Aquaculture intelligent oxygenation device based on neural network

Abstract

The invention discloses an aquaculture intelligent oxygenation device based on a neural network, which adopts a low-energy-consumption and high-efficiency technology to remarkably improve the dissolved oxygen content of a water body and promote the healthy growth of aquatic animals such as shrimps, crabs and fish schools. The device is suitable for small, medium and large culture pond environments, and the frequency of an air blower is adjusted through an intelligent control system to adaptively adjust the water dissolved oxygen level. The system integrates the fuzzy neural network model with strong self-learning ability, and can predict the dissolved oxygen amount according to the real-time data collected by the water quality monitoring module, thereby realizing accurate environment control. In addition, due to the modular design of the device, installation and maintenance are more convenient, and the technical threshold of farmers is greatly lowered. The device is also integrated with a water quality monitoring and communication module, supports real-time environment monitoring and data remote transmission, and realizes remote monitoring and management functions. The highly integrated intelligent solution provides powerful technical support for modernized and intelligent aquaculture.

Description

Technical Field

[0001] The present invention belongs to the field of aquaculture technology, and in particular is an intelligent oxygenation device based on a neural network, which is intended to improve the dissolved oxygen content of water and promote the healthy growth of aquaculture organisms. The device is suitable for a variety of aquaculture environments, including but not limited to various shrimps, crabs, and fish schools. Background Art

[0002] Faced with the dual challenges of global environmental change, resource shortage and population growth, green, low-carbon and energy-saving sustainable aquaculture models have become an inevitable trend in the development of the industry. As an important source of efficient protein production, the sustainable development capacity of aquaculture is directly related to food safety and ecological environmental protection. In this context, achieving energy conservation and environmental protection in the breeding process, especially optimizing water quality management and improving oxygenation efficiency, has become the key to improving the economic and environmental benefits of breeding. Although traditional extensive aeration equipment and methods have met basic needs to a certain extent, their low energy efficiency, potential damage to the environment and high operating costs have limited the sustainable development of the breeding industry. These traditional methods often cannot accurately respond to the dynamically changing dissolved oxygen demand in the breeding water, resulting in insufficient or excessive dissolved oxygen supply, affecting the healthy growth of aquatic organisms, and may cause water quality deterioration and increase environmental pressure.

[0003] Therefore, developing an intelligent, efficient and environmentally friendly oxygenation device that can accurately adjust the dissolved oxygen content according to real-time monitored water quality data can not only significantly improve the growth efficiency and production quality of farmed organisms, but also reduce energy consumption and environmental impact, and has important social and economic value. The intelligent oxygenation device of the present invention integrates advanced neural network algorithms and automatic control technologies to achieve real-time and accurate regulation of dissolved oxygen conditions in a variety of breeding environments, which not only meets the needs of various aquaculture organisms for dissolved oxygen, but also promotes the development of aquaculture industry in the direction of green, low-carbon and energy-saving. Summary of the invention

[0004] The present invention aims to solve the limitations of existing aquaculture oxygenation technology in terms of water quality regulation ability, energy efficiency, artificial dependence, oxygenation effect and technical level, and provide a neural network-based intelligent aeration device for aquaculture ponds that integrates the advantages of high integration, easy operation and maintenance. This device can effectively improve the growth efficiency and health of aquaculture organisms through accurate monitoring and automatic regulation of the real-time water quality environment.

[0005] An intelligent oxygenation device for aquaculture based on neural network, characterized in that: in view of the problems of limited regulation capacity, low energy utilization efficiency, great dependence on manual labor, limited oxygenation effect, low technical level, etc. of the oxygenation devices currently used in breeding ponds, the device can automatically and accurately control the dissolved oxygen required by white shrimp according to the actual water quality environment, and it has a high degree of integration, simple operation and easy maintenance. The dissolved oxygen is predicted by the built-in neural network model, and each operating component in the device is accurately controlled by the intelligent control algorithm, thereby greatly improving the shrimp breeding efficiency on the basis of strictly controlling the dissolved oxygen concentration. The device is composed of a carrier, a water quality monitoring module, an intelligent control module, a communication module, and an oxygenation device. It includes a float, a support rod, a carrier plate, a temperature sensor, a dissolved oxygen sensor, a pH sensor, an ammonia nitrogen sensor, a salinity sensor, a liquid level sensor, a light sensor, a motor, a telescopic rod, a blower, a main pipe, a branch pipe, and a microporous plastic pipe.

[0006] The carrier is composed of a float, a support rod and a carrier plate. There are four floats placed on the water surface and connected to the carrier plate through the support rods. The buoyancy of the floats enables the carrier plate to withstand a certain weight without sinking. The four sides of the carrier plate can be connected to the pond bank with ropes to fix the carrier on the water surface.

[0007] The water quality monitoring module integrates a temperature sensor, a dissolved oxygen sensor, a pH sensor, an ammonia nitrogen sensor, a salinity sensor, a liquid level sensor, a light sensor, and a camera. The module is arranged underwater and is further connected to the intelligent control module by a rotatable telescopic rod, which can adjust the height and rotate, thereby realizing water quality monitoring in different water layers.

[0008] The intelligent control module is composed of a central processing unit, a plc controller, and a built-in communication module. After receiving the signal from the water quality monitoring module, the central processing unit further uses the built-in neural network model to predict the amount of dissolved oxygen required by the water body. The processor further converts the predicted value into a control signal and transmits it to the plc controller, which further controls the blower to compress the air to achieve water oxygenation. At the same time, the built-in communication module of the intelligent control module realizes 5G wireless communication with the host computer through the cloud server, and further transmits information such as the remaining battery power, real-time water quality data, and water body scene to the host computer in real time.

[0009] The neural network prediction process is as follows: s1. Collect a large amount of training data from the actual breeding environment, including parameters such as pH value, ammonia nitrogen, temperature, salinity, light, and the corresponding dissolved oxygen required by white shrimp; after the collection is completed, further check the collected abnormal data, remove abnormal values, duplicate values ​​and incomplete records; further use 70% of the collected data as a training set, and the remaining 30% as a test set; further normalize the data and scale each feature to between 0 and 1; further divide the normalized training set data into input and output. s2. Construct the bp neural network structure: further set the number of neurons in the input layer to 5, corresponding to pH value, ammonia nitrogen, temperature, salinity and light; further use one hidden layer in the hidden layer, the number of neurons is 10, and the activation function uses ReLU; further the number of neurons in the output layer is 1, indicating the predicted dissolved oxygen required by white shrimp, and the activation function selects the linear activation function; further use the mean square error (MSE) as the loss function; further use the stochastic gradient descent optimization algorithm. s3, training bp neural network: further use smaller random numbers to initialize the weights w and bias b of the neural network; further set the learning rate, number of iterations and batch size; further for each iteration, divide the training data into the next small batch of the batch size, further for each small batch, forward propagation: calculate the output of the neural network according to the current weights and biases; calculate the loss: use the mean square error (MSE) loss function to calculate the difference between the predicted value and the actual value; back propagation: calculate the gradient of the loss function with respect to the weights and biases; update the weights and biases: update the weights and biases according to the gradient and learning rate; further calculate the loss values ​​of the training set and the test set to evaluate the performance of the model; further if the stopping condition is met (such as reaching the maximum number of iterations or the test set loss value is less than the preset threshold), stop training. s4, predicting dissolved oxygen: through the new water quality parameter input collected by the water quality detection module, further use the trained neural network for forward propagation to obtain the predicted dissolved oxygen required by the white shrimp; further denormalize the prediction structure, and denormalize the prediction result from the range of 0-1 to the range of actual dissolved oxygen.

[0010] The oxygenation device is composed of a blower, a main pipe, a branch pipe, and a microporous plastic hose. The blower is mounted on a carrier plate and connected to the main pipe. The main pipe is further fixed to the carrier plate through a support rod, and is provided with 4 branch pipe openings for connecting the branch pipes. The microporous plastic pipe is placed at the bottom of the pond through the branch pipe. The plastic pipe is provided with micropores. When pressurized air passes through the plastic pipe, tiny bubbles are discharged. In the process of the bubbles rising, water movement is formed, and part of the oxygen is dissolved in the water to achieve the purpose of oxygenation. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1The overall structural diagram of the present invention is as follows:

[0012] Figure 2 The top view of the overall structure of the present invention is:

[0013] Figure 3 Detailed diagram of a portion of the oxygenation device of the present invention

[0014] Figure 4 Schematic diagram of the neural network model prediction process of the present invention

[0015] Figure 5 The control system structure diagram of the present invention is: DETAILED DESCRIPTION

[0016] The preferred technical solutions of the present invention are described in detail below with reference to the accompanying drawings.

[0017] The invention discloses an intelligent oxygenation device for aquaculture based on a neural network, which is characterized in that: in view of the problems of limited water quality regulation ability, low energy utilization efficiency, large dependence on manual labor, limited oxygenation effect, low technical level, etc. in the current oxygenation devices used in aquaculture ponds, the device can automatically and accurately control the dissolved oxygen required by white shrimp according to the actual water quality conditions, and it has a high degree of integration, simple operation and easy maintenance. Through the built-in neural network model, each operating component in the device is accurately controlled, thereby greatly improving the breeding efficiency of white shrimp on the basis of strictly controlling the oxygen concentration. The structure includes: 1. pond, 2. float, 3. support rod, 4. carrier plate, 5. main support rod, 6. blower, 7. hose, 8. intelligent control module, 9. rotatable telescopic rod, 10. water quality detection module, 11. main pipe, 12. branch pipe, 13. plastic aeration pipe, 14. micropore.

[0018] The float, support rod and carrier plate constitute a carrier. There are four floats placed on the water surface and connected to the carrier plate through the support rod. The buoyancy of the float enables the carrier plate to bear a certain weight without sinking. The four sides of the carrier plate can be connected to the bank of the pond with ropes to fix the carrier on the water surface.

[0019] The water quality monitoring module integrates a temperature sensor, a dissolved oxygen sensor, a pH sensor, an ammonia nitrogen sensor, a salinity sensor, a liquid level sensor, a light sensor, and a camera. The module is arranged underwater and is further connected to the intelligent control module by a rotatable telescopic rod, which can adjust the height and rotate, thereby realizing water quality monitoring in different water layers.

[0020] The intelligent control module is composed of a central processing unit, a plc controller, and a built-in communication module. After receiving the signal from the water quality monitoring module, the central processing unit further uses the built-in neural network model to predict the amount of dissolved oxygen required by the water body. The processor further converts the predicted value into a control signal and transmits it to the plc controller, which further controls the blower to compress the air to achieve water oxygenation. At the same time, the built-in communication module of the intelligent control module realizes 5G wireless communication with the host computer through the cloud server, and further transmits information such as the remaining battery power, real-time water quality data, and water body scene to the host computer in real time.

[0021] The neural network prediction process is as follows: s1. Collect a large amount of training data from the actual breeding environment, including parameters such as pH value, ammonia nitrogen, temperature, salinity, light, and the corresponding dissolved oxygen required by white shrimp; after the collection is completed, further check the collected abnormal data, remove abnormal values, duplicate values ​​and incomplete records; further use 70% of the collected data as a training set, and the remaining 30% as a test set; further normalize the data and scale each feature to between 0 and 1; further divide the normalized training set data into input and output. s2. Construct the bp neural network structure: further set the number of neurons in the input layer to 5, corresponding to pH value, ammonia nitrogen, temperature, salinity and light; further use one hidden layer in the hidden layer, the number of neurons is 10, and the activation function uses ReLU; further the number of neurons in the output layer is 1, indicating the predicted dissolved oxygen required by white shrimp, and the activation function selects the linear activation function; further use the mean square error (MSE) as the loss function; further use the stochastic gradient descent optimization algorithm. s3, training bp neural network: further use smaller random numbers to initialize the weights w and bias b of the neural network; further set the learning rate, number of iterations and batch size; further for each iteration, divide the training data into the next small batch of the batch size, further for each small batch, forward propagation: calculate the output of the neural network according to the current weights and biases; calculate the loss: use the mean square error (MSE) loss function to calculate the difference between the predicted value and the actual value; back propagation: calculate the gradient of the loss function with respect to the weights and biases; update the weights and biases: update the weights and biases according to the gradient and learning rate; further calculate the loss values ​​of the training set and the test set to evaluate the performance of the model; further if the stopping condition is met (such as reaching the maximum number of iterations or the test set loss value is less than the preset threshold), stop training. s4, predicting dissolved oxygen: through the new water quality parameter input collected by the water quality detection module, further use the trained neural network for forward propagation to obtain the predicted dissolved oxygen required by the white shrimp; further denormalize the prediction structure, and denormalize the prediction result from the range of 0-1 to the range of actual dissolved oxygen.

[0022] The blower is mounted on a carrier plate and connected to a main pipe, and a further main pipe is fixed to the carrier plate through a main pipe support rod, and is provided with 4 branch pipe openings for connecting branch pipes. The microporous plastic pipe is placed at the bottom of the pond through the branch pipe, and micropores are provided on the plastic pipe. When pressurized air passes through the plastic pipe, tiny bubbles are discharged, and water movement is formed during the rising process of the bubbles, and a part of the oxygen is dissolved in the water to achieve the purpose of oxygenation.

[0023] The above description is only a preferred specific implementation manner of the present invention, and the protection scope of the present invention is not limited thereto. Any simple change or equivalent replacement of the technical solution that can be obviously obtained by any technician familiar with the technical field within the technical scope disclosed in the present invention falls within the protection scope of the present invention.

Claims

1. An intelligent oxygenation device for aquaculture based on neural network, characterized in that: The pressurized air is sent into the microporous plastic tube installed at the bottom of the water through the blower, so that tiny bubbles are discharged, and water movement is formed during the rising process. Part of the oxygen is integrated into the water to oxygenate the water. This device introduces an intelligent control algorithm with autonomous learning ability and inference, judgment and decision-making ability in the control system, so that the control system can predict and control the amount of compressed air introduced according to the real-time water quality parameters and their changes. This device consists of a carrier, a water quality monitoring module, an intelligent control module, a communication module, and an oxygenation device. It includes a float, a support rod, a carrier plate, a temperature sensor, a dissolved oxygen sensor, a pH sensor, an ammonia nitrogen sensor, a salinity sensor, a liquid level sensor, a light sensor, a battery, a telescopic rod, a blower, a main pipe, a branch pipe, and a microporous plastic tube.

2. The carrier according to claim 1, characterized in that: The carrier is composed of a floating ball, a supporting rod and a carrier plate. There are four floating balls arranged on both sides of the carrier plate, floating on the water surface. The carrier plate mainly bears the load, and four ropes can be used to connect the bank of the pond to fix the carrier in the pond.

3. The water quality monitoring module according to claim 1 is characterized in that: The water quality monitoring module integrates a temperature sensor, a dissolved oxygen sensor, a pH sensor, an ammonia nitrogen sensor, a salinity sensor, a liquid level sensor, a light sensor, and a camera, and is connected to the intelligent control module through a wired connection in a rotatable waterproof telescopic rod. The telescopic rod can adjust the height and rotate, connect the water quality monitoring module and the intelligent control module, and cooperate with the water quality monitoring module to realize water quality monitoring of different water layers.

4. The intelligent control module according to claim 1 is characterized in that: The intelligent control module is composed of a central processing unit and a PLC controller. After receiving the signal from the water quality monitoring module, the central processing unit uses the built-in neural network model to predict the amount of dissolved oxygen required by the water body. The processor further converts the predicted value into a control signal and transmits it to the PLC controller, which controls the blower to compress air to achieve water oxygenation. The intelligent control module also has a built-in battery to power the entire system.

5. The neural network model for predicting dissolved oxygen according to claim 4, characterized in that: The central processing unit includes a data preprocessing module and a neural network prediction module. The neural network has a strong parallel structure and parallel implementation capability, is suitable for real-time control and dynamic control, and its internal distributed information storage and data processing structure give it unique fault tolerance. The neural network can also learn by studying the system's past data records, has the ability to summarize all data, and can quickly implement a large number of complex control algorithms. The data processing structure performs normalization processing on the collected data, and the processed data can be used as input parameters of the neural network. The normalization processing formula is: Where X * represents the normalized data, X max is the maximum value of X among all data, X min is the minimum value of X among all data, X n is the new value obtained after normalization.

6. The neural network prediction module according to claim 5, characterized in that The neural network prediction module includes an artificial neural network ANN, and the steps of setting parameters of the artificial neural network ANN and building an ANN model include: A 3-layer BP neural network is used as an ANN model, and the topological structure and minimum allowable error parameters of the BP neural network model are input; wherein the topological structure of the BP neural network model includes the number of input layers, the number of hidden layers, and the number of output layers; and the hidden layer activation function and the output layer activation function are set; Calculate the input α of the hidden layer neurons in the bp neural network model h (k) and output β h (K), and the input y of the output layer neuron j (k), and output y o (k), which is calculated as follows: b h (k)=f1(α h (k)) I j (k)=f2(y j (k)) In the formula, w ih is the connection weight between the input layer and the hidden layer, b h is the threshold of each neuron in the hidden layer, w hj is the connection weight between the hidden layer and the output layer, b j is the threshold value of each neuron in the output layer, where h = 1, 2, ..., m, m is the number of neurons in the hidden layer, j = 1, 2, ..., q, q is the number of neurons in the output layer; x i (k) represents the i-th sample in the k-th input sample set x(k), where x(k) = [x1(k), x2(k), ..., x n (k)];d j (k) represents the expected output of the k-th input sample set x(k); f1 is the hidden layer activation function, and f2 is the output layer activation function; the error function E is defined, and the connection weight between the hidden layer and the output layer, the output layer threshold, the connection weight between the input layer and the hidden layer, and the correction amount of the hidden layer threshold are calculated according to the error function E. The expression formula is as follows: In the formula, η is the correction coefficient, δ j (k) is the sensitivity of the output layer, δ h (k) is the sensitivity of the hidden layer; f1' is the derivative of the hidden layer activation function, and f2' is the derivative of the output layer activation function; the corresponding connection weights and thresholds are adjusted according to the above correction amount, and the actual output value d of the BP network model is calculated according to the adjusted connection weights and thresholds i and the expected output value y i , the error function fit, and determine the optimal positions of the input layer, hidden layer, and output layer in the BP network model according to the error function fit; the calculation formula of the error function fit is as follows: Where N is the number of training samples.

7. The bp neural network prediction model according to claim 6, characterized in that: The specific dissolved oxygen prediction process is as follows: s1. Collect a large amount of training data from the actual breeding environment, including parameters such as pH value, ammonia nitrogen, temperature, salinity, light, and the dissolved oxygen required by the corresponding white shrimp; After the collection is completed, the collected abnormal data is further checked to remove outliers, duplicate values ​​and incomplete records; 70% of the collected data is further used as a training set, and the remaining 30% is used as a test set; the data is further normalized to scale each feature to between 0 and 1; the normalized training set data is further divided into input and output. s2. Construct the BP neural network structure: further set the number of neurons in the input layer to 5, corresponding to pH value, ammonia nitrogen, temperature, salinity and light respectively; further use one hidden layer, the number of neurons is 10, and the activation function is ReLU; further the number of neurons in the output layer is 1, indicating the predicted dissolved oxygen required by white shrimp, and the activation function selects the linear activation function; further use the mean square error (MSE) as the loss function; further use the stochastic gradient descent optimization algorithm. s3, training bp neural network: further use smaller random numbers to initialize the weights w and bias b of the neural network; further set the learning rate, number of iterations and batch size; further for each iteration, divide the training data into small batches of the batch size, and further for each small batch, forward propagation: calculate the output of the neural network based on the current weights and biases; calculate the loss: use the mean square error (MSE) loss function to calculate the difference between the predicted value and the actual value; Back propagation: calculate the gradient of the loss function with respect to weights and biases; update weights and biases: update weights and biases according to gradients and learning rates; further calculate the loss values ​​of the training set and test set to evaluate the performance of the model; further, if the stopping condition is met (such as reaching the maximum number of iterations or the test set loss value is less than the preset threshold), stop training. s4. Predict dissolved oxygen: The new water quality parameters collected by the water quality detection module are input, and the trained neural network is further used for forward propagation to obtain the predicted dissolved oxygen required by the white shrimp; the prediction structure is further denormalized, and the prediction result is denormalized from the range of 0-1 to the range of actual dissolved oxygen.

8. The communication module according to claim 1, characterized in that: The communication module is built into the intelligent control system, and realizes 5G wireless communication with the host computer through the cloud server. The further communication module can transmit information such as the remaining battery power, real-time water quality data, and water scene to the host computer in real time.

9. The oxygen enrichment device according to claim 1, characterized in that: The oxygenation device is composed of a blower, a main pipe, a branch pipe, and a microporous plastic hose. The blower is mounted on a carrier plate and connected to the main pipe. The main pipe is fixed to the carrier plate through a support rod, and is provided with four branch pipe openings for connecting the branch pipes. The microporous plastic pipe is placed at the bottom of the pond through the branch pipe. The plastic pipe is provided with micropores. When pressurized air passes through the plastic pipe, tiny bubbles are discharged. In the process of the bubbles rising, water movement is formed, and part of the oxygen is dissolved in the water to achieve the purpose of oxygenation.

Images