Method and device for predicting dissolved oxygen in crab pond based on neural network

Abstract

The invention discloses a crab pond dissolved oxygen prediction method and device based on a neural network, and belongs to the technical field of crab pond dissolved oxygen prediction. Cleaning the meteorological data in a previous period of time, and combining the cleaned meteorological data in the previous period of time; six indexes of each month, each day, each time point, the air temperature of each time point, the air pressure of each time point and the humidity of each time point in a previous period are selected as main basis for predicting the dissolved oxygen; dividing the sample set into a training set and a test set according to a set quantitative proportion; sending the training set and the test set into a training model for training; in the crab pond culture process, the current dissolved oxygen amount of the crab pond is predicted, and an oxygenation pump is started when the dissolved oxygen amount of the crab pond is too low; therefore, the production cost of a farmer is reduced, and the farmer can scientifically breed crabs without purchasing a dissolved oxygen probe with thousands of values; and the neural network technology is combined, so that the current dissolved oxygen concentration of the crab pond can be accurately predicted.

Description

Technical Field

[0001] The invention belongs to the technical field of crab pond dissolved oxygen prediction, and in particular relates to a method and device for predicting crab pond dissolved oxygen based on a neural network. Background Art

[0002] In river crab farming, the dissolved oxygen content of the crab pond is a key water quality indicator, which directly affects the growth and health of river crabs. The water quality standard for crab ponds used for river crab farming requires that the dissolved oxygen content be maintained at an appropriate value to ensure that river crabs can feed, move and grow normally. If the dissolved oxygen content is lower than the appropriate value for a long time, river crabs may experience problems such as decreased feeding, slow growth, chronic enteritis, bacterial infection, etc.

[0003] Digital aquaculture is currently being vigorously developed, and there are many technical problems in aquaculture that need to be solved. During the management process, crab pond farmers must pay close attention to the current dissolved oxygen level in the crab pond to ensure that it is sufficient to support the growth needs of crabs. To detect the dissolved oxygen level in the crab pond where river crabs are located, the existing technical solution with patent publication number "CN106259082A" is often used to detect the dissolved oxygen level in the crab pond where river crabs are located. Based on this consideration, it can be determined whether it is necessary to start the oxygen pump to maintain the ideal dissolved oxygen concentration in the crab pond.

[0004] Defects and shortcomings of the existing technology:

[0005] 1. Given that the cost of dissolved oxygen probes is significantly high, this high cost undoubtedly poses a certain economic pressure on the majority of crab pond farmers, making it difficult for them to bear.

[0006] 2. In view of the fragility of dissolved oxygen probes in practical applications, combined with the actual production environment of crab pond farmers, faced with complex and changeable water quality conditions and biological activities, dissolved oxygen probes are prone to functional failure, damage and other adverse situations, which causes certain troubles to production activities.

[0007] 3. Given the high cost of dissolved oxygen probes, their large-scale promotion is significantly restricted. Therefore, crab pond farmers currently still rely mainly on experience when deciding whether to start the aeration pump, which leads to instability in crab production and quality. Summary of the invention

[0008] In order to solve the defects in the prior art, the present invention proposes a method and device for predicting dissolved oxygen in crab ponds based on neural networks, which collect meteorological data provided by the Meteorological Bureau; clean the meteorological data in the past period, and merge the cleaned meteorological data in the past period; select six indicators of each month, each day, each time point, temperature at each time point, air pressure at each time point, and humidity at each time point in the past period as the main basis for predicting dissolved oxygen; divide the sample set into a training set and a test set according to a set quantity ratio; send the training set and the test set to a training model for training; in crab pond breeding, the current dissolved oxygen content of the crab pond is predicted based on sending the currently collected meteorological data to a high-level model, and an oxygenation pump is turned on when the dissolved oxygen content in the crab pond is too low; thereby reducing the production cost of farmers, and farmers do not need to purchase dissolved oxygen probes worth thousands of dollars to scientifically breed crabs; combined with neural network technology, the current dissolved oxygen concentration in the crab pond can be accurately predicted.

[0009] The present invention uses the following technical solutions.

[0010] A method for predicting dissolved oxygen in crab ponds based on neural networks, comprising:

[0011] Step 1: Collect meteorological data provided by the Meteorological Bureau;

[0012] Step 2: Clean the meteorological data in the past period and merge the cleaned meteorological data in the past period;

[0013] Step 3: Select six indicators of each month, each day, each time point, temperature at each time point, air pressure at each time point, and humidity at each time point in the past period as the main basis for predicting dissolved oxygen;

[0014] Step 4: Divide the sample set into training set and test set according to the set quantity ratio;

[0015] Step 5: Send the training set and test set to the training model for training;

[0016] Step 6: In crab pond farming, the current dissolved oxygen content in the crab pond is predicted based on the currently collected meteorological data sent to the high-level model, and the oxygen pump is turned on when the dissolved oxygen content in the crab pond is too low.

[0017] Preferably, in step 1, the meteorological data provided by the Meteorological Bureau includes meteorological data of the crab pond where the river crabs are located in the past period of time, and the meteorological data includes the temperature, air pressure and humidity at each time point in the period.

[0018] Preferably, in step 2, the method for cleaning the meteorological data in the past period includes: sequentially cleaning the meteorological data in the past period using a binning method, a clustering method and a regression method, so as to obtain the cleaned meteorological data in the past period.

[0019] Preferably, the length of the past period may be several years.

[0020] Preferably, step 3 also includes obtaining the dissolved oxygen content of the crab pond where the river crabs are located at each time point, and the six indicators plus the dissolved oxygen content of the crab pond where the river crabs are located at each time point constitute a sample set.

[0021] Preferably, the method for obtaining the dissolved oxygen content of the crab pond where the river crabs are located at each time point is: using the iodine titration method to obtain the dissolved oxygen content of the crab pond where the river crabs are located at each time point.

[0022] Preferably, in step 4, the set quantity ratio is 4:1, that is, the quantity ratio of the training set and the test set into which the sample set is divided is 4:1.

[0023] Preferably, in step 5, the training models include four major models: machine forest regression, LightGBM regression, XGBoost regression and CatBoost regression.

[0024] Preferably, step 5 specifically includes: using the four major models of random forest regression, LightGBM regression, XGBoost regression and CatBoost regression in the training model to perform model fusion.

[0025] Preferably, the method of using the four major models of random forest regression, LightGBM regression, XGBoost regression and CatBoost regression in the training model to perform model fusion specifically includes:

[0026] Different basic models were obtained by training the training sets using CatBoost, LightGBM, XGBoost, random forest, and DNN-BN respectively; finally, the test sets were input into the basic models respectively to obtain the output results, and the results were used as input. The dissolved oxygen content of the crab pond in the test set was used as the training target, and the high-level model was trained using the logistic regression algorithm.

[0027] Preferably, in step 6, starting the aeration pump when the dissolved oxygen content in the crab pond is too low means: when the dissolved oxygen content in the crab pond is lower than 5 mg / L, starting the aeration pump to oxygenate the crab pond.

[0028] A device for predicting dissolved oxygen in a crab pond based on a neural network, comprising:

[0029] A controller and an oxygenation pump connected thereto for oxygenating the crab pond;

[0030] The modules running on the controller include:

[0031] A collection module, which is used to collect meteorological data provided by the Meteorological Bureau;

[0032] A merging module, which is used to clean the meteorological data in the past period and merge the cleaned meteorological data in the past period;

[0033] A selection module is used to select six indicators of each month, each day, each time point, temperature at each time point, air pressure at each time point, and humidity at each time point in the past period as the main basis for predicting dissolved oxygen;

[0034] A partitioning module is used to divide the sample set into a training set and a test set according to a set quantity ratio;

[0035] A training module, which is used to send the training set and the test set to the training model for training;

[0036] The prediction module is used in crab pond farming to predict the current dissolved oxygen content in the crab pond based on sending the currently collected meteorological data into the high-level model, and to start the oxygenation pump when the dissolved oxygen content in the crab pond is too low.

[0037] The beneficial effects of the present invention are that, compared with the prior art, the technical effects of the present invention include:

[0038] 1. It reduces the production costs of farmers. Farmers can scientifically breed crabs without having to purchase dissolved oxygen probes worth thousands of dollars.

[0039] 2. Combined with neural network technology, it can accurately predict the current dissolved oxygen concentration in the crab pond. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a flow chart of the method for predicting dissolved oxygen in crab ponds based on neural network described in the present invention;

[0041] Figure 2 It is a partial structural diagram of the device for predicting dissolved oxygen in crab ponds based on neural network described in the present invention;

[0042] Figure 3 This is the error distribution diagram of two dissolved oxygen probes detecting dissolved oxygen in crab ponds;

[0043] Figure 4 It is the error distribution map of detecting dissolved oxygen in crab ponds using only air temperature;

[0044] Figure 5 It is the error distribution diagram of detecting dissolved oxygen in crab ponds using only air pressure;

[0045] Figure 6 It is the error distribution diagram of detecting dissolved oxygen in crab ponds using only humidity;

[0046] Figure 7 It is an error distribution diagram of the method for predicting dissolved oxygen in crab ponds based on neural network described in the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the present invention clearer, the following will be combined with the drawings in the embodiments of the present invention to clearly and completely express the technical solution of the present invention. The embodiments expressed in this application are only some embodiments of the present invention, not all embodiments. According to the spirit of the present invention, other embodiments obtained by ordinary technicians in this field without creative work are all within the protection scope of the present invention.

[0048] like Figure 1 As shown, the method for predicting dissolved oxygen in crab ponds based on neural network of the present invention comprises:

[0049] Step 1: Collect meteorological data provided by the Meteorological Bureau;

[0050] In a preferred but non-limiting embodiment of the present invention, in step 1, the meteorological data provided by the Meteorological Bureau includes meteorological data of the crab pond where the river crabs are located in the past period of time, and the meteorological data includes the temperature, air pressure and humidity at each time point in the period.

[0051] Step 2: Clean the meteorological data in the past period and merge the cleaned meteorological data in the past period;

[0052] In a preferred but non-limiting embodiment of the present invention, in step 2, the method for cleaning the meteorological data in the past period includes: sequentially using binning, clustering and regression methods to clean the meteorological data in the past period, so as to obtain the cleaned meteorological data in the past period.

[0053] In a preferred but non-limiting embodiment of the present invention, the duration of the past period may be several years, such as from 2019 to 2023.

[0054] Step 3: Considering the difficulty of measurement in actual production, six indicators, namely, temperature at each month, day, time, air pressure at each time, and humidity at each time in the past period, are selected as the main basis for predicting dissolved oxygen;

[0055] In a preferred but non-limiting embodiment of the present invention, step 3 also includes obtaining the dissolved oxygen content of the crab pond where the river crabs are located at each time point, and the six indicators plus the dissolved oxygen content of the crab pond where the river crabs are located at each time point constitute a sample set.

[0056] In a preferred but non-limiting embodiment of the present invention, the method for obtaining the dissolved oxygen content of the crab pond where the river crabs are located at each time point is: using iodine titration to obtain the dissolved oxygen content of the crab pond where the river crabs are located at each time point.

[0057] Step 4: Divide the sample set into training set and test set according to the set quantity ratio;

[0058] In a preferred but non-limiting embodiment of the present invention, in step 4, the set quantity ratio is 4:1, that is, the quantity ratio of the training set and the test set into which the sample set is divided is 4:1.

[0059] Step 5: Send the training set and test set to the training model for training;

[0060] In a preferred but non-limiting embodiment of the present invention, in step 5, the training models include four major models: machine forest regression, LightGBM regression, XGBoost regression and CatBoost regression.

[0061] In a preferred but non-limiting embodiment of the present invention, step 5 specifically includes: using the four major models of random forest regression, LightGBM regression, XGBoost regression and CatBoost regression in the training model to perform model fusion to improve the accuracy of prediction.

[0062] In a preferred but non-limiting embodiment of the present invention, the method for model fusion using four major models of random forest regression, LightGBM regression, XGBoost regression and CatBoost regression in the training model specifically includes:

[0063] Different basic models were obtained by training the training sets using CatBoost, LightGBM, XGBoost, random forest, and DNN-BN respectively; finally, the test sets were input into the basic models respectively to obtain the output results, and the results were used as input. The dissolved oxygen content of the crab pond in the test set was used as the training target, and the high-level model was trained using the logistic regression algorithm.

[0064] In practice, model evaluation is performed on the high-level model. In the actual production process, the error of predicted dissolved oxygen content is within 2 in 95% of the cases, that is, at a confidence level of 95%, the absolute error is basically within 2%.

[0065] Step 6: In crab pond farming, the current dissolved oxygen content in the crab pond is predicted based on the currently collected meteorological data sent to the high-level model, and the oxygen pump is turned on when the dissolved oxygen content in the crab pond is too low.

[0066] The present invention mainly realizes the real-time prediction of dissolved oxygen in crab ponds by combining neural network technology, reduces the production cost of farmers, and accelerates the process of digital aquaculture.

[0067] In a preferred but non-limiting embodiment of the present invention, in step 6, starting the aeration pump when the dissolved oxygen content in the crab pond is too low is: when the dissolved oxygen content in the crab pond is lower than 5 mg / L, starting the aeration pump to oxygenate the crab pond.

[0068] like Figure 2 As shown, the device for predicting dissolved oxygen in crab ponds based on neural networks of the present invention comprises:

[0069] A controller and an oxygenation pump connected thereto for oxygenating the crab pond;

[0070] The modules running on the controller include:

[0071] A collection module, which is used to collect meteorological data provided by the Meteorological Bureau;

[0072] A merging module, which is used to clean the meteorological data in the past period and merge the cleaned meteorological data in the past period;

[0073] A selection module is used to select six indicators of each month, each day, each time point, temperature at each time point, air pressure at each time point, and humidity at each time point in the past period as the main basis for predicting dissolved oxygen;

[0074] A partitioning module is used to divide the sample set into a training set and a test set according to a set quantity ratio;

[0075] A training module, which is used to send the training set and the test set to the training model for training;

[0076] The prediction module is used in crab pond farming to predict the current dissolved oxygen content in the crab pond based on sending the currently collected meteorological data into the high-level model, and to start the oxygenation pump when the dissolved oxygen content in the crab pond is too low.

[0077] exist Figure 3 In actual situations, the actual error of the two dissolved oxygen probes is within 2 (the absolute error is within 2%), accounting for 40%, which shows that the error is very large.

[0078] exist Figures 4 to 7 In the forecast results based on temperature, air pressure and humidity, the error compared with the actual situation is within 2, accounting for about 94%, and the error compared with the actual situation is within 1, accounting for about 70%. The increase in the number of indicators cannot significantly improve the accuracy of the forecast results. In actual situations, any indicator of temperature, humidity or atmospheric pressure can have a certain forecast effect, but considering that a single indicator may fail in the actual production process, multiple indicators can alleviate this problem to a certain extent.

[0079] The beneficial effects of the present invention are that, compared with the prior art, the technical effects of the present invention include:

[0080] 1. It reduces the production costs of farmers. Farmers can scientifically breed crabs without having to purchase dissolved oxygen probes worth thousands of dollars.

[0081] 2. Combined with neural network technology, it can accurately predict the current dissolved oxygen concentration in the crab pond.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not deviate from the spirit and scope of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A method for predicting dissolved oxygen in crab ponds based on neural networks, characterized in that: include: Step 1: Collect meteorological data provided by the Meteorological Bureau; Step 2: Clean the meteorological data in the past period and merge the cleaned meteorological data in the past period; Step 3: Select six indicators of each month, each day, each time point, temperature at each time point, air pressure at each time point, and humidity at each time point in the past period as the main basis for predicting dissolved oxygen; Step 4: Divide the sample set into training set and test set according to the set quantity ratio; Step 5: Send the training set and test set to the training model for training; Step 6: In crab pond farming, the current dissolved oxygen content in the crab pond is predicted based on the currently collected meteorological data sent to the high-level model, and the oxygen pump is turned on when the dissolved oxygen content in the crab pond is too low.

2. The method for predicting dissolved oxygen in crab ponds based on neural networks according to claim 1, characterized in that: In step 1, the meteorological data provided by the Meteorological Bureau includes the meteorological data of the crab pond where the river crabs are located in the past period of time, and the meteorological data includes the temperature, air pressure and humidity at each time point in the period.

3. The method for predicting dissolved oxygen in crab ponds based on neural networks according to claim 2, characterized in that: In step 2, the method for cleaning the meteorological data in the past period includes: sequentially cleaning the meteorological data in the past period using a binning method, a clustering method, and a regression method, so as to obtain the cleaned meteorological data in the past period; In the past, a period of time could be several years in length.

4. The method for predicting dissolved oxygen in crab ponds based on neural networks according to claim 3, characterized in that: Step 3 also includes obtaining the dissolved oxygen content of the crab pond where the river crabs are located at each time point, and the six indicators plus the dissolved oxygen content of the crab pond where the river crabs are located at each time point constitute a sample set; The method for obtaining the dissolved oxygen content of the crab pond where the river crabs are located at each time point is: using the iodine titration method to obtain the dissolved oxygen content of the crab pond where the river crabs are located at each time point.

5. The method for predicting dissolved oxygen in crab ponds based on neural networks according to claim 4, characterized in that: In step 4, the ratio is set to 4:1, that is, the ratio of the training set and the test set into which the sample set is divided is 4:

1.

6. The method for predicting dissolved oxygen in crab ponds based on neural networks according to claim 5, characterized in that: In step 5, the training models include machine forest regression, LightGBM regression, XGBoost regression and CatBoost regression.

7. The method for predicting dissolved oxygen in crab ponds based on neural networks according to claim 6, characterized in that: Step 5 specifically includes: using the four major models of random forest regression, LightGBM regression, XGBoost regression and CatBoost regression in the training model to perform model fusion.

8. The method for predicting dissolved oxygen in crab ponds based on neural networks according to claim 7, characterized in that: The method of model fusion using the four major models of random forest regression, LightGBM regression, XGBoost regression and CatBoost regression in the training model includes: Different basic models were obtained by training the training sets using CatBoost, LightGBM, XGBoost, random forest, and DNN-BN respectively; finally, the test sets were input into the basic models respectively to obtain the output results, and the results were used as input. The dissolved oxygen content of the crab pond in the test set was used as the training target, and the high-level model was trained using the logistic regression algorithm.

9. The method for predicting dissolved oxygen in crab ponds based on neural networks according to claim 8, characterized in that: In step 6, starting the aeration pump when the dissolved oxygen content in the crab pond is too low means: when the dissolved oxygen content in the crab pond is lower than 5 mg / L, starting the aeration pump to oxygenate the crab pond.

10. A device for predicting dissolved oxygen in crab ponds based on neural networks, characterized in that: include: A controller and an oxygenation pump connected thereto for oxygenating the crab pond; The modules running on the controller include: A collection module, which is used to collect meteorological data provided by the Meteorological Bureau; A merging module, which is used to clean the meteorological data in the past period and merge the cleaned meteorological data in the past period; A selection module is used to select six indicators of each month, each day, each time point, temperature at each time point, air pressure at each time point, and humidity at each time point in the past period as the main basis for predicting dissolved oxygen; A partitioning module is used to divide the sample set into a training set and a test set according to a set quantity ratio; A training module, which is used to send the training set and the test set to the training model for training; The prediction module is used in crab pond farming to predict the current dissolved oxygen content in the crab pond based on sending the currently collected meteorological data into the high-level model, and to start the oxygenation pump when the dissolved oxygen content in the crab pond is too low.

Images