JP6849958B2 - Prediction method, prediction program, and prediction device - Google Patents

Description

The present disclosure relates to techniques for predicting crop yields.

A technique for predicting the yield of agricultural products using a neural network or chaos analysis is known (see Patent Document 1). In this technique, the first predicted yield is calculated based on the cultivation plan of the crop. In addition, a correction value based on the weather data is calculated using a neural network or the like. Then, the yield of the agricultural product is calculated based on the first predicted yield and the corrected value.

JP-A-2002-136223

However, if the weather fluctuates greatly, there is a risk that the yield of crops cannot be predicted accurately.
One aspect of the present disclosure is preferably to accurately predict the crop yield.

The prediction method according to one aspect of the present disclosure predicts a predetermined crop yield in a predetermined area. In the prediction method, the arrival time when the crop reaches a predetermined growth stage is determined by the target year, which is the year including the arrival time, and any year before the target year, or one of the regions. Predict based on the weather information of. Further, in the prediction method, the harvest situation in the target year is predicted based on the meteorological information of the measurement time determined based on the arrival time in both or one of the target year and an arbitrary year.

The weather at the time when the crop reaches a predetermined growth stage, or before or after that time, may have a relatively large effect on the growth of the crop. In particular, when the weather fluctuates drastically, the weather at such times may differ significantly from the usual year, and as a result, the harvest situation may fluctuate significantly compared to the usual year.

On the other hand, according to the above prediction method, the arrival time when the crop reaches a predetermined growth stage is predicted. Then, the harvest situation of the crop is predicted based on the meteorological information of the measurement time determined based on the predicted arrival time. Therefore, it is possible to predict the harvest situation based on the weather at the time when the crop growth is relatively significantly affected. Therefore, it is possible to accurately predict the harvest status of crops.

In the above configuration, the day when the crop was planted in the target year is set as the start date, and the arrival time is the meteorological information of the day determined based on the start date in both or one of the target year and any year. It may be predicted based on. Further, the measurement time may be a day determined based on the arrival time in both or one of the target year and the arbitrary year.

The weather on a particular day during the growing season of the crop can have a relatively large effect on the growth of the crop. On the other hand, according to the above configuration, the arrival time is predicted based on the meteorological information of the day determined based on the start date. Then, the harvest situation is predicted based on the meteorological information of the measurement time, which is the day determined based on the predicted arrival time. Therefore, it is possible to predict the arrival time and the harvest situation based on the weather on a specific day, which has a relatively large effect on the growth of crops. Therefore, it is possible to accurately predict the harvest situation.

Further, in the above configuration, the arrival time may be the time when the spikes or flowers of the crop reach a predetermined growth stage.
The weather at the time when the spikes or flowers of the crop reach a predetermined growth stage, or before or after that time, may have a relatively large effect on the growth of the crop. On the other hand, according to the above configuration, the time when the crop reaches such a growth stage is predicted as the time of arrival. Then, the harvest situation is predicted by the weather information of the measurement time determined based on the predicted arrival time. Therefore, it is possible to accurately predict the harvest status of crops.

Further, in the above configuration, the crop may be rice.
With such a configuration, it is possible to accurately predict the rice harvest situation.
Further, one aspect of the present disclosure relates to a prediction program that causes a computer to execute the above-mentioned prediction method.

Further, one aspect of the present disclosure relates to a prediction device that executes the above-mentioned prediction method.
According to such a prediction program or a prediction device, it is possible to accurately predict the harvest situation of the crop.

FIG. 1A is a block diagram of the prediction system of the present embodiment. FIG. 1B is a flowchart of the prediction process of the present embodiment. It is explanatory drawing of the neural network used in the prediction processing.

Hereinafter, modes for carrying out the invention will be described with reference to the drawings.
[1. Constitution]
The prediction system 1 shown in FIG. 1A includes a server 10, a terminal 20, and a measuring device 30. These are connected to a communication network such as the Internet.

The server 10 communicates with the terminal 20, the measuring device 30, and the like via the communication network. The server includes a storage unit 11, a control unit, a display unit, an operation unit, and the like. In this embodiment, as an example, a PC is used as the server 10. However, the present invention is not limited to this, and another electronic device may be used as the server 10.

The display unit has a liquid crystal display or the like and displays various information. In addition, the operation unit has a keyboard, a mouse, and the like, and accepts various operations.
The storage unit 11 has a rewritable non-volatile storage device such as a flash memory or an HDD. The storage unit 11 corresponds to a non-transitional substantive recording medium and stores various data. The storage unit 11 stores a prediction program, which is an application installed on the server 10. The prediction program predicts the harvest status of crops (hereinafter, target crops) in a predetermined farmland (hereinafter, target farmland) in a predetermined area (hereinafter, target area). In addition, the weather history and cultivation history are stored in the storage unit 11. The meteorological history has a plurality of meteorological information measured in the target area. In addition, the cultivation history shows information on the cultivation of the target crop on the agricultural land in the target area in the past. As an example, the cultivation history may show information on the cultivation of the target crop on the target agricultural land in the past.

The control unit includes a well-known computer having a CPU and a semiconductor memory such as a ROM, RAM, and a flash memory. The function of the server 10 is realized by operating the CPU according to a program stored in the semiconductor memory. In addition, by executing the program, the method corresponding to the program is executed. The prediction program is loaded in the semiconductor memory. The method of realizing the function of the server 10 is not limited to software. A part or all of the functions may be realized by a logic circuit or the like (in other words, hardware) of the server 10.

Further, the server 10 has a heading date prediction unit 12, a yield prediction unit 13, and a quality prediction unit 14. These correspond to the functions realized by the control unit operating according to the prediction program. The functions of these parts will be described later.

In addition, the measuring device 30 measures the weather in the target area and generates daily weather information in the target area. The measuring device 30 may measure the weather of the target farmland in the target area and generate daily weather information of the target farmland. Specifically, as the meteorological information, for example, types such as average temperature, maximum temperature, minimum temperature, sunshine duration, precipitation amount, and temperature-daily difference in one day may be provided. The daily temperature difference is the difference between the maximum temperature and the minimum temperature on that day. The measuring device 30 transmits weather information to the server 10 at any time. On the other hand, the server 10 stores the weather information received from the measuring device 30 in the storage unit 11 as a part of the weather history.

In addition, the measuring device 30 may transmit the measurement results such as the temperature, precipitation, and sunshine condition to the server 10 in real time. Then, the server 10 may generate the above-mentioned meteorological information based on the measurement result received from the measuring device 30.

Further, the terminal 20 is configured as an electronic device such as a PC or a smartphone as an example. The terminal 20 accesses the server 10 in response to an instruction from the user, and acquires a prediction result such as a harvest status of the target crop. The terminal 20 may further acquire information on the cultivation of the target crop, such as the weather history or cultivation history of the target area, from the server 10. Then, the terminal 20 displays the prediction result such as the harvest status of the target crop or the information regarding the cultivation of the target crop by using a browser or the like.

[2. processing]
Next, the processing by the prediction program will be described. In the present embodiment, as an example, the prediction program predicts the harvest situation in the target year by using a predetermined variety of rice as the target crop. The rice may be paddy rice or upland rice. Further, the harvest status may be, for example, the yield of the target crop and the quality of the harvested product harvested from the target crop. Further, the quality of the harvested product may be information on the grade of the harvested rice, for example, when the target crop is rice. The grade of rice is determined based on the ratio of sized rice contained in the rice and the water content in the rice. The sizing is rice grains other than damaged rice grains (specifically, damaged grains, dead rice, and colored rice). The ratio of sized rice contained in the harvested rice, the water content of the rice, and the like may be used as the quality of the rice as the target crop. In addition to this, for example, the taste of the harvested product, the amount of nutrition contained in the harvested product, and the like may be used as the quality. Further, for example, the harvest time of the target crop may be predicted as the harvest status.

Hereinafter, the year immediately before the target year will be described as the previous year. The meteorological history stored in the storage unit 11 has a plurality of meteorological information for a plurality of years (hereinafter, a recording period) from the past to the target year. Hereinafter, the weather history of the previous year will be described as the previous year's weather history, and the weather history of the target year will be described as the target year's weather history. In addition, the cultivation history stored in the storage unit 11 indicates information on the cultivation of rice, which is the target crop, on the agricultural land in the target area during the recording period. The cultivation history shows the rice planting date, heading date, harvest date, harvest amount, quality, etc. of each year during the recording period. The cultivation history may be generated based on the information input via the operation unit. The weather history and cultivation history are used to predict the harvest situation.

The heading date prediction unit 12 of the prediction program predicts the heading date. The heading date is the day when the ratio of the headed rice to the rice cultivated on the agricultural land reaches a predetermined threshold value. The threshold value may be, for example, a value of 40 to 50%. A plurality of predetermined types of meteorological information on a predetermined rice planting reference date are used for predicting the heading date. The rice planting reference date is the date set based on the rice planting date in the target year. The rice planting date is the day when the rice, which is the target crop, was planted on the target farmland in the target year (in other words, the day when the rice, which is the target crop, was planted). A plurality of meteorological information used for predicting the heading date is extracted from the target year meteorological history or the previous year meteorological history. The heading date may be predicted based on one type of meteorological information on one rice planting reference date in the previous year's meteorological history or the target year's meteorological history.

Then, the yield prediction unit 13 predicts the yield of rice based on the heading date, and the quality prediction unit 14 predicts the quality of rice based on the heading date. A plurality of predetermined types of meteorological information on a predetermined heading reference date are used for these predictions. The heading reference date is a date determined based on the predicted heading date in the target year. Such a plurality of meteorological information is extracted from the target year's meteorological history or the previous year's meteorological history. The yield or quality of rice may be predicted based on one type of meteorological information on one heading reference date in the previous year's meteorological history or the target year's meteorological history.

In the following, a prediction process for predicting the yield and quality of rice, which is a target crop, will be described (FIG. 1B). This process is executed by the server 10 that operates according to the prediction program. In the prediction processing, various areas are targeted, rice of other varieties is targeted as crops, and various crops (for example, crops having spikes such as wheat, or vegetables such as tomatoes or green soybeans). It is possible to predict the harvest situation by targeting various varieties. When crops such as vegetables without spikes are the target crops, the harvest status is predicted using the arrival time instead of the heading date. The arrival time is the time when the target crop reaches a predetermined growth stage other than heading. In addition, the meteorological information or mathematical formulas used for the prediction processing are determined according to the target area or the target crop.

In S100, the server 10 calculates an input variable used for predicting the heading date. The server 10 uses the predetermined one or more types of weather information on the predetermined one or more rice planting reference dates as explanatory variables. Input variables are generated with a number of different explanatory variables. The number of the plurality of explanatory variables is N. In addition, each of the plurality of explanatory variables is also described as E0 [i]. i is an integer of any of 1 to N.

Meteorological information that has a relatively large effect on the heading of rice is an explanatory variable. As an example, the explanatory variables are the precipitation on the 3rd day from the rice planting date, the minimum temperature on the 22nd day from the rice planting date, the average temperature on the 26th day from the rice planting date, and 26 from the rice planting date. It may be the maximum temperature of the day. The rice planting reference date related to the explanatory variable and the type of meteorological information used as the explanatory variable are determined based on the meteorological history and cultivation history during the recording period. The number of explanatory variables, the rice planting reference date related to the explanatory variables, the type of weather information used as the explanatory variables, etc. differ depending on the type of the target crop, the variety of the target crop, the target area, and the like.

The server 10 extracts explanatory variables from both or one of the previous year's meteorological history and the target year's meteorological history stored in the storage unit 11. At this time, the server 10 may extract the explanatory variables that are the meteorological information of the rice planting reference date before the forecasting date from the meteorological history of the target year. On the other hand, the explanatory variables that are the meteorological information of the rice planting reference date after the forecasting date may be extracted from the previous year's meteorological history. That is, the same type of meteorological information as the explanatory variable on the day corresponding to the rice planting reference date in the previous year may be extracted from the previous year's meteorological history. If the rice planting reference date is the X month and Y day of the target year, the day corresponding to the rice planting reference date in another year is the X month and Y day of the other year. That is, in principle, the server 10 may extract explanatory variables from the target year's meteorological history, and if it cannot be extracted, it may extract explanatory variables from the previous year's meteorological history. Of course, the server 10 may extract all the explanatory variables from the previous year's weather history.

Next, the server 10 normalizes the explanatory variable E0 [i] by the equation (1) to obtain a scale (E0 [i]).

Here, the date corresponding to the rice planting reference date pertaining to E0 [i] in each year included in the recording period is defined as the corresponding date. Further, the same type of weather information as E0 [i] on the corresponding date is used as the corresponding explanatory variable. ave (E0 [i]) is the average value of the corresponding explanatory variables for each year included in the recording period, and σ (E0 [i]) is the standard deviation of the corresponding explanatory variables for each year.

Then, the server 10 analyzes the principal components of N scales (E0 [i]). The synthetic variable generated by the principal component analysis is the input variable X0. Specifically, M input variables X0 [1] to [M] are generated by the equation (2). Note that M is an integer and 1 <M <N.

_{The principal component load amounts R0 11 to R0 NM} in the equation (2) are calculated based on the corresponding explanatory variables of each explanatory variable in each year included in the recording period. Further, the number of input variables, the value of the principal component load, and the like differ depending on the type of the target crop, the variety of the target crop, the target area, and the like.

In S105, the server 10 predicts the heading date by a neural network that is a multi-layer perceptron. Note that 200 in FIG. 2A is an example of such a neural network. The input layer of the neural network corresponds to M input variables. The neural network has one intermediate layer, and M neurons are provided in the intermediate layer. Further, the output layer of the neural network outputs one output variable. Further, in the neural network, the sigmoid function shown in the equation (3) is used as the output function g (x).

M input variables X0 [1] to [M] are input to the neural network. Hereinafter, the j-th input variable is set to X0 [j]. In addition, j is an integer of any one of 1 to M.

Further, the output from the k-th neuron among the M neurons in the intermediate layer is defined as Y0 [k]. In addition, k is an integer of any one of 1 to M. Further, the weight parameter corresponding to X0 [j] in the k-th neuron among the M neurons in the intermediate layer is W0 _kj . Further, the bias parameter in the k-th neuron is set to θ0 _k . Y0 [k] is calculated by the formula (4).

Further, the output layer of the neural network outputs Z0 indicating the heading date. Let the weight parameter in the neurons of the output layer be W0 _k and the bias parameter be Θ0. Z0 is calculated by the formula (5).

Z0 is a value obtained by normalizing the predicted value of the heading date of the target year. Therefore, the format of Z0 is changed by the equation (6). As a result, Z0', which is a predicted value of the heading date, can be obtained. The heading date may be represented by, for example, the number of days elapsed from the rice planting date or the like.

Note that σ (z0) indicates the standard deviation of the heading date of each year during the recording period. Further, ave (z0) indicates the average value of the heading date of each year in the recording period. Further, the weight parameters W0 _kj and W0 _k and the bias parameters θ0 _k and Θ0 in the neural network are determined by machine learning using the past heading date indicated by the cultivation history in the recording period as teacher data. The weight parameters W0 _kj , W0 _k , or the bias parameters θ0 _k , Θ0 have different values depending on the type of the target crop, the variety of the target crop, the target area, and the like. Further, the configuration of the neural network may be changed according to the target type, the variety of the target crop, the target area, and the like.

Subsequently, in S110, the server 10 calculates an input variable used for predicting the yield. The server 10 uses a predetermined one or more types of weather information on a predetermined heading reference date as an explanatory variable. The heading reference date is determined based on the heading date predicted in S105. Input variables are generated with a number of different explanatory variables. Further, let O be the number of the plurality of explanatory variables. In addition, each of the plurality of explanatory variables is also described as E1 [i]. i is an integer of any of 1 to O.

Meteorological information, which has a relatively large effect on rice yield, is an explanatory variable. As an example, the explanatory variables may be precipitation on the day 13 days before the heading date, temperature-day difference on the 10th day from the heading date, and sunshine hours on the 10th day from the heading date. good. The heading reference date related to the explanatory variable and the type of meteorological information used as the explanatory variable are determined based on the meteorological history and cultivation history during the recording period. The number of explanatory variables, the heading reference date related to the explanatory variables, the type of weather information used as the explanatory variables, etc. differ depending on the type of the target crop, the variety of the target crop, the target area, and the like.

The server 10 extracts explanatory variables from both or one of the previous year's meteorological history and the target year's meteorological history stored in the storage unit 11. At this time, the server 10 may extract the explanatory variables from the target year's meteorological history in principle in the same manner as in S100, and if it cannot be extracted, the explanatory variables may be extracted from the previous year's meteorological history. Of course, the server 10 may extract all the explanatory variables from the previous year's weather history.

Next, the server 10 normalizes the explanatory variable E1 [i] by the equation (7) to obtain a scale (E1 [i]).

Here, the date corresponding to the heading reference date related to E1 [i] in each year included in the recording period is defined as the corresponding date. Further, the same type of weather information as E1 [i] on the corresponding date is used as the corresponding explanatory variable. ave (E1 [i]) is the average value of the corresponding explanatory variables for each year included in the recording period, and σ (E1 [i]) is the standard deviation of the corresponding explanatory variables for each year.

Then, the server 10 analyzes the principal components of O scales (E1 [i]). The synthetic variable generated by the principal component analysis is the input variable X1. Specifically, P input variables X1 [1] to [P] are generated by the equation (8). Note that P is an integer and 1 <P <O.

_{The principal component load amounts R1 11 to R1 OP} in the equation (8) are calculated based on the corresponding explanatory variables of each explanatory variable in each year included in the recording period. Further, the number of input variables, the value of the principal component load, and the like differ depending on the type of the target crop, the variety of the target crop, the target area, and the like.

In S115, the server 10 predicts the yield by means of a neural network that is a multi-layer perceptron. The neural network is configured in the same manner as the neural network used for predicting the heading date, but the number of input variables corresponding to the input layer and the number of neurons contained in the intermediate layer are different. That is, the input layer of the neural network used for predicting the yield corresponds to P input variables. Further, in the neural network, P neurons are provided in the intermediate layer. Further, in the neural network, the sigmoid function shown in the equation (3) is used as the output function g (x).

P input variables X1 [1] to [P] are input to the neural network. Hereinafter, the j-th input variable will be referred to as X1 [j]. In addition, j is an integer of any one of 1 to P.

Further, the output from the k-th neuron among the P neurons in the intermediate layer is defined as Y1 [k]. In addition, k is an integer of any one of 1 to P. Further, the weight parameter corresponding to X1 [j] in the k-th neuron among the P neurons in the intermediate layer is W1 _kj . Further, the bias parameter in the k-th neuron is set to θ1 _k . Y1 [k] is calculated by the equation (9).

Further, the output layer of the neural network outputs Z1 indicating the yield. Let the weight parameter in the neurons of the output layer be W1 _k and the bias parameter be Θ1. Z1 is calculated by the formula (10).

Z1 is a value obtained by normalizing the predicted value of the yield in the target year. Therefore, the format of Z1 is changed by the equation (11). As a result, Z1', which is a predicted value of the yield, can be obtained.

Note that σ (z1) indicates the standard deviation of the yield for each year during the recording period. In addition, ave (z1) indicates the average value of the yield of each year in the recording period. Further, the weight parameters W1 _kj and W1 _k and the bias parameters θ1 _k and Θ1 in the neural network are determined by machine learning using the past yield indicated by the cultivation history in the recording period as teacher data. The weight parameters W1 _kj and W1 _k and the bias parameters θ1 _k and Θ1 have different values depending on the type of the target crop, the variety of the target crop, the target area, and the like. Further, the configuration of the neural network used for prediction may be changed according to the type of the target crop, the variety of the target crop, the target area, and the like.

Subsequently, in S120, the server 10 calculates an input variable used for quality prediction. Similar to S110, the server 10 uses a predetermined one or a plurality of types of weather information on a predetermined heading reference date as an explanatory variable. Input variables are generated with a number of different explanatory variables. Further, let Q be the number of the plurality of explanatory variables. In addition, each of the plurality of explanatory variables is also described as E2 [i]. i is an integer of any of 1 to Q.

Meteorological information, which has a relatively large effect on rice quality, is an explanatory variable. As an example, the explanatory variables may be the average temperature on the day 10 days before the heading date, the maximum temperature on the 8th day from the heading date, the sunshine hours on the 19th day from the heading date, and the like. .. The heading reference date related to the explanatory variable and the type of meteorological information used as the explanatory variable are determined based on the meteorological history and cultivation history during the recording period. The number of explanatory variables, the heading reference date related to the explanatory variables, the type of weather information used as the explanatory variables, etc. differ depending on the type of the target crop, the variety of the target crop, the target area, and the like.

The server 10 extracts explanatory variables from both or one of the previous year's meteorological history and the target year's meteorological history stored in the storage unit 11. At this time, the server 10 may extract the explanatory variables from the target year's meteorological history in principle in the same manner as in S100, and if it cannot be extracted, the explanatory variables may be extracted from the previous year's meteorological history. Of course, the server 10 may extract all the explanatory variables from the previous year's weather history.

Then, the server 10 analyzes the principal components of Q E2 [i]. The synthetic variable generated by the principal component analysis is the input variable X2. Specifically, R input variables X2 [1] to [R] are generated by the equation (12). R is an integer, and 1 <R <Q.

Here, the date corresponding to the heading reference date related to E2 [i] in each year included in the recording period is defined as the corresponding date. Further, the same type of weather information as E2 [i] on the corresponding date is used as the corresponding explanatory variable. _{The principal component load amounts R2 11 to R2 QR} in the formula (12) are calculated based on the corresponding explanatory variables of each explanatory variable in each year included in the recording period. The number of input variables, the value of the principal component load, etc. differ depending on the type of target crop, the variety of target crop, the target area, and the like.

In S125, the server 10 predicts the quality by multiple regression analysis using X2 [j] as an explanatory variable. In addition, j is an integer of any one of 1 to R. Specifically, Y2, which is a predicted quality value, is calculated by the equation (13).

Then, the server 10 ends this process.
_{The partial regression coefficients a 1 to} a _R in the equation (13) are calculated as follows. That is, the server 10 generates the input variable X2 in the same manner as in S120 for each year included in the recording period. That is, the server 10 extracts the explanatory variable E2 [i] corresponding to the year from the weather history of each year in the same manner as S120. At this time, the heading reference date for E2 [i] of each year is determined based on the heading date of the year. Then, the server 10 generates X2 [j] based on the explanatory variable E2 [i] according to the equation (12) as in S120. Further, the server 10 calculates the _{partial regression coefficients a 0 to} a _R based on X2 [j] of each year and the quality of each year. The partial regression coefficients a _{1 to} a _R have different values depending on the type of the target crop, the variety of the target crop, the target area, and the like.

[3. effect]
According to the above-described embodiment described in detail above, the following effects can be obtained.
(1) The weather at the time when the crop reaches a predetermined growth stage or before or after the stage may have a relatively large influence on the growth of the crop. In particular, when the weather fluctuates drastically, the weather at such times may differ significantly from the usual year, and as a result, the harvest situation may fluctuate significantly compared to the usual year.

On the other hand, according to the prediction process of the present embodiment, the heading date of the target crop, rice, is predicted. Then, the rice harvest situation is predicted based on the weather information such as the heading reference date determined based on the predicted heading date. Therefore, it is possible to predict the harvest situation based on the weather at the time when the rice growth is relatively significantly affected. Therefore, it is possible to accurately predict the rice harvest situation.

(2) In addition, the weather on a specific day during the cultivation period of the crop may have a relatively large effect on the growth of the crop. On the other hand, according to the prediction processing of the present embodiment, the heading date is predicted based on the weather information such as the rice planting reference date, which is the date determined based on the rice planting date. Then, the harvest situation is predicted based on the weather information such as the heading reference date, which is the date determined based on the predicted heading date. Therefore, it is possible to predict the heading date and the harvest situation based on the weather on a specific day, which has a relatively large effect on the growth of the target crop. Therefore, it is possible to accurately predict the rice harvest situation.

(3) In addition, the weather when the spikes or flowers of the crop reach a predetermined growth stage, or before or after that stage may have a relatively large effect on the growth of the crop. .. On the other hand, according to the prediction process of the present embodiment, the rice harvest situation is predicted based on the weather information at the time determined based on the predicted heading date. Therefore, it is possible to accurately predict the harvest situation.

[4. Other embodiments]
Although the embodiment for carrying out the present disclosure has been described above, the present disclosure is not limited to the above embodiment, and can be implemented in various modifications.

(1) In the prediction process of the above embodiment, the heading date is predicted based on the weather information of the target year or the previous year. Then, the harvest situation is predicted based on the weather information of the heading reference date determined based on the predicted heading date or the day corresponding to the heading reference date in the previous year. Hereinafter, the time when the meteorological information used for the prediction is measured will be referred to as the measurement time. However, the prediction program may predict the arrival time when the target crop reaches a growth stage other than heading in the prediction process. Then, the prediction program may determine the measurement time based on the predicted arrival time and predict the harvest situation based on the predetermined type of meteorological information at the measurement time in the same manner as in the present embodiment.

Specifically, the arrival time may be, for example, the time when the spikes or flowers of the target crop reach a predetermined growth stage. Specifically, the target crop may reach a growth stage such as ear differentiation, flower bud differentiation, flowering, or falling flowers. The ear differentiation may be, for example, a growth stage at which ears have begun to be formed on the target crop, and the flower bud differentiation may be, for example, a growth stage at which flower buds have begun to be formed on the target crop. In addition to this, the growth stage related to the arrival time may be, for example, pollination of the target crop, and the size, length, number, etc. of the stem, branch, leaf, etc. of the target crop may be determined in advance. It may be at the stage when the set standard is reached.

The arrival time is the day when the ratio of the target crops cultivated on the target farmland to the target crops that have reached the growth stage reaches a predetermined threshold value (for example, 40 to 50%). Is also good. Further, the arrival time or the measurement time is not limited to any day, and may be, for example, a period spanning a plurality of days, a time zone on any day, or the like. It is preferable to determine the arrival time and the measurement time according to the type of the target crop, the variety of the target crop, the target area, and the like.

(2) In the prediction process of the above embodiment, the heading date is predicted based on the weather information of the rice planting reference date determined based on the rice planting date. However, the forecasting program may predict the heading date in the forecasting process, for example, based on the meteorological information of the specified date regardless of the rice planting date in the target year or the previous year.

Further, for example, when the seeds of the target crop are planted and cultivated on the target agricultural land, the start reference date may be used instead of the rice planting reference date in the prediction processing. The start reference date is a date set based on the start date, which is the date when the seeds of the target crop are planted on the target farmland. That is, the prediction program may use the start reference date instead of the rice planting reference date in the prediction process to predict the arrival time when the target crop reaches a predetermined growth stage in the same manner as in the present embodiment. .. Then, the prediction program may predict the harvest status of the target crop based on the meteorological information of the measurement time determined based on the arrival time.

(3) In the prediction process of the above embodiment, a neural network is used for predicting the heading date and yield of rice as a target crop, and multiple regression analysis is used for predicting quality. However, the heading date and yield may be predicted by multiple regression analysis, and the quality may be predicted by a neural network. In addition, the heading date, yield, and quality may be predicted by other methods.

(4) In the prediction process of the above embodiment, the harvest situation is predicted using the previous year's weather history, which is the weather information of the previous year of the target year. However, in the prediction process, the harvest situation may be predicted using not only the previous year's meteorological history but also the meteorological history of any one or more years before the target year. That is, in the prediction process, the explanatory variables may be extracted in the same manner from the meteorological history of any year before the previous year instead of the previous year's meteorological history. In addition to this, the harvest situation may be predicted using the meteorological history of any plurality of years prior to the target year. That is, the average weather history may be generated based on the weather history of the plurality of years. The average weather history shows the average value of each type of weather information on each day of these years. Then, in the prediction process, the explanatory variables may be extracted from the average weather history in the same manner instead of the previous year's weather history.

(5) A plurality of functions possessed by one component in the above embodiment may be realized by a plurality of components, or one function possessed by one component may be realized by a plurality of components. .. Further, a plurality of functions possessed by the plurality of components may be realized by one component, or one function realized by the plurality of components may be realized by one component. Further, a part of the configuration of the above embodiment may be omitted. In addition, at least a part of the configuration of the above embodiment may be added or replaced with the configuration of the other above embodiment. It should be noted that all aspects included in the technical idea specified only by the wording described in the claims are embodiments of the present disclosure.

(6) In addition to the prediction system 1 described above, there are various devices such as a device corresponding to the prediction system 1, a non-transitional actual recording medium such as a semiconductor memory in which the prediction program in the prediction system 1 is recorded, and a method corresponding to the prediction processing. The present disclosure can also be realized in the form of.

[Correspondence with claims]
The correspondence between the terms used in the description of the above-described embodiment and the terms used in describing the scope of claims is shown.

The method realized by the prediction process corresponds to an example of the prediction method, and the server 10 corresponds to an example of the prediction device.
In addition, the heading of rice, which is the target crop, corresponds to an example of the growth stage, and the heading date corresponds to an example of the arrival time. In addition, the heading reference date and the day corresponding to the heading reference date in the previous year correspond to an example of the measurement time. In addition, the rice planting date corresponds to an example of the start date.

Further, S100 and S105 of the prediction process correspond to an example of the arrival time prediction unit, and S110 to S125 correspond to an example of the harvest status prediction unit.

1 ... Prediction system, 10 ... server, 11 ... storage unit, 12 ... heading date prediction unit, 13 ... yield prediction unit, 14 ... quality prediction unit, 20 ... terminal, 30 ... measuring device.

Claims (8)

It is a predictive method for predicting the harvest situation of a predetermined crop in a predetermined area.
The time of arrival of the crop at a predetermined growth stage is the weather in the area in both or one of the target year, which is the year including the arrival time, and any year prior to the target year. Informed predictions
The harvest situation in the target year is predicted by a neural network which is a multi-layer perceptron, and the first input variable input to the neural network is the arrival in both or one of the target year and the arbitrary year. The meteorological information on the first reference date, which is a specific day determined based on the time , and is generated based on the meteorological information having a relatively large influence on the harvest situation.
A prediction method characterized by. In the prediction method according to claim 1,
The first input variable is generated by performing principal component analysis on the meteorological information on the first reference date.
A prediction method characterized by. In the prediction method according to claim 1 or 2.
The day when the crop was planted in the target year was set as the start date.
The arrival time is predicted by a neural network that is a multi-layer perceptron, and the second input variable input to the neural network is based on the start date in both or one of the target year and the arbitrary year. The meteorological information on the second reference date, which is a specified specific day, and is generated based on the meteorological information having a relatively large influence on the crop reaching the growth stage.
A prediction method characterized by. In the prediction method according to claim 3,
The second input variable is generated by performing principal component analysis on the meteorological information on the second reference date.
A prediction method characterized by. In the prediction method according to any one of claims 1 to 4,
The arrival time is the time when the spikes or flowers of the crop reach a predetermined growth stage.
A prediction method characterized by. In the prediction method according to any one of claims 1 to 5,
The crop is rice
A prediction method characterized by. A forecasting program that predicts the harvest status of predetermined crops in a predetermined area.
The time of arrival of the crop at a predetermined growth stage is the weather in the area in both or one of the target year, which is the year including the arrival time, and any year prior to the target year. The arrival time prediction unit that predicts based on information,
The part that predicts the harvest situation in the target year by a neural network that is a multi-layer perceptron, and the first input variable input to the neural network is both the target year and the arbitrary year, or On the other hand, it is the meteorological information on the first reference date, which is a specific day determined based on the arrival time , and is a part generated based on the meteorological information having a relatively large influence on the harvest situation. As a harvest status prediction department
A prediction program characterized by running a computer. It is a predictor that predicts the harvest situation of a predetermined crop in a predetermined area.
The time of arrival of the crop at a predetermined growth stage is the weather in the area in both or one of the target year, which is the year including the arrival time, and any year prior to the target year. The arrival time prediction unit that predicts based on information,
The part that predicts the harvest situation in the target year by a neural network that is a multi-layer perceptron, and the first input variable input to the neural network is both the target year and the arbitrary year, or On the other hand, it is the meteorological information on the first reference date, which is a specific day determined based on the arrival time , and is a part generated based on the meteorological information having a relatively large influence on the harvest situation. Harvest status prediction department and
Predictor device.

Images