time-series feature data extracted from the standardized dataset includes:

[0020] Arrange the standardized dataset in chronological order to obtain the time series data of the standardized dataset;

[0021] In a sliding window, the time series data is segmented to obtain a data window of the standardized dataset;

[0022] Statistical features are extracted from the data window to obtain the initial feature set of the marine fish fry breeding pond;

[0023] By retaining the features in the initial feature set that are highly correlated with environmental changes, the time-series feature data of the marine fish broodstock pond are obtained.

[0024] In a preferred embodiment, predicting the changing trends of environmental parameters in the marine fish broodstock pond to obtain environmental change data for the marine fish broodstock pond includes:

[0025] The time-series feature data is divided into a training set and a validation set;

[0026] The pre-constructed long short-term memory neural network model is optimized based on the training set, and the prediction accuracy of the optimized long short-term memory neural network model is evaluated through the validation set.

[0027] The standardized dataset is input into the long short-term memory neural network model with the required prediction accuracy to obtain the environmental change data of the marine fish fry pond.

[0028] In a preferred embodiment, optimizing the pre-constructed long short-term memory neural network model based on the training set, and simultaneously evaluating the prediction accuracy of the optimized long short-term memory neural network model using the validation set, includes:

[0029] The training set is input into the long short-term memory neural network model for forward propagation to obtain the training prediction results of the training set;

[0030] The model loss of the long short-term memory neural network model is calculated based on the difference between the training prediction results and the true values.

[0031] The parameters of the long short-term memory neural network model are optimized based on the model loss.

[0032] The validation set is input into the optimized long short-term memory neural network model to obtain the validation prediction results of the validation set;

[0033] The prediction accuracy of the optimized long short-term memory neural network model is evaluated based on the error between the verification prediction results and the corresponding true values.

[0034] The hyperparameters of the long short-term memory neural network model are adjusted based on the predicted accuracy.

[0035] In a preferred embodiment, the model loss is calculated using the following formula:

[0036] ;

[0037] In the formula, The loss of the model, The number of samples in the training set. Let be the ordinal number of the sample in the training set. For the first The true value of each sample, For the long short-term memory neural network model, the first... The training prediction results for each sample.

[0038] In a preferred embodiment, the step of performing multi-objective optimization of the control parameters of the marine fish broodstock pond based on the environmental change data in the database to obtain the optimal control parameters of the marine fish broodstock pond includes:

[0039] Based on the environmental change data, the optimization objectives of the marine fish fry rearing pond are determined;

[0040] Set constraints on the control parameters in the marine fish broodstock pond;

[0041] Under the constraints, evaluate the degree to which all combinations of control parameters in the marine fish broodstock pond meet the optimization objective;

[0042] The parameter combination that achieves the highest degree of satisfaction is selected as the optimal control parameter for the marine fish broodstock pond.

[0043] In a preferred embodiment, the step of transmitting the optimal control parameters to the local actuator of the marine fish fry pond via the cloud-based remote monitoring platform, and adjusting the pond environment based on the optimal control parameters, includes:

[0044] The optimal control parameters are converted into a set of equipment control instructions;

[0045] The device control command set is transmitted to the local actuator of the fish breeding pond through the encrypted communication link of the cloud-based remote monitoring platform.

[0046] The local actuator parses the received set of device control instructions to obtain the device drive signal corresponding to the local actuator;

[0047] Based on the device drive signal, the working status of the heating device, oxygenation device and chemical regulator are synchronously controlled to adjust the pool environment.

[0048] To address the above problems, the present invention also provides a remote intelligent control system for marine fish fry rearing environments, the system comprising:

[0049] The environmental data transmission module is used to transmit environmental monitoring data collected by the multi-parameter sensor array in the marine fish breeding pond to the cloud-based remote monitoring platform via the Internet of Things channel;

[0050] The data cleaning module is used to remove noise and outliers from the environmental monitoring data in the cloud-based remote monitoring platform to obtain a standardized dataset for the marine fish fry pond.

[0051] The water quality change prediction module is used to predict the environmental parameter change trend of the marine fish broodstock pond based on the time-series feature data extracted from the standardized dataset, and obtain the environmental change data of the marine fish broodstock pond.

[0052] The seedling control module is used to perform multi-objective optimization of the control parameters of the marine fish seedling pond based on the environmental change data in the database, so as to obtain the optimal control parameters of the marine fish seedling pond.

[0053] The pool adjustment module is used to transmit the optimal control parameters to the local actuator of the marine fish fry pond through the cloud-based remote monitoring platform, and adjust the pool environment of the marine fish fry pond based on the optimal control parameters.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] 1. This invention enables remote transmission of data collected by a multi-parameter sensor array in a marine fish nursery pond via an IoT channel. Combined with a data header addition and verification mechanism, it ensures that environmental monitoring data is transmitted to a cloud-based remote monitoring platform in real time and completely. At the same time, on the platform, by identifying and removing abnormal data points, interpolating missing data, and standardizing the data, noise and bias in the data are effectively eliminated, resulting in a high-quality standardized dataset. This provides accurate and reliable data support for subsequent environmental parameter analysis, significantly improving the accuracy and effectiveness of environmental monitoring.

[0056] 2. This invention extracts time-series feature data based on standardized datasets, uses an optimized long short-term memory neural network model to predict the changing trends of environmental parameters, and can accurately obtain environmental change data. Then, it determines the optimal control parameters through multi-objective optimization and converts them into equipment control commands, which are transmitted to the local actuator via an encrypted communication link to synchronously control the heating device, oxygenation equipment, and chemical regulator. This enables remote, automated, and precise adjustment of the environment in the seedling pond, significantly improving the response speed and accuracy of environmental control, effectively ensuring the stability of the marine fish seedling environment, and thus improving seedling efficiency and seedling growth quality. Attached Figure Description

[0057] Figure 1 A flowchart illustrating a remote intelligent control method for marine fish fry rearing environment provided in an embodiment of the present invention;

[0058] Figure 2 A functional module diagram of a remote intelligent control system for marine fish fry rearing environment provided in an embodiment of the present invention;

[0059] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0060] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0061] This application provides a method for remote intelligent control of the marine fish fry rearing environment. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for remote intelligent control of the marine fish fry rearing environment can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0062] Reference Figure 1 The diagram shown is a flowchart illustrating a remote intelligent control method for the marine fish fry rearing environment according to an embodiment of the present invention. In this embodiment, the remote intelligent control method for the marine fish fry rearing environment includes:

[0063] S1. Transmit the environmental monitoring data collected by the multi-parameter sensor array in the marine fish breeding pond to the cloud-based remote monitoring platform through the Internet of Things channel;

[0064] In this embodiment of the invention, transmitting the environmental monitoring data collected by the multi-parameter sensor array in the marine fish nursery pond to the cloud-based remote monitoring platform via the Internet of Things channel includes:

[0065] Add data headers and verification information to the environmental monitoring data collected by the multi-parameter sensor array in the marine fish breeding pond to obtain the transmission data packet of the marine fish breeding pond;

[0066] The data packets are sent to the cloud-based remote monitoring platform via the Internet of Things (IoT) channel.

[0067] The cloud-based remote monitoring platform verifies the received data packets and restores the environmental monitoring data of the marine fish fry pond.

[0068] Specifically, the environmental monitoring data collected by the multi-parameter sensor array in the marine fish breeding pond is first determined. A data header containing the unique device identifier of the sensor array, the data acquisition time, and the monitoring parameter type is constructed. The data header is appended to the beginning of the environmental monitoring data. Then, the number of "1"s in the combination of the data header and the environmental monitoring data is counted using a parity check method. When the number is odd, the check information is set to "1" and when it is even, it is set to "0". The check information is appended to the end of the overall data to finally obtain the transmission data packet of the marine fish breeding pond.

[0069] Furthermore, a LoRa IoT channel is selected, and a LoRa IoT channel connection is established between the sensor array data processing module at the marine fish fry pond and the cloud-based remote monitoring platform. After ensuring a stable connection, the data packets from the marine fish fry pond are sent to the cloud-based remote monitoring platform through this channel until the platform reports successful reception.

[0070] Furthermore, in the cloud-based remote monitoring platform, the verification information in the transmitted data packets is extracted. The number of "1"s in the data packets excluding the verification information is recounted, and the required verification information is determined. This information is then compared with the extracted verification information. If they match, the verification passes. After that, the data header and verification information are removed from the data packets, and the remaining part is the restored environmental monitoring data of the marine fish fry pond. If they do not match, a retransmission command is sent to the sensor array data processing module until the verification passes and the environmental monitoring data is successfully restored.

[0071] In summary, relying on a multi-parameter sensor array, multi-dimensional environmental monitoring data in marine fish breeding ponds can be collected simultaneously, avoiding the limitations of single-parameter collection, achieving a comprehensive perception of the environmental status of the breeding pond, providing a complete and multi-dimensional data foundation for subsequent environmental analysis, and ensuring an accurate grasp of the overall environmental conditions of the breeding pond.

[0072] In summary, by adding data headers and verification information during transmission, and by having the cloud-based remote monitoring platform verify and restore the received data packets, the system effectively avoids data loss and distortion during transmission, ensuring the integrity and accuracy of environmental monitoring data and guaranteeing that the data transmitted to the platform can be directly used for subsequent processing.

[0073] In summary, data transmission to the cloud platform breaks the spatial limitations of local monitoring, allowing staff to remotely obtain real-time environmental data of the seedling ponds without on-site supervision, thus enabling remote dynamic monitoring.

[0074] In summary, this transmission link is a core prerequisite for subsequent data cleaning, temporal feature extraction, environmental trend prediction, and control parameter optimization. It provides reliable data input for the entire remote intelligent control process, supports subsequent precise control actions, and helps ensure the stability of the seedling environment.

[0075] S2. Eliminate noise and outliers in the environmental monitoring data in the cloud-based remote monitoring platform to obtain a standardized dataset of the marine fish breeding pond.

[0076] In this embodiment of the invention, the step of eliminating noise and outliers in the environmental monitoring data on the cloud-based remote monitoring platform to obtain a standardized dataset for the marine fish fry pond includes:

[0077] Based on preset thresholds, missing and abnormal data points in the environmental monitoring data are identified.

[0078] Remove the abnormal data points and perform interpolation on the missing data points to obtain the environmental data of the marine fish broodstock pond after cleaning;

[0079] The environmental data after cleaning is scaled to a preset numerical range to obtain a standardized dataset for the marine fish fry breeding pond.

[0080] Specifically, a preset threshold is first obtained, which includes the upper and lower limits of normal values ​​for each environmental monitoring parameter. Then, each data point in the environmental monitoring data of the marine fish broodstock pond is checked one by one. If a data point has no specific value, it is identified as a missing data point. If the value of a data point exceeds the upper and lower limits of the preset threshold of the corresponding environmental monitoring parameter, it is identified as an abnormal data point. Finally, the missing data points and abnormal data points in the environmental monitoring data are obtained.

[0081] Furthermore, the identified abnormal data points are directly removed from the environmental monitoring data. For the identified missing data points, linear interpolation is used. Specifically, the two adjacent valid data points before and after the missing data point are found, the numerical difference between the two valid data points is calculated, and the numerical difference is proportionally allocated according to the position of the missing data point between the two valid data points to obtain the interpolated value of the missing data point and fill it in. After completing the interpolation of all missing data points and the removal of abnormal data points, the environmental data of the marine fish fry pond after cleaning is obtained.

[0082] Furthermore, a preset numerical range is determined, and then the maximum and minimum values ​​of the corresponding environmental monitoring parameters in the cleaned environmental data are found. For each data point in the cleaned environmental data, the minimum value of the corresponding parameter is subtracted from the value of the data point, and then the result is divided by the difference between the maximum and minimum values ​​of the corresponding parameter. The result is the scaled data. After all data points are scaled, a standardized dataset of the marine fish broodstock pond is obtained.

[0083] In summary, it can effectively eliminate noise and outliers in environmental monitoring data caused by sensor errors and transmission interference, avoid distorted data from misleading subsequent analysis, ensure the authenticity and reliability of data, provide a high-quality data foundation for subsequent prediction of environmental parameter change trends and optimization of control parameters, and reduce the impact of data deviation on the overall control process from the source.

[0084] In summary, by imputing missing data points, incompleteness in the data acquisition process can be compensated for, ensuring the continuity and integrity of the dataset and avoiding analytical loopholes caused by data gaps. Scaling the cleaned data to a preset numerical range can unify the units and numerical scales of different environmental parameters (such as water temperature and dissolved oxygen), eliminating the interference of numerical differences between parameters on time-series feature extraction (such as sliding window segmentation and statistical feature calculation) and model operation. This lays a regular data foundation for the subsequent accurate prediction of environmental change data by long short-term memory neural network models, and further supports the accurate generation of subsequent optimal control parameters.

[0085] S3. Based on the time-series feature data extracted from the standardized dataset, predict the changing trend of environmental parameters in the marine fish broodstock pond to obtain environmental change data of the marine fish broodstock pond.

[0086] In this embodiment of the invention, the time-series feature data extracted from the standardized dataset includes:

[0087] Arrange the standardized dataset in chronological order to obtain the time series data of the standardized dataset;

[0088] In a sliding window, the time series data is segmented to obtain a data window of the standardized dataset;

[0089] Statistical features are extracted from the data window to obtain the initial feature set of the marine fish fry breeding pond;

[0090] By retaining the features in the initial feature set that are highly correlated with environmental changes, the time-series feature data of the marine fish broodstock pond are obtained.

[0091] In this embodiment of the invention, predicting the changing trends of environmental parameters in the marine fish broodstock pond to obtain environmental change data of the marine fish broodstock pond includes:

[0092] The time-series feature data is divided into a training set and a validation set;

[0093] The pre-constructed long short-term memory neural network model is optimized based on the training set, and the prediction accuracy of the optimized long short-term memory neural network model is evaluated through the validation set.

[0094] The standardized dataset is input into the long short-term memory neural network model with the required prediction accuracy to obtain the environmental change data of the marine fish fry pond.

[0095] In this embodiment of the invention, the optimization of the pre-constructed long short-term memory neural network model based on the training set, and the evaluation of the prediction accuracy of the optimized long short-term memory neural network model using the validation set, includes:

[0096] The training set is input into the long short-term memory neural network model for forward propagation to obtain the training prediction results of the training set;

[0097] The model loss of the long short-term memory neural network model is calculated based on the difference between the training prediction results and the true values.

[0098] The parameters of the long short-term memory neural network model are optimized based on the model loss.

[0099] The validation set is input into the optimized long short-term memory neural network model to obtain the validation prediction results of the validation set;

[0100] The prediction accuracy of the optimized long short-term memory neural network model is evaluated based on the error between the verification prediction results and the corresponding true values.

[0101] The hyperparameters of the long short-term memory neural network model are adjusted based on the predicted accuracy.

[0102] In this embodiment of the invention, the formula for calculating the model loss is as follows:

[0103] ;

[0104] In the formula, The loss of the model, The number of samples in the training set. Let be the ordinal number of the sample in the training set. For the first The true value of each sample, For the long short-term memory neural network model, the first... The training prediction results for each sample.

[0105] Specifically, the collection time corresponding to each data point in the standardized dataset is first extracted. Based on the order of collection time, all data points in the standardized dataset are arranged in order from the earliest collection time to the latest collection time. After the arrangement is completed, the time series data of the standardized dataset is obtained.

[0106] Furthermore, a fixed length for the sliding window is determined, which is the number of consecutive time series data points it contains. Starting from the first data point of the time series data, consecutive data points of the same length as the sliding window are selected to form the first data window. Then, the sliding window is moved back one data point, and consecutive data points of the same length as the window are selected at the new position to form the next data window. This moving and selection operation is repeated until the sliding window covers the last data point of the time series data, and finally the data window of the standardized dataset is obtained.

[0107] Furthermore, for each data window, the average value of all data points within the window is calculated by summing the values ​​of all data points within the window and dividing by the total number of data points; the maximum value of all data points within the window is calculated by finding the largest value among all data points within the window; the minimum value of all data points within the window is calculated by finding the smallest value among all data points within the window; and the variance of all data points within the window is calculated by first calculating the difference between each data point value and the window's average value, squared each difference, summing them, and then dividing by the total number of data points. The average value, maximum value, minimum value, and variance of each data window are used as the statistical characteristics of that window. The statistical characteristics of all data windows are collected and integrated to obtain the initial feature set of the marine fish fry breeding pond.

[0108] Furthermore, environmental change indicators for marine fish broodstock ponds, such as water temperature and salinity changes, are identified. For each feature in the initial feature set, the trend of the feature value over time is compared with the trend of the environmental change indicator value over time. The proportion of times the two trends are consistent is counted out of the total number of comparisons. If this proportion exceeds the set high correlation standard, the feature is determined to be highly correlated with environmental changes. All features with high correlation with environmental changes in the initial feature set are selected and integrated to obtain the time-series feature data of marine fish broodstock ponds.

[0109] Specifically, the ratio of time series feature data to be divided is determined to be 8:2. According to the chronological order of the time series feature data, the first 80% of the time series feature data is selected as the training set, and the last 20% of the time series feature data is selected as the validation set, thus completing the operation of dividing the time series feature data into training set and validation set.

[0110] Furthermore, the training set data is input into the pre-built long short-term memory neural network model in fixed batches. After receiving the data, the model calculates the prediction results and compares them with the actual results of the corresponding data in the training set. Based on the differences obtained from the comparison, the weight parameters inside the model are adjusted. This input, calculation, comparison, and adjustment operation is repeated until the prediction error of the model on the training set no longer decreases, thus completing the model optimization. Then, the validation set data is input into the optimized long short-term memory neural network model in the same batches. The model outputs the prediction results of the validation set, and the degree of agreement between the prediction results and the actual results of the corresponding data in the validation set is statistically analyzed to evaluate the prediction accuracy of the optimized long short-term memory neural network model.

[0111] Furthermore, it is first determined whether the prediction accuracy of the optimized long short-term memory neural network model meets the preset requirements. If it does, the standardized dataset is organized according to the format of the data received by the model and input into the model in batches. The model processes and calculates the input standardized dataset according to the internally optimized parameters and outputs the corresponding prediction results. These prediction results are the environmental change data of the marine fish fry pond.

[0112] Specifically, the training set data is organized according to the input format required by the Long Short-Term Memory Neural Network Model. The data is first input into the input layer of the model, and the input layer passes the data to the hidden layer. The hidden layer retains the historical information of the temporal feature data through the memory unit, and performs calculations in combination with the current input data. The calculation results are passed to the output layer, and the output layer processes the calculation results of the hidden layer and outputs them to obtain the training prediction results of the training set.

[0113] Furthermore, the true value corresponding to each data point in the training set is collected, and the training prediction result is compared with the corresponding true value one by one. The difference between the training prediction result and the true value in each comparison group is calculated, and the absolute value of each difference is taken. All absolute values ​​are added together and then divided by the total number of data in the training set. The result is the model loss of the long short-term memory neural network model.

[0114] Furthermore, determine the acceptable range of model loss. If the model loss exceeds this range, start adjusting the model's weight parameters. First, calculate the direction of the model loss's influence on each weight parameter, i.e., the trend of model loss change when the weight parameter increases. If increasing the weight parameter leads to an increase in model loss, decrease the weight parameter. If increasing the weight parameter leads to a decrease in model loss, increase the weight parameter. After adjustment, recalculate the model loss. Repeat this adjustment and calculation process until the model loss drops to an acceptable range, thus completing the optimization of the long short-term memory neural network model parameters.

[0115] Furthermore, the validation set data is organized according to the input format of the optimized long short-term memory neural network model. The organized validation set data is then input into the model, and the data passes through the input layer, hidden layer, and output layer of the model in sequence. The hidden layer uses the optimized parameters to process the time series information, and the output layer outputs values ​​based on the processing results of the hidden layer. These values ​​are the validation prediction results of the validation set.

[0116] Furthermore, obtain the true value corresponding to each data point in the validation set, set the allowable error range, check one by one whether the difference between the validation prediction result and the corresponding true value is within the allowable error range, count the number of validation data points with the difference within the allowable range, divide this number by the total number of data points in the validation set, and the resulting ratio is the prediction accuracy of the optimized long short-term memory neural network model.

[0117] Furthermore, a required standard for prediction accuracy is set. If the prediction accuracy of the optimized model does not meet this standard, the hyperparameters of the model are selected for adjustment. Taking the number of hidden layer nodes as an example, the number of hidden layer nodes is increased first, the model parameters are re-optimized using the training set, and then the prediction accuracy of the model after adjusting the hyperparameters is evaluated using the validation set. If the accuracy improves, the number of nodes is retained; if the accuracy decreases, the number of hidden layer nodes is reduced. This process of hyperparameter adjustment, model optimization, and accuracy evaluation is repeated until the model prediction accuracy meets the required standard, thus completing the adjustment of the hyperparameters of the long short-term memory neural network model based on prediction accuracy.

[0118] Specifically, the model loss formula is used to calculate the model loss of a long short-term memory neural network model. The specific calculation process is as follows: First, for each sample in the training set, use the... The true value of the nth sample minus the model's effect on the nth sample The training prediction results for each sample are used to obtain the difference between the actual value and the training prediction result for each sample. Then, each difference is squared to avoid the positive and negative differences canceling each other out. Next, the squared differences of all samples are summed to obtain the sum of the squared differences of all samples. Finally, this sum is divided by the number of samples in the training set to obtain the model loss. The model loss can reflect the overall degree of difference between the training prediction results of the model on the training set samples and the actual values ​​of the samples.

[0119] In summary, by sorting by time, segmenting by sliding window, and filtering by highly correlated features, time-series feature data can be extracted from standardized data. This not only fully captures the dynamic patterns of environmental parameters changing over time, but also eliminates redundant features with low correlation to environmental changes, reducing the interference of invalid information on subsequent predictions. This makes the features input into the model more accurate and focused, laying a high-quality foundation for the prediction process.

[0120] In summary, employing a long short-term memory neural network model optimized on the training set, with accuracy assessed on the validation set and hyperparameters adjusted, can fully leverage the model's advantages in processing time-series data, improving the accuracy and reliability of predicting environmental parameter change trends. This ensures that the output environmental change data accurately reflects the future environmental state of the nursery pond. Accurate environmental change data provides a clear basis for subsequent multi-objective optimization of control parameters, helping to predict environmental anomalies (such as decreased dissolved oxygen or water temperature imbalance) in advance, avoiding passive regulation, and facilitating the generation of more targeted optimal control parameters to ensure environmental stability in the nursery pond.

[0121] S4. Based on the environmental change data described in the database, perform multi-objective optimization on the control parameters of the marine fish broodstock pond to obtain the optimal control parameters of the marine fish broodstock pond;

[0122] In this embodiment of the invention, the step of performing multi-objective optimization of the control parameters of the marine fish broodstock pond based on the environmental change data in the database to obtain the optimal control parameters of the marine fish broodstock pond includes:

[0123] Based on the environmental change data, the optimization objectives of the marine fish fry rearing pond are determined;

[0124] Set constraints on the control parameters in the marine fish broodstock pond;

[0125] Under the constraints, evaluate the degree to which all combinations of control parameters in the marine fish broodstock pond meet the optimization objective;

[0126] The parameter combination that achieves the highest degree of satisfaction is selected as the optimal control parameter for the marine fish broodstock pond.

[0127] Specifically, we first extract key environmental indicators from the environmental change data of the marine fish fry pond, such as water temperature change data and salinity change data. Combined with the suitable environmental indicator range required for the growth of marine fish fry, we determine the types of indicators in the environmental change data that deviate from the suitable range. If the environmental change data shows that the water temperature is consistently higher than the suitable range, then "stabilizing the water temperature of the marine fish fry pond within the suitable water temperature range for fry" is taken as the core direction. We then clarify the specific requirements of this direction and obtain the optimization goal of the marine fish fry pond.

[0128] Furthermore, first determine the control parameters used to regulate the environment in the marine fish broodstock pond, such as the operating power of the heating equipment, the water exchange frequency of the water exchange device, and the working intensity of the aeration equipment. Consult the safety operation specifications of marine fish broodstock equipment and the environmental control standards of the broodstock industry, and clarify the safe operating limits and stable requirements of the broodstock environment for each control parameter. For example, the operating power of the heating equipment must not exceed the rated maximum safe power of the equipment, and the water exchange frequency must not be less than once every two days to avoid water quality deterioration. Compile these clear restrictions into the constraints of the control parameters in the marine fish broodstock pond.

[0129] Furthermore, based on the constraints of the control parameters, all possible values ​​of each control parameter within the constraint range are listed one by one. Then, all combinations of control parameters that meet the constraints are generated through combination. For each combination of control parameters, the effect of the combination on the marine fish broodstock pond environment is simulated, such as the change in pond water temperature under the combined effect of a certain heating power and water exchange frequency. The simulation results are compared with the optimization objectives of the marine fish broodstock pond, and the degree of agreement between the simulation results and the optimization objectives is calculated. The degree of satisfaction of the optimization objectives by the control parameter combination is expressed as a specific percentage, and the satisfaction degree evaluation of all control parameter combinations is completed.

[0130] Furthermore, all combinations of control parameters and their corresponding satisfaction levels are organized and sorted from highest to lowest satisfaction level. The control parameter combination with the highest satisfaction level in the sorted list is identified, and it is confirmed that this combination still meets the constraints of the control parameters. This combination is then determined as the optimal control parameters for the marine fish fry breeding pond.

[0131] In summary, determining optimization targets based on environmental change data allows for precise matching of control directions with the future environmental dynamics of the seedling pond, preventing optimization from deviating from actual needs and ensuring that the final control parameters can specifically address subsequent environmental changes, thus providing targeted support for maintaining a stable seedling environment.

[0132] In summary, by setting control parameter constraints, parameters can be limited to the safe operating range of the equipment and the suitable range for seedling growth, effectively avoiding equipment failures or environmental unsuitability caused by parameter exceeding limits, and ensuring the safety and rationality of the control process. Furthermore, a comprehensive evaluation of the satisfaction level of all control parameter combinations allows for the selection of the optimal parameter combination from multiple dimensions, overcoming the limitations of single-objective optimization. This approach considers multiple needs such as seedling efficiency, energy consumption costs, and water quality balance, ensuring that the optimal control parameters not only meet environmental maintenance requirements but also have practical application value. This lays a crucial foundation for subsequent precise adjustments to the pond environment, helping to improve overall seedling efficiency.

[0133] S5. The optimal control parameters are transmitted to the local actuator of the marine fish fry pond through the cloud-based remote monitoring platform, and the pond environment of the marine fish fry pond is adjusted based on the optimal control parameters.

[0134] In this embodiment of the invention, the step of transmitting the optimal control parameters to the local actuator of the marine fish fry pond through the cloud-based remote monitoring platform, and adjusting the pond environment based on the optimal control parameters, includes:

[0135] The optimal control parameters are converted into a set of equipment control instructions;

[0136] The device control command set is transmitted to the local actuator of the fish breeding pond through the encrypted communication link of the cloud-based remote monitoring platform.

[0137] The local actuator parses the received set of device control instructions to obtain the device drive signal corresponding to the local actuator;

[0138] Based on the device drive signal, the working status of the heating device, oxygenation device and chemical regulator are synchronously controlled to adjust the pool environment.

[0139] Specifically, first identify the control objects corresponding to the optimal control parameters, namely heating devices, aeration equipment, and chemical regulators. For the optimal power parameters of the heating devices, generate the instruction "adjust the heating devices to XX power level". For the optimal working intensity parameters of the aeration equipment, generate the instruction "set the aeration intensity of the aeration equipment to XX". For the optimal reagent release parameters of the chemical regulator, generate the instruction "control the chemical regulator to release XX amount of reagent per hour". Classify and organize these three types of instructions according to equipment type to form a set containing all equipment control instructions, thus obtaining the equipment control instruction set.

[0140] Furthermore, it was confirmed that the encrypted communication link between the cloud-based remote monitoring platform and the local actuator in the marine fish breeding pond was connected. On the cloud-based remote monitoring platform, the device control command set was converted according to the character encoding rules preset by the encrypted communication link to ensure that the command content was not leaked. Then, the encoded device control command set was sent to the local actuator through the link. After the local actuator responded with a signal that the reception was successful, the transmission operation was completed.

[0141] Furthermore, the local actuator stores instruction parsing rules corresponding to the heating device, aeration equipment, and chemical regulator. For example, "adjust the heating device to XX power level" corresponds to a specific current signal driving the heating device, and "set the aeration intensity of the aeration equipment to XX" corresponds to a specific pulse signal driving the aeration equipment. The local actuator reads each instruction in the received device control instruction set one by one, and converts each instruction into an electrical signal that the corresponding device can recognize according to the preset parsing rules. These electrical signals are the device drive signals corresponding to the local actuator.

[0142] Furthermore, the device drive signal corresponding to the heating device is transmitted to the control port of the heating device. After receiving the signal, the heating device immediately adjusts the working power of its internal heating element to reach the required level. At the same time, the device drive signal corresponding to the aeration device is transmitted to the control module of the aeration device. The aeration device adjusts the operating speed of the aeration pump according to the signal to achieve the specified aeration intensity. Then, the device drive signal corresponding to the chemical regulator is transmitted to the chemical regulator's chemical control unit. The chemical regulator opens or adjusts the opening of the chemical release valve according to the signal to control the amount of chemical released. During this process, the operating status of the three devices is monitored in real time to ensure that they work synchronously according to the device drive signals, and finally the adjustment of the environment of the marine fish fry pond is completed.

[0143] In summary, transmitting optimal control parameters through the encrypted communication link of the cloud-based remote monitoring platform can effectively resist interference and data tampering risks during transmission, ensure the security and integrity of parameter transmission, and ensure that the control commands received by the local actuator are accurate, laying the foundation for the accuracy of subsequent environmental adjustments.

[0144] In summary, converting optimal control parameters into a set of equipment control instructions enables the transformation of abstract parameters into executable signals, resolving the connection between parameters and actuator actions. This allows the actuator to quickly parse and generate corresponding equipment drive signals. Furthermore, the simultaneous control of heating devices, aeration equipment, and chemical regulators by the actuators achieves coordinated action of multiple environmental control devices, avoiding environmental parameter imbalances caused by adjustments to a single device and ensuring that the pool environment reaches a suitable state for seedling cultivation simultaneously across multiple dimensions.

[0145] In summary, remote transmission and automated adjustment eliminate the need for manual on-site operation, significantly improving the response speed of environmental control. It is particularly suitable for multi-pond management scenarios in large-scale seedling bases, ensuring the timeliness and consistency of control actions, and ultimately maintaining the optimal environment required for marine fish seedling cultivation, thus helping to improve the survival rate and growth quality of seedlings.

[0146] like Figure 2 The diagram shown is a functional block diagram of a remote intelligent control system for marine fish fry rearing environment provided in an embodiment of the present invention.

[0147] The remote intelligent control system 100 for marine fish fry rearing environment described in this invention can be installed in an electronic device. Depending on the functions implemented, the remote intelligent control system 100 may include an environmental data transmission module 101, a data cleaning module 102, a water quality change prediction module 103, a fry rearing control module 104, and a pond adjustment module 105. The module described in this invention can also be called a unit, referring to a series of computer program segments that can be executed by the processor of an electronic device and perform a fixed function, stored in the memory of the electronic device.

[0148] In this embodiment, the functions of each module / unit are as follows:

[0149] The environmental data transmission module 101 is used to transmit the environmental monitoring data collected by the multi-parameter sensor array in the marine fish breeding pond to the cloud-based remote monitoring platform through the Internet of Things channel.

[0150] The data cleaning module 102 is used to remove noise and outliers in the environmental monitoring data in the cloud-based remote monitoring platform to obtain a standardized dataset of the marine fish breeding pond.

[0151] The water quality change prediction module 103 is used to predict the environmental parameter change trend of the marine fish broodstock pond based on the time-series feature data extracted from the standardized dataset, and obtain the environmental change data of the marine fish broodstock pond.

[0152] The seedling control module 104 is used to perform multi-objective optimization of the control parameters of the marine fish seedling pond based on the environmental change data in the database, so as to obtain the optimal control parameters of the marine fish seedling pond.

[0153] The pool adjustment module 105 is used to transmit the optimal control parameters to the local actuator of the marine fish fry pond through the cloud-based remote monitoring platform, and adjust the pool environment of the marine fish fry pond based on the optimal control parameters.

[0154] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0155] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0156] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0157] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0158] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0159] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for remotely intelligent regulation of a marine fish larviculture environment, characterized in that, The method comprises: S1, transmitting the environmental monitoring data collected by the multi-parameter sensor array in the seawater fish breeding pond to the cloud-based remote monitoring platform through an Internet of Things channel; S2, eliminating noise points and abnormal values in the environmental monitoring data in the cloud-based remote monitoring platform to obtain a standardized data set of the seawater fish breeding pond; S3, predicting the environmental parameter change trend of the seawater fish breeding pond based on the time series feature data extracted from the standardized data set, and obtaining the environmental change data of the seawater fish breeding pond; S4, multi-objective optimization of the control parameters of the seawater fish breeding pond according to the environmental change data, and obtaining the optimal control parameters of the seawater fish breeding pond; S5, transmitting the optimal control parameters to the local actuator of the seawater fish breeding pond through the cloud-based remote monitoring platform, and adjusting the pond environment of the seawater fish breeding pond based on the optimal control parameters.

2. The method for remote intelligent control of marine fish fry rearing environment as described in claim 1, characterized in that, The method comprises: Adding data header and check information to the environmental monitoring data collected by the multi-parameter sensor array in the seawater fish breeding pond to obtain the transmission data packet of the seawater fish breeding pond; Sending the transmission data packet to the cloud-based remote monitoring platform through the Internet of Things channel; In the cloud-based remote monitoring platform, the received transmission data packet is checked, and the environmental monitoring data of the seawater fish breeding pond is restored.

3. The method for remote intelligent control of marine fish fry rearing environment as described in claim 1, characterized in that, The method comprises: Identifying missing data points and abnormal data points in the environmental monitoring data based on a preset threshold; Removing the abnormal data points and performing interpolation processing on the missing data points to obtain the cleaned environmental data of the seawater fish breeding pond; Scaling the cleaned environmental data to a preset value range to obtain the standardized data set of the seawater fish breeding pond.

4. The method for remote intelligent control of marine fish fry rearing environment as described in claim 1, characterized in that, The time series feature data extracted from the standardized data set comprises: Arranging the standardized data set in time sequence to obtain the time series data of the standardized data set; Segmenting the time series data in a sliding window to obtain the data window of the standardized data set; Extracting statistical features from the data window to obtain an initial feature set of the seawater fish breeding pond; Retaining features with high environmental change correlation in the initial feature set to obtain the time series feature data of the seawater fish breeding pond.

5. The method for remote intelligent control of marine fish fry rearing environment as described in claim 4, characterized in that, The method comprises: Dividing the time series feature data into a training set and a validation set; Optimizing the pre-constructed long short-term memory neural network model based on the training set, and evaluating the prediction accuracy of the optimized long short-term memory neural network model through the validation set; Inputting the standardized data set into the long short-term memory neural network model with the required prediction accuracy to obtain the environmental change data of the seawater fish breeding pond.

6. The method of claim 5, wherein the method further comprises: determining the water quality of the marine fish hatchery environment; and adjusting the water quality of the marine fish hatchery environment based on the determined water quality. The long short-term memory neural network model is optimized based on the training set, and the prediction accuracy of the optimized long short-term memory neural network model is evaluated through the validation set, including: The training set is input into the long short-term memory neural network model for forward propagation to obtain a training prediction result of the training set; A model loss of the long short-term memory neural network model is calculated based on the difference between the training prediction result and the true value; Parameters of the long short-term memory neural network model are optimized according to the model loss; The validation set is input into the optimized long short-term memory neural network model to obtain a validation prediction result of the validation set; The prediction accuracy of the optimized long short-term memory neural network model is evaluated according to the error between the validation prediction result and the corresponding true value; The hyperparameters of the long short-term memory neural network model are adjusted based on the prediction accuracy.

7. The method of claim 6, wherein the method further comprises: determining the water quality of the marine fish hatchery environment; and adjusting the water quality of the marine fish hatchery environment based on the determined water quality of the marine fish hatchery environment. The calculation formula of the model loss is as follows: ； In the formula, is the model loss, is the number of samples in the training set, is the ordinal number of the sample in the training set, is the true value of the first sample, is the training prediction result of the first sample by the long short-term memory neural network model.

8. The method of claim 1, wherein the method further comprises: determining the water quality of the marine fish hatchery environment; and adjusting the water quality of the marine fish hatchery environment based on the determined water quality. The control parameters of the seawater fish seedling pool are multi-objective optimized based on the environmental change data to obtain optimal control parameters of the seawater fish seedling pool, including: Based on the environmental change data, the optimization target of the seawater fish seedling pool is determined; The constraint condition of the control parameter in the seawater fish seedling pool is set; Under the constraint condition, the satisfaction degree of all control parameter combinations in the seawater fish seedling pool to the optimization target is evaluated; The parameter combination with the highest satisfaction degree is selected as the optimal control parameter of the seawater fish seedling pool.

9. The method for remote intelligent control of marine fish fry rearing environment as described in claim 1, characterized in that, The optimal control parameter is transmitted to the local actuator of the seawater fish seedling pool through the cloud-based remote monitoring platform, and the pool body environment of the seawater fish seedling pool is adjusted based on the optimal control parameter, including: The optimal control parameter is converted into a device control instruction set; The device control instruction set is transmitted to the local actuator of the seawater fish seedling pool through the encrypted communication link of the cloud-based remote monitoring platform; The local actuator parses the received device control instruction set to obtain a device driving signal corresponding to the local actuator; Based on the device driving signal, the working states of the heating device, the oxygenation equipment and the chemical regulator are synchronously controlled to complete the adjustment of the pool body environment.

10. A remote intelligent control system for a marine fish larviculture environment, characterized in that, The system comprises: An environmental data transmission module for transmitting environmental monitoring data collected by a multi-parameter sensor array in a seawater fish seedling pool to a cloud-based remote monitoring platform through an Internet of Things channel; A data cleaning module for eliminating noise and outliers in the environmental monitoring data in the cloud-based remote monitoring platform to obtain a standardized data set of the seawater fish seedling pool; A water quality change prediction module for predicting the environmental parameter change trend of the seawater fish seedling pool based on time series feature data extracted from the standardized data set to obtain environmental change data of the seawater fish seedling pool; A seedling control module for multi-objective optimizing control parameters of the seawater fish seedling pool based on the environmental change data to obtain optimal control parameters of the seawater fish seedling pool; The pool body adjusting module is configured to transmit the optimal control parameter to a local actuator of the seawater fish breeding pool through the cloud-based remote monitoring platform, and adjust the pool body environment of the seawater fish breeding pool based on the optimal control parameter.

Images