JP2023061400A - Aquaculture support system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To predict a state of marine products.

SOLUTION: A system for supporting aquaculture of marine products includes: a model storage part for storing a learning model created by machine learning, with the number of aquaculture days, statistical values of water quality data and an amount of feed consumption defined as input data and a weight of marine products defined as training data; a water quality prediction part for acquiring prediction values for the water quality data; a feed consumption prediction part for acquiring prediction values of the amount of feed consumption; and a weight prediction part for giving the number of aquaculture days of prediction targets, the acquired statistical values of the water quality data, and prediction values of the amount of consumption to the learning model to predict the weight.

SELECTED DRAWING: Figure 1

COPYRIGHT: (C)2023,JPO&INPIT

Description

The present invention relates to an aquaculture support system.

Aquaculture is managed by a system (see Patent Document 1, for example).

Japanese Patent Application Laid-Open No. 2021-061782

Predicting the growth rate of farmed seafood has been based on experience.

The present invention has been made in view of such a background, and an object of the present invention is to provide a technique capable of predicting the state of marine products.

The main invention of the present invention for solving the above problems is a system for supporting aquaculture of marine products, wherein the number of days of cultivation, statistical values of water quality data, and feed consumption are used as input data, and the weight of marine products is used as teacher data. A model storage unit that stores a learning model created by machine learning, a water quality prediction unit that obtains a predicted value for the water quality data, a feed consumption prediction unit that obtains the predicted value of the feed consumption, a weight prediction unit that predicts the weight by providing the learning model with the number of farming days to be predicted, the statistical value of the obtained predicted value of the water quality data, and the predicted value of the consumption amount. .

Other problems disclosed by the present application and solutions thereof will be clarified by the section of the embodiment of the invention and the drawings.

According to the present invention, the state of marine products can be predicted.

2 is a diagram showing a hardware configuration example of the aquaculture support device 10. FIG. 2 is a diagram showing a software configuration example of the aquaculture support device 10. FIG. It is a figure explaining the flow of the process which predicts the weight of marine products. FIG. 3 is a diagram showing a flow of processing for obtaining an estimated value (estimated weight) of the weight of marine products. FIG. 4 is a diagram showing the flow of biomass prediction processing; It is a figure which shows the flow of calculation processing of a proposal yield. FIG. 10 is a diagram showing the flow of latent history prediction processing;

<Overview of the invention>
The contents of the embodiments of the present invention are listed and explained. The present invention has, for example, the following configurations.
[Item 1]
A system for supporting aquaculture of marine products,
A model storage unit that stores a learning model created by machine learning using the number of days of farming, statistical values of water quality data, and feed consumption as input data and the weight of fishery products as teacher data;
a water quality prediction unit that acquires a predicted value for the water quality data;
a feed consumption prediction unit that acquires a predicted value of the feed consumption;
a weight prediction unit that predicts the weight by providing the learning model with the number of days of farming to be predicted, the statistical value of the predicted value of the obtained water quality data, and the predicted value of the consumption amount;
An aquaculture support system comprising:
[Item 2]
The aquaculture support system according to item 1,
The weight prediction unit predicts the weight for a plurality of the farming days,
Further comprising a weight adjusting unit that acquires actual values of the weight of the marine product corresponding to each of the plurality of farming days, and corrects the predicted weight value according to a difference between the actual value and the predicted weight value. matter,
An aquaculture support system characterized by:
[Item 3]
The aquaculture support system according to item 1 or 2,
A value obtained by subtracting the amount of the marine product that died by the farming days to be predicted from the amount of the marine product that was put into the aquaculture pond on the cultivation start date and the amount corresponding to the predetermined mortality rate, further comprising a biomass prediction unit that calculates the biomass of the aquaculture pond by multiplying the predicted value;
An aquaculture support system characterized by:
[Item 4]
The aquaculture support system according to item 3,
further comprising a yield proposal unit that proposes harvesting of the marine product when the biomass corresponding to the number of days of farming to be predicted is equal to or greater than the capacity of the farming pond;
An aquaculture support system characterized by:
[Item 5]
The aquaculture support system according to item 4,
The harvest amount proposing unit provides a first predicted value of the weight related to the first number of days of farming, which is the number of days of farming to be predicted, and the weight related to the second number of days of farming after the first number of days of farming. The ratio of the second predicted value exceeds a predetermined value, and the second number of farming days that specifies the first predicted value of the first number of farming days is specified. calculating a proposed yield by multiplying the biomass by 1 minus the value divided by the second predicted value, and outputting the calculated yield;
An aquaculture support system characterized by:

<Overview>
The aquaculture support system according to this embodiment will be described below. The aquaculture support system of the present embodiment calculates parameters related to aquaculture such as shrimp farming, and provides them to aquaculture workers. Parameters include fish weight, biomass, survival rate, optimal feeding, harvest recommendations, and potential profit based on feeding frequency, pond area, juvenile abundance, actual feeding and weight sampling, etc. offer.

In this embodiment, the aquaculture support system includes an aquaculture support device 10 . The aquaculture support device 10 may be, for example, a general-purpose computer such as a workstation or personal computer, or may be logically realized by cloud computing. In addition, the aquaculture support device 10 is communicably connected to other computers such as user terminals, sensors, and the like via a communication network. The communication network is, for example, the Internet, and is constructed by a public telephone line network, a mobile telephone line network, a wireless communication path, Ethernet (registered trademark), and the like. The aquaculture support device 10 can, for example, receive input of various data from a user terminal or acquire it from a sensor.

<Hardware configuration example>
FIG. 1 is a diagram showing a hardware configuration example of an aquaculture support device 10. As shown in FIG. Note that the illustrated configuration is an example, and other configurations may be employed. The aquaculture support device 10 includes a CPU 101 , a memory 102 , a storage device 103 , a communication interface 104 , an input device 105 and an output device 106 . The storage device 103 is, for example, a hard disk drive, solid state drive, flash memory, etc., which stores various data and programs. The communication interface 104 is an interface for connecting to a communication network. Examples include a USB (Universal Serial Bus) connector and an RS232C connector for serial communication. The input device 105 is, for example, a keyboard, a mouse, a touch panel, a button, a microphone, etc. for inputting data. The input device 105 may receive input from an external device such as a user terminal via the communication interface 104, for example. The output device 106 is, for example, a display, printer, speaker, or the like that outputs data. Each functional unit of the aquaculture support device 102 described later is realized by the CPU 101 reading out a program stored in the storage device 103 into the memory 102 and executing it. and part of the storage area provided by the storage device 103 .

FIG. 2 is a diagram showing a software configuration example of the aquaculture support device 10. As shown in FIG. The aquaculture support device 10 includes a water quality data acquisition unit 110, a water quality prediction unit 111, a feed consumption amount prediction unit 112, a weight prediction unit 113, a weight adjustment unit 114, a biomass prediction unit 115, a feed amount prediction unit 116, and a proposed yield calculation. Unit 117, potential profit prediction unit 118, unit price prediction unit 119, weight prediction model storage unit 131, setting information storage unit 132, water quality data storage unit 150, weight performance storage unit 151, death performance storage unit 152, fry harvest storage unit 153 , a feeding performance storage unit 154 , a harvest performance storage unit 155 , and a sales performance storage unit 156 .

<Memory unit>
The weight prediction model storage unit 131 stores a prediction model for predicting the weight of marine products (for example, shrimp). The weight prediction model can be a learning model learned by machine learning using an MLP network, etc., with data on the number of days of farming, water quality and feed consumption as input data, and weight per unit amount of marine products as teacher data. . Parameters representing water quality data may include temperature, dissolved oxygen content, salinity, pH, and the like. These parameters can be measured, for example, by installing sensors in the pond. In addition, average, variance, deterioration probability, etc. can be calculated for these parameters indicating water quality. Food consumption data are measurements of the amount of food actually consumed. For the amount of feed consumed, for example, subtract the amount of feed remaining, which is estimated by multiplying the amount of feed per unit amount detected in the pond by the capacity of the pond, from the amount of feed fed (feed amount). can be calculated by

The setting information storage unit 132 stores various setting information. The set information can include the start date of aquaculture, the initial amount of juvenile fish, the standard feed amount, the average feeding unit price, the maximum breeding amount, and the set survival rate. The farming start date is the date on which farming of marine products (for example, shrimp) was started in the farming pond. The initial amount of juvenile fish is the amount of juvenile fish introduced on the start date of aquaculture. The amount of fry can be expressed, for example, by weight (kg). The standard feed amount is the standard feed amount given per day per unit amount of juvenile fish. The average feed unit price is the average price (cost) per unit amount of feed. The maximum rearing capacity is the maximum amount of fish that can be reared in the pond. The set survival rate is the assumed value (%) of the survival rate of the fishery product at the time of the target farming days to be predicted (hereinafter referred to as the target farming days). It is assumed that setting information is stored in the setting information storage unit 132 in advance by accepting an input of the setting information from the user.

The water quality data storage unit 150 stores information necessary for predicting water quality data of aquaculture ponds. In this embodiment, the water quality data storage unit 150 stores actual values of water quality data. The water quality data storage unit 150 can store water quality data in association with dates. Water quality data includes, for example, temperature, dissolved oxygen content, salinity, and pH. The water quality data storage unit 150 can also store a prediction model for predicting future water quality data as information for prediction based on the measured water quality data.

The actual weight storage unit 151 stores the actual value obtained by measuring the weight per fish (for example, shrimp) or per unit amount (for example, 1 kg). The actual weight storage unit 151 can store the measured weight in association with the date.

The actual death record storage unit 152 stores the actual amount (death amount) of dead marine products (for example, shrimp). The death record storage unit 152 can store the amount of death in association with the date.

The juvenile fish harvest amount storage unit 153 stores the actual amount (juvenile fish harvest amount) of the marine products being raised. In the present embodiment, marine products that have not reached the specified weight for shipment are referred to as "fry fish". The fry harvested amount storage unit 153 stores the fry harvested amount in association with the date.

The actual feed storage unit 154 stores the actual amount of feed given (feed amount). The feeding record storage unit 154 can store the feeding amount in association with the date.

The harvest record storage unit 155 stores the actual amount (harvest amount) of harvested (shipped) marine products. The harvest performance storage unit 155 can store harvest amounts in association with dates.

The sales record storage unit 156 stores the actual amount (sales amount) of the sold marine products. The sales record storage unit 156 can store sales amounts in association with dates. In this embodiment, the sales record storage unit 156 stores the unit sales price per predetermined unit of weight (for example, 1 kg), the weight, and the sales amount (unit price x weight + other fees, etc.) in association with the date. shall be

<Function part>
The water quality data acquisition unit 110 stores data used for prediction of water quality data. In this embodiment, the water quality data acquisition unit 110 acquires water quality data of aquaculture ponds. The water quality data acquisition unit 110 may, for example, directly acquire water quality data from a sensor, or receive an input of a file in which water quality data is stored or a designation of the location of the file, and read the water quality data from the file. can be The water quality data acquisition unit 110 can register the acquired water quality data in the water quality data storage unit 150 in association with the measurement date.

The water quality prediction unit 111 acquires predicted values of water quality data of the farm. In this embodiment, the water quality prediction unit 111 can predict future water quality data based on the water quality data stored in the water quality data storage unit 150 . The water quality prediction unit 111 can receive the prediction target number of farming days or the prediction target date as an argument. The water quality prediction unit 111 can predict water quality data based on past water quality data, for example, by a known technique such as time series analysis. The water quality prediction unit 111, for example, divides the water quality data into day and night time zones, and analyzes the trend of water quality parameters from the past 10 days of data consisting of different data to obtain trend analysis data and signal white noise by a linear model. can do. The water quality prediction unit 111 can use an SVM (Support Vector Machine) model to predict the random noise of the water quality parameters due to the received white noise. Also, the water quality prediction unit 111 can analyze actual values of water quality parameters from the predicted noise.

The feed consumption prediction unit 112 acquires a predicted value of feed consumption. The feed consumption prediction unit 112 can, for example, assume that 100% of the feed is always consumed, and use the feed amount obtained by multiplying the standard feed amount by the initial juvenile fish amount as the consumption amount prediction value. In addition, in the present embodiment, the feed consumption prediction unit 112 sets the consumption rate in advance, and multiplies the feeding amount by the predetermined consumption rate. The amount of feed consumed can also be calculated based on actual feed consumption, for example, by time-series prediction. For example, the feed consumption prediction unit 112 takes out some fishery products (shrimp) by a method of feeding sampling, and calculates the amount of feed that was put in and the amount of feed that remains in the water after a predetermined time (for example, after 1 hour). can be manually investigated to obtain the feed consumption rate, and the amount of feed fed is multiplied by the consumption rate to calculate the consumption amount. The feed consumption prediction unit 112 can also calculate the consumption per unit amount of juvenile fish.

The weight prediction unit 113 predicts the weight of marine products. The weight predictor 113 can use a growth model of seafood to estimate the weight and compensate the output based on the next day's actual sampling. It is assumed that the weight prediction unit 113 is given the number of culture days as a parameter. The details of the processing for calculating the predicted weight value will be described later.

The weight adjustment unit 114 can correct the weight of the seafood predicted by the weight prediction unit 113 . The weight adjustment unit 114 can perform adjustment according to the difference between the actual measured fishery product weight and the predicted value. The details of the processing for calculating the value (estimated amount) obtained by correcting the predicted weight value will be described later.

The biomass prediction unit 115 calculates a predicted value (estimated biomass) of the biomass (biomass) of the culture pond. Estimated biomass is calculated using estimated weight and total juvenile mass, including calculation of expected daily mortality, actual mortality, actual feed affecting survival, and recommended and actual yields. calculated by Details of the processing for calculating the estimated biomass will be described later.

The feed amount prediction unit 116 predicts the amount of feed to be given (feed amount). A feed amount prediction unit 116 is supplied with the amount of juvenile fish, and predicts the feed amount corresponding to the amount of juvenile fish. In this embodiment, the feed amount prediction unit 116 calculates the feed amount by multiplying the standard feed amount stored in the setting information storage unit 132 by the fry amount. In addition, when the number of feeding days is given, the feeding amount prediction unit 116 calculates the feeding amount if the feeding amount corresponding to the date after the feeding start date is registered in the feeding result storage unit 154. can be returned.

The proposed yield calculation unit 117 calculates the amount of marine products to be harvested (proposed yield). Proposed yields can be calculated based on a comparison of the estimated amount of biomass and the carrying capacity of the pond. Based on the comparison between the estimated biomass and the carrying capacity of the farming pond, the proposed yield calculation unit 117 can calculate that the portion exceeding the holding capacity of the farming pond should be harvested. The details of the processing for calculating the proposed yield will be described later.

In addition, the proposed harvest amount calculation unit 117 can propose harvesting to the user. The suggested yield calculator 117 can suggest harvesting, for example, when the estimated biomass exceeds the maximum capacity of the pond. At that time, the suggested yield calculation unit 117 can also output the suggested yield.

The potential profit prediction unit 118 calculates the profit expected to be obtained from the sale of marine products. The potential profit prediction unit 118 subtracts the estimated total cost, which is estimated assuming that half of the total cost is the feed cost, from the biomass estimated to be the actual selling price after harvest and the current shrimp price. can be used to calculate potential profit. The details of the processing for calculating the potential profit will be described later.

The unit price prediction unit 119 predicts the sales unit price of marine products. The unit selling price is the price of a predetermined unit of weight (eg, kg). The unit price prediction unit 119 predicts the sales unit price based on the sales unit price stored in the sales record storage unit 156 . The unit price prediction unit 119, for example, re-samples the history of the unit selling price by cubic spline interpolation, obtains weekly data, removes noise using the difference in the data, calculates the moving average, and autoregresses. Analyze terms. The unit price prediction unit 119 can estimate the unit selling price by a Bayesian regression technique using these values.

<Weight Prediction>

FIG. 3 is a diagram illustrating the flow of processing for predicting the weight of marine products.

The weight prediction unit 113 receives the number of days for aquaculture (S301) and acquires water quality data (S302). The weight prediction unit 113 reads out the water quality data stored in the water quality data storage unit 150 from the start date of the farming to the target number of days of farming. can obtain the predicted value predicted by The weight prediction unit 113 calculates statistical values of water quality data (S303). The quantity prediction unit 113 can calculate the total consumption by multiplying the consumption predicted by the feed consumption prediction 112 by the fry quantity (S304). Here, the amount of fry can be the initial amount of fry registered in the setting information storage unit 132, but the amount of fry adjusted according to the mortality according to the survival rate in the same manner as the biomass prediction process described later can also be The weight prediction unit 113 provides the number of days of farming, the statistical value of the water quality data, and the total amount of feed consumed to the prediction model stored in the weight prediction model storage unit 131 to obtain a predicted weight value. Yes (S305).

<Weight adjustment>
FIG. 4 is a diagram showing the flow of processing for obtaining an estimated value (estimated weight) of the weight of marine products.

The weight adjustment unit 114 sets the adjustment coefficient factor to 1, and sets the variable doc of the number of culture days to 0 (S321). The weight adjustment unit 114 reads out the farming start date from the setting information storage unit 132 (S322).

If the actual weight (actual weight [doc-1]) after doc-1 days from the start date of aquaculture exists in the actual weight storage unit 151 (S323: YES), the weight adjustment unit 114 adjusts the actual weight from the start date of aquaculture to doc- The actual weight after one day is divided by the weight prediction value (weight prediction (doc-1)) predicted by giving doc-1 to the weight prediction unit 113, and the result is set as a factor (S324).

The weight adjustment unit 114 multiplies the predicted value of weight (predicted weight (doc)) predicted by giving doc to the weight prediction unit 113 by the factor to obtain the estimated weight (estimated weight [doc]) of the number of farming days of doc. is calculated (S325). The weight adjustment unit 114 increments doc (S326), and repeats the process from step S303 while doc is less than the target culture days.

As described above, the predicted weight value obtained by the weight prediction unit 113 can be corrected according to the actual weight value.

<Calculation of estimated biomass>
FIG. 5 is a diagram showing the flow of biomass prediction processing.

The biomass prediction unit 115 initializes variables, that is, sets each variable of estimated mortality total, actual mortality total, and juvenile harvest total to 0, and calculates each juvenile fish quantity from 0 to the target culture days. The initial amount of fry is set in (amount of fry [doc]) (S341). In addition, the biomass prediction unit 115 sets 0 to the variable doc of the culture days (S342).

The biomass prediction unit 115 divides the set survival rate by 100 and subtracts it from 1, and multiplies it by the initial amount of juvenile fish. The expected value of the mortality rate of juvenile fish is divided by the target farming days. (S343). If the mortality on the doc-1 day from the culture start date is registered in the mortality record storage unit 152 (S344: YES), the biomass prediction unit 115 predicts the mortality on the doc-1 day from the culture start date. is added to the actual mortality total (S345).

If the fry harvest amount on the doc day from the culture start date is registered in the fry harvest amount storage unit 153 (S346: YES), the biomass prediction unit 115 predicts the fry harvest amount on the doc day from the culture start date. , is added to the total amount of juvenile fish harvested (S347).

The biomass prediction unit 115 subtracts the estimated total mortality, the total actual mortality, and the total harvest of juvenile fish from the initial fry volume to determine the juvenile volume on the doc day (S348).

If the amount of feed on the doc-1 day from the culture start date is registered in the feed performance storage unit 154 (S349: YES), the biomass prediction unit 115 predicts the feed according to the estimated weight on the doc-1 day. The survival rate is obtained by dividing the feed amount on the doc-1 day from the start of aquaculture by the predicted amount (can be calculated by multiplying the estimated weight by the standard feed amount) (S350), and the survival rate is calculated as the initial fry amount. The value obtained by subtracting the fry amount on the doc day by the product multiplied by is subtracted to calculate the amount of juvenile fish on the doc day (S352).

The biomass prediction unit 115 calculates the estimated biomass (estimated biomass [doc]) on the doc day by multiplying the estimated weight on the doc day by the fry amount on the doc day (S353), increments doc (S354 ), and the processing from S323 is repeated until doc reaches the target culture days.

As described above, the biomass predicted value (estimated biomass) is calculated.

<Proposed yield>
FIG. 6 is a diagram showing a flow of processing for calculating a proposed yield. The suggested harvest amount calculation unit 117 may output that harvesting is not recommended when the suggested harvest amount is equal to or less than a predetermined value (which may be 0).

The proposed harvest amount calculation unit 117 sets the variable doc for the cultivation days to 0 (S381). S382: YES), the harvest amount registered in the harvest performance storage unit 155 is set for the proposed harvest amount [doc] after doc days from the culture start date (S383).

On the other hand, if the harvest amount after doc days is not registered (S382: NO), and if the estimated biomass after doc days [doc] is greater than or equal to the maximum rearing capacity of the pond (S384: YES), the proposed harvest amount is calculated. The unit 117 initializes the variable i to 0, and initializes the variable next_doc to the target culture days (S385). The proposed harvest amount calculation unit 117 calculates the predicted weight after doc days (predicted value calculated by giving doc to the weight prediction unit 113) to the predicted weight after next_doc days (prediction calculated by giving next_doc to the weight prediction unit 113). value) is 0.5 or less, set the value obtained by dividing the value obtained by subtracting doc from the target culture days by i+1 and the sum of doc to next_doc (S386), and increment i. (S387). The proposed yield calculation unit 117 subtracts the value obtained by dividing the predicted weight after doc days by the predicted weight after next_doc days from 1 to obtain the harvest rate (S388), multiplies the estimated biomass after doc days by the harvest rate, and obtains the harvest rate after doc days. is the proposed yield (proposed yield [doc]) (S389).

If the estimated biomass after doc days is less than the maximum capacity (S384: NO), the proposed harvest amount calculation unit 117 can set the proposed harvest amount after doc days to 0 (S390). It should be noted that the maximum capacity can be determined based on various parameters and can be set by the user based on multiple factors such as pond area, number of turbines, density, and the like. The maximum capacity can be set to 1.25 kg/m2 as a default value, for example.

The proposed yield is calculated as described above. Since next_doc is the optimum number of days for harvesting aquaculture, it is possible to determine the harvest rate according to the predicted value of weight in next_doc.

<Potential profit>
FIG. 7 is a diagram showing the flow of latent history prediction processing.

The potential profit prediction unit 118 initializes the total sales performance and the total feed, which are variables, to 0 (S401), and sets 0 to the variable doc of the culture days (S402).

If the sales amount after doc days from the culture start date is registered in the sales performance storage unit 156 (S403: YES), the potential profit prediction unit 118 adds the sales amount to the total sales performance (S404).

The potential profit prediction unit 118 gives the feed amount prediction unit 116 the amount of juvenile fish on the doc day (the amount of juvenile fish calculated by calculating the estimated biomass) and doc to calculate the feed amount on the doc day. is added to the feeding total (S405). The potential profit prediction unit 118 gives the estimated weight on the doc day to the unit price prediction unit 119 to predict the sales unit price, and multiplies the unit price by the estimated biomass on the doc day to calculate the sales amount (S406).

The potential profit prediction unit 118 calculates estimated sales by adding the sales amount to the actual sales total (S407), multiplies the feed total by the average feed unit price stored in the setting information storage unit 132, and calculates the amount. The doubled amount is set as the estimated total cost (S408), and the estimated profit for the doc day is calculated by subtracting the estimated total cost from the estimated sales (S409).

In the present embodiment, the ratio of feed to the cost is set to 50%, but this is not restrictive, and the ratio of the feed to the cost is registered in the setting information storage unit 132. The estimated total cost can also be calculated by multiplying the average feed unit price and dividing by the ratio.

As described above, the estimated profit (potential profit) can be calculated.

Although the present embodiment has been described above, the above-described embodiment is intended to facilitate understanding of the present invention, and is not intended to limit and interpret the present invention. The present invention can be modified and improved without departing from its spirit, and the present invention also includes equivalents thereof.

For example, in the present embodiment, the set survival rate is used as the survival rate, but the survival rate may be predicted. In this case, the aquaculture support device 10 can be provided with a survival rate prediction unit. In the processing described above, the survival rate predicted by the survival rate prediction unit can be used in all or part of the places where the survival rate is used. The survival rate prediction unit can predict the survival rate using aquaculture information such as the density of marine products in the aquaculture pond, the number of days of aquaculture, the predetermined cycle period, and the start date of aquaculture. It is also possible to use water quality parameters such as temperature, dissolved oxygen, salinity, and day/night pH.

The survival rate predictors can be classified, for example, according to four seasons (DJF, MAM, JA, SON). DJF indicates December, January and February, MAM indicates March, April and May, JA indicates June, July and August, and SON indicates September, October and November. indicates the month. The survival rate prediction part determines whether the temperature is 26 degrees or less or 32 degrees or more, whether the dissolved oxygen is less than 4, whether the salinity is less than 15 or 30 or more, and whether the pH is less than 7.5 or It is possible to determine whether the water quality is good or bad depending on whether it is 8.5 or higher. The survival rate prediction unit can calculate the ratio of the number of days with poor water quality. The survival rate predictor can train a random forest model to predict survival rate with the above variables as input. It is possible to further include a survival rate prediction model storage unit that stores learning models of random forest models. The survival rate prediction unit can predict the survival rate by inputting the four seasons, water quality data, density, number of days of cultivation, etc. into the survival rate prediction model.

110 Water quality data acquisition unit 111 Water quality prediction unit 112 Feed consumption prediction unit 113 Weight prediction unit 114 Weight adjustment unit 115 Biomass prediction unit 116 Feeding amount prediction unit 117 Proposed yield calculation unit 118 Potential profit prediction unit 119 Unit price prediction unit 131 Weight prediction Model storage unit 132 Setting information storage unit 150 Water quality data storage unit 151 Weight performance storage unit 152 Death performance storage unit 153 Fry harvest storage unit 154 Feeding performance storage unit 155 Harvest performance storage unit 156 Sales performance storage unit

Claims (5)

A system for supporting aquaculture of marine products,
A model storage unit that stores a learning model created by machine learning using the number of days of farming, statistical values of water quality data, and feed consumption as input data and the weight of fishery products as teacher data;
a water quality prediction unit that acquires a predicted value for the water quality data;
a feed consumption prediction unit that acquires a predicted value of the feed consumption;
a weight prediction unit that predicts the weight by providing the learning model with the number of days of farming to be predicted, the statistical value of the predicted value of the obtained water quality data, and the predicted value of the consumption amount;
An aquaculture support system comprising: The aquaculture support system according to claim 1,
The weight prediction unit predicts the weight for a plurality of the farming days,
Further comprising a weight adjusting unit that acquires actual values of the weight of the marine product corresponding to each of the plurality of farming days, and corrects the predicted weight value according to a difference between the actual value and the predicted weight value. matter,
An aquaculture support system characterized by: The aquaculture support system according to claim 1 or 2,
A value obtained by subtracting the amount of the marine product that died by the farming days to be predicted from the amount of the marine product that was put into the aquaculture pond on the cultivation start date and the amount corresponding to the predetermined mortality rate, further comprising a biomass prediction unit that calculates the biomass of the aquaculture pond by multiplying the predicted value;
An aquaculture support system characterized by: The aquaculture support system according to claim 3,
further comprising a yield proposal unit that proposes harvesting of the marine product when the biomass corresponding to the number of days of farming to be predicted is equal to or greater than the capacity of the farming pond;
An aquaculture support system characterized by: The aquaculture support system according to claim 4,
The harvest amount proposing unit provides a first predicted value of the weight related to the first number of days of farming, which is the number of days of farming to be predicted, and the weight related to the second number of days of farming after the first number of days of farming. The ratio of the second predicted value exceeds a predetermined value, and the second number of farming days that specifies the first predicted value of the first number of farming days is specified. calculating a proposed yield by multiplying the biomass by 1 minus the value divided by the second predicted value, and outputting the calculated yield;
An aquaculture support system characterized by:

Images