Harvest timing prediction system, harvest timing prediction method, and harvest timing prediction program

Abstract

We provide a harvest timing prediction system that enables highly accurate prediction of harvest time. [Solution] The algae culture apparatus 1 comprises a measurement unit 102 for measuring the culture state of microalgae, an analysis unit 108 that uses the measurement data acquired by the measurement unit 102 to perform machine learning in the learning stage using a predetermined prediction model, and a prediction unit 110 that uses the measurement data in the prediction stage to predict the harvest time of the microalgae using the prediction model on which machine learning has been performed. The measurement data includes the photon flux density of sunlight supplied to the algae culture apparatus 1. In machine learning, the analysis unit 108 divides the microalgae culture period into multiple batches in a time series, sets a culture period of multiple days for each batch, uses five-day data acquired during the culture period as an explanatory variable, and the planned harvest date as the dependent variable.

Description

[Technical Field]

[0001] The disclosures herein relate to a harvest timing prediction system, a harvest timing prediction method, and a harvest timing prediction program. [Background technology]

[0002] Patent Document 1 (Claim 1, etc.), cited below, discloses that "the cultivation state of algae is determined based on information indicating the properties of the culture water that provides a cultivation environment for algae having the ability to adsorb and recover microplastics." Furthermore, paragraph 0078 of Patent Document 1 states that "the machine learning unit 122 generates a predictive model by machine learning. By using the generated predictive model, cell concentration, cell size, proliferation rate, cell lifespan, and adhesive substance secretion concentration can be predicted from newly captured image data from the camera 282, and the cultivation state, growth state, and health state of the algae can be determined."

[0003] Furthermore, Patent Document 2 states that "the harvest prediction means is a means for predicting at least one of the harvest yield and harvest timing of the algae based on the state of the algae determined by the algae cultivation state determination system" (paragraph 0093, etc.) and that "it is preferable that the harvest prediction means is a model obtained by machine learning for each type of algae, which includes state information of the algae at the start of cultivation and after the start of cultivation, and at least one of the harvest yield and harvest timing of the algae in the state information" (paragraph 0096, etc.).

[0004] Furthermore, Patent Document 3 (paragraph 0011, Figure 1, etc.) discloses the present applicant's invention of a closed-system photobioreactor (algae cultivation device). [Prior art documents] [Patent Documents]

[0005] [Patent Document 1] Japanese Patent Publication No. 2023-64772 [Patent Document 2] Japanese Patent Publication No. 2021-132642 [Patent Document 3] Patent No. 7216239 [Overview of the project] [Problems that the invention aims to solve]

[0006] Incidentally, microorganisms such as artificially cultured microalgae (hereinafter sometimes referred to as "microalgae, etc.") can serve as a source of energy, food, and pharmaceuticals in a sustainable society. Furthermore, microalgae, etc. are resources that can make a significant contribution to environmental issues such as carbon dioxide capture and water quality improvement. Moreover, if the harvest time can be predicted with high accuracy during the cultivation of microalgae, etc., it is thought that the use of microalgae, etc. will be promoted in various fields, and the utilization rate of microalgae, etc., and algae cultivation devices disclosed in Patent Document 3 will improve.

[0007] This disclosure aims to provide a harvest timing prediction system, a harvest timing prediction method, and a harvest timing prediction program that are capable of highly accurate harvest timing prediction. [Means for solving the problem]

[0008] (1) The disclosed harvest timing prediction system is A measuring unit for measuring the culture state of microorganisms in a microbial culture device, An analysis unit that uses the measurement data acquired by the measurement unit to perform machine learning in the learning phase of a predetermined predictive model, A harvest time prediction system comprising: a prediction unit that uses the measurement data in the prediction stage to the prediction model on which machine learning has been performed to predict the harvest time of the microorganism, The measurement data includes the photon flux density of the culture light supplied to the microbial culture apparatus. The aforementioned analysis unit is In the aforementioned machine learning process, the culture period of the microorganism is divided into multiple batches in a time series, and a culture period of multiple days is set for each batch. Use the measurement data obtained during the cultivation period as an explanatory variable and the planned harvest date as a target variable. (2) The disclosed harvest time prediction method is the harvest time prediction method performed by the harvest time prediction system of (1) above. (3) The disclosed harvest time prediction program is a harvest time prediction program that causes the harvest time prediction system of (1) above to perform the harvest time prediction method of (2) above. [Advantages of the Invention]

[0009] According to the harvest time prediction system, the harvest time prediction method, and the harvest time prediction program according to the present disclosure, it is possible to perform highly accurate harvest time prediction. [Brief Description of the Drawings]

[0010] [Figure 1] [[ID=**20**]]A diagram showing an example of an algae cultivation apparatus that can be combined with a harvest time prediction system. [Figure 2] [[ID=**23**]]A diagram schematically showing the configuration of the harvest time prediction system according to the embodiment. [Figure 3] [[ID=**26**]](a) is a time chart showing the cultivation period in the learning stage, and (b) is a time chart showing the cultivation period in the prediction stage. [Figure 4] [[ID=**29**]]A graph showing the relationship between turbidity and SS (suspended solids) in the culture solution. [Figure 5] [[ID=**32**]]A chart showing an example of measurement data. [Modes for Carrying Out the Invention]

[0011] [[ID=**40**]]Hereinafter, various embodiments of the present invention will be described while referring to the drawings. However, note that the technical scope of the present invention is not limited to those embodiments, and extends to the invention described in the claims and its equivalents.

[0012] **Note**: The text in bold is the translated content. The text tags to are preserved as they are. If there are any specific requirements or corrections regarding these tags, please let me know.First, an example of an algal culture device 1 (FIG. 1) as a microbial culture device will be described below. The algal culture device 1 shown in FIG. 1 has many of the same configurations as the algal culture device disclosed in Patent Document 3. After the description of the algal culture device 1, a harvest time prediction system 100 (FIG. 3) that can be used in combination with the algal culture device 1 will be described. The harvest time prediction system 100 predicts the harvest time of the microorganisms (here, microalgae) cultured in the algal culture device 1.

[0013] <Closed-system photobioreactor> <<Basic configuration>> The algal culture device 1 in the example of FIG. 1 is a tubular closed-system photobioreactor (PBR). The algal culture device 1 is installed outdoors where it can receive sufficient irradiation of culture light (e.g., sunlight, LED light, etc.), or indoors in a building 72 through which the culture light can penetrate.

[0014] The algal culture device 1 includes a circulation tank 2, a reactor 3, a pump 4, a gas dissolution tube 5, etc. Although details will be described later, the circulation tank 2 houses a culture solution containing microalgae (hereinafter sometimes simply referred to as "algae") inside and is used for storing and supplying the culture solution. The culture solution is a liquid for culturing algae, and for example, is water containing nutrients such as nitrogen and phosphorus, and carbon dioxide ( CO2). The culture solution flows between the circulation tank 2 and the reactor 3. In FIG. 1, the flow of the culture solution is indicated by an arrow M (only some arrows are labeled).

[0015] The reactor 3 cultures algae, and the pump 4 pumps the culture solution (circulation liquid) discharged from the bottom of the circulation tank 2 to the reactor 3. The pump 4 and the flow path inlet of the reactor 3 are connected via a first connection pipe 6. The flow path outlet of the reactor 3 and the circulation tank 2 are connected via a second connection pipe 7. The gas dissolution tube 5 is located closer to the circulation tank 2 between the circulation tank 2 and the reactor 3 and is connected to the second connection pipe 7. Further, in the gas dissolution tube 5, carbon dioxide (CO2) is supplied to the culture solution.

[0016] The first connecting pipe 6 is equipped with a flow meter 61 and a drain valve 62. The flow meter 61 measures the flow rate of the culture medium (circulating fluid) from the pump 4. The drain valve 62 is used for draining water when cleaning the inside of the first connecting pipe 6. In addition, a measuring tank 8 is provided in the middle of the second connecting pipe 7. The measuring tank 8 is used to measure the pH (hydrogen ion concentration), dissolved carbon dioxide concentration, dissolved oxygen concentration, and liquid temperature of the culture medium. The measuring tank 8 is located in the middle section of the second connecting pipe 7. The details of these components are described below.

[0017] <<Circulation Tank 2>> The circulation tank 2 includes a tank body 21, a tank lid 22, an exhaust pipe 23, a jacket 24, and a frame 25, etc. Of these, the tank body 21 has a tank-like space, and drain valves 26 and 27 are provided on the bottom surface of the tank body 21. The tank body 21 is supported by the frame 25.

[0018] The tank lid 22 is formed, for example, in the shape of a plate, and closes the top of the tank body 21 to prevent foreign matter such as debris from entering. A culture solution containing algae is injected into the circulation tank 2.

[0019] The circulation tank 2 is equipped with a liquid level meter (not shown). The liquid level meter detects the liquid level of the culture medium in the circulation tank 2. Based on the detection result of the liquid level meter, the liquid level of the culture medium is controlled to stay within a predetermined range.

[0020] The circulation tank 2 is equipped with a liquid thermometer (T1) 8a. The liquid thermometer (T1) 8a measures the temperature of the culture medium. The liquid thermometer 8a is, for example, a thermocouple-type industrial thermometer using a temperature sensor composed of two different metal conductors. The measurement result is sent to the control system 9, which will be described later, and used as an instruction parameter for the inverter (INV) 9c related to the flow rate control of the pump 4. The inverter (INV) 9c will be described later.

[0021] Next, the exhaust pipe 23 is U-shaped and protrudes from the tank lid 22. This releases impurities that could adversely affect the culture medium in the circulation tank 2 into the atmosphere. For example, if the gas supplied from the gas dissolution pipe 5 to the second connecting pipe 7 is exhaust gas generated when fossil fuels are burned, then gases other than carbon dioxide, such as nitrogen and oxygen, will be released from the circulation tank 2 into the atmosphere via the exhaust pipe 23. A net 23a is provided at the outer end of the exhaust pipe 23. This net 23a prevents foreign matter from entering the circulation tank 2 from the outside.

[0022] Next, the jacket 24 is formed by cylindrical piping that covers the circulation tank 2. The jacket 24 has a heating hot water inlet 24a and a heating hot water outlet 24b. The heating hot water inlet 24a is located at a lower position than the heating hot water outlet 24b. Hot water flows through the jacket 24. The jacket 24 maintains the temperature of the culture medium in the circulation tank 2 at a predetermined level. It is used to maintain a certain temperature.

[0023] For example, if the temperature of the culture medium in the circulation tank 2 drops due to the cold in winter, algae growth will be inhibited. Similarly, if the temperature of the culture medium in the circulation tank 2 rises due to the heat in summer, algae growth will also be inhibited. However, the warm water in the jacket 24 maintains the temperature of the culture medium at a predetermined temperature (for example, about 23°C).

[0024] Of the drain valves 26 and 27 shown at the bottom of the circulation tank 2, one drain valve 26 is opened when the culture medium in the circulation tank 2 is discharged. Drain valve 26 is used for draining liquid when cleaning the circulation tank 2, etc. Drain valve 26 is also used when the liquid level of the culture medium detected by a liquid level meter (not shown) exceeds a predetermined value and the culture medium is recovered. Drain valve 26 may be opened and closed manually or by automatic control.

[0025] Next, of the drain valves 26 and 27 shown at the bottom of the circulation tank 2, the other drain valve 27 is used when sampling the culture medium inside the circulation tank 2. Furthermore, drain valve 27 is also used for checking the drainage status inside the circulation tank 2.

[0026] A shut-off valve 28 is located in the first connecting pipe 6 and is provided between the circulation tank 2 and the pump 4. It is used to stop the flow of culture solution, such as when performing maintenance work on the algae cultivation apparatus 1, by stopping the pump 4 for an extended period. The shut-off valve 28 may be operated manually or automatically.

[0027] <<Reactor 3>> Reactor 3 is used to enable photosynthesis in algae contained in the culture medium. Reactor 3 is formed in a meandering shape using reactor tubes (glass tubes) 32. Reactor tubes 32 are formed in a cylindrical tubular shape using a material (described later) that can transmit light (culture light).

[0028] In reactor 3, a multi-stage spiral-shaped pipeline (flow channel) is formed by connecting the ends of multiple straight pipes vertically with U-shaped pipes. The joints of the pipeline in reactor 3 are smoothly formed so as not to create gaps or steps. Therefore, for example, a cleaning pig (not shown) for the pipeline can move smoothly through the pipe without stopping. In addition, reactor 3 is supported by a reactor frame 31 which has a vibration isolation structure.

[0029] As described above, the reactor 3 is formed in a multi-stage spiral shape. Therefore, even in the case of relatively large-scale outdoor closed-system photobioreactors (for example, with pipeline (flow channel) lengths of 30m or 50m), it is possible to increase the surface area of ​​the reactor 3 while keeping the installation area of ​​the reactor 3 low.

[0030] As a result, more culture light with wavelengths suitable for photosynthesis (400-700 nm) can be irradiated onto the algae in reactor 3. Although not shown in the diagram, an artificial light source (not shown) using LEDs or the like may be provided near reactor 3.

[0031] In this embodiment, the material used for the reactor tube 32 is, for example, a silicate (SiO2) glass tube, which has a high softening temperature, a low coefficient of thermal expansion, and is chemically stable. However, the material is not limited to this, and it is also possible to use a light-transmitting synthetic resin, such as soft plastic (LDPE), as the material for the reactor tube 32.

[0032] The lowest piping in reactor 3 is provided with a flow channel inlet 3a to which the first connecting pipe 6 is connected. It is provided. In addition, the uppermost piping in reactor 3 is provided with a flow outlet 3b that is connected to the second connecting pipe 7.

[0033] <<Pump 4>> Pump 4 is capable of pumping culture medium and washing pigs (not shown) to reactor 3. In this embodiment, pump 4 is a positive displacement pump with a relatively low rotational speed, rather than a rotary pump with a high rotational speed. This is because if the culture medium is supplied by the strong flow of a rotary pump, depending on the type of algae (algae species), the algae may be torn (divided). A positive displacement pump can deliver only a certain amount of liquid to a fixed volume, generating a gentle flow suitable for algae growth.

[0034] <<Gas dissolution tube 5>> The gas dissolution pipe 5 is a U-shaped pipe and has a first pipe 5a and a second pipe 5b. The first pipe 5a is located on the side of the reactor 3, and the second pipe 5b is located on the side of the circulation tank 2. The first pipe 5a and the second pipe 5b each extend vertically. At the top, the first pipe 5a and the second pipe 5b are connected to the second connecting pipe 7, and at the bottom, they are bent in a U-shape and connected to each other.

[0035] A carbon dioxide gas supply unit 5c is provided near the U-shape of the first pipe 5a. The carbon dioxide gas supply unit 5c supplies carbon dioxide gas to the culture medium in the first pipe 5a from the supply port 5d, dissolving carbon dioxide (CO2) in the culture medium. The supplied carbon dioxide forms bubbles in the culture medium. To facilitate the dissolution of carbon dioxide, the diameter of the bubbles is preferably, for example, 5 mm or less. Furthermore, the amount of gas in the cross-sectional area of ​​the flow path of the first pipe 5a is preferably, for example, 2.0 NL (normal liters) / cm³. 2 Less than or equal to / min, more preferably 1.0 NL (normal liters) / cm³ 2 It is less than / min.

[0036] The gas supplied from the gas dissolution pipe 5 may be exhaust gas generated when fossil fuels are burned. For example, carbon dioxide emitted when a hydrogen production device produces hydrogen from city gas can be directly introduced into the algae cultivation device 1. Since such exhaust gas contains carbon dioxide, it is possible to reduce the amount of carbon dioxide released into the atmosphere and contribute to environmental conservation activities. Furthermore, in order to efficiently remove oxygen molecules produced by photosynthesis from the culture medium, it is desirable to keep the oxygen concentration in the exhaust gas below 10%.

[0037] Since the supply port 5d of the carbon dioxide gas supply unit 5c is located on the lower side of the first pipe 5a, fine bubbles of carbon dioxide gas are generated at the bottom of the first pipe 5a. When these generated bubbles rise in the first pipe 5a, the culture medium in the first pipe 5a receives an upward thrust. Therefore, the culture medium in the first pipe 5a rises and flows into the second connecting pipe 7.

[0038] Meanwhile, culture medium is drawn into the second pipe 5b from the second connecting pipe 7. Furthermore, because the culture medium in the second pipe 5b flows downward, a portion of the culture medium flowing through the second connecting pipe 7 is drawn into the second pipe 5b. As a result, a portion of the culture medium that flowed from the first pipe 5a into the second connecting pipe 7 is drawn back into the gas dissolution pipe 5.

[0039] The distance from the supply port 5d of the gas dissolution pipe 5 to the liquid surface of the second connecting pipe 7 is preferably 1.5 m or more, more preferably 1.5 m or more and less than 6 m, and even more preferably 1.5 m or more and less than 2.5 m.

[0040] By doing this, for example, when the bubble size is 5 mm or less, the dissolution rate of carbon dioxide becomes 90% or more, enabling efficient dissolution. In other words, carbon dioxide dissolves Because sufficient distance is maintained, the amount of carbon dioxide dissolved in the culture medium increases dramatically compared to conventional methods.

[0041] Furthermore, the first pipe 5a and the second pipe 5b of the gas dissolution pipe 5 are not inclined with respect to the vertical. Therefore, air bubbles are less likely to adhere to the inner surface of the pipes. Specifically, if air bubbles adhere to the inner surface of the pipes, they come into contact with each other, forming larger bubbles, which reduces the carbon dioxide dissolution efficiency. However, in this embodiment, such a situation can be avoided.

[0042] <<Measurement tank 8>> The measuring tank 8 is positioned on the second connecting pipe 7, between the gas dissolution pipe 5 and the reactor 3. The measuring tank 8 is used to measure the pH, dissolved carbon dioxide concentration, dissolved oxygen concentration, and liquid temperature of the culture medium flowing out of the reactor 3.

[0043] The measurement tank 8 is a storage tank into which the culture medium flows. The measurement tank 8 is filled to a certain liquid level with the culture medium flowing out of the reactor 3. The measurement tank 8 is equipped with a dissolved oxygen meter 8c for measuring the dissolved oxygen concentration and a pH meter 8b for measuring the pH, etc.

[0044] Measurement results from the dissolved oxygen meter 8c, pH meter 8b, etc., are transmitted via wireless communication (not shown) to an external monitoring device (e.g., a personal computer (PC) or tablet terminal). The monitoring device graphs the measurement results at predetermined time scales and monitors them. The measurement results from the dissolved oxygen meter 8c, pH meter 8b, etc., are used as control parameters for a pump (pump 4 in this case) that automatically controls the circulation flow rate of the culture medium. In this way, by utilizing the measurement results from the dissolved oxygen meter 8c, pH meter 8b, etc., it becomes possible to optimize the photosynthetic efficiency (production efficiency) of algae.

[0045] Furthermore, the location where sensors such as the dissolved oxygen meter 8c and pH meter 8b are installed is not limited to the measurement tank 8. For example, it is possible to use embedded sensors and embed them in the second connecting pipe 7 for monitoring.

[0046] Although not shown in the diagram, the measuring tank 8 may also be equipped with a dissolved oxygen concentration meter, a liquid thermometer (T2), etc. These measurement results are also transmitted via wireless communication to an external instrument for the work manager (PC (personal computer) or tablet terminal, etc.) for monitoring.

[0047] <Harvest time prediction> <<Background to the development of the Harvest Time Prediction System 100>> Next, we will explain how to predict the harvest time for algae using the algae cultivation apparatus 1 described above. First, when cultivating microorganisms (such as microalgae) using a photobioreactor like the algae cultivation apparatus 1, the growth speed and harvest time of the algae differ depending on various factors such as room temperature, dissolved oxygen concentration, and weather. Therefore, it is thought that the harvest time can be predicted with high accuracy by using machine learning to incorporate these various factors. Furthermore, by predicting the harvest time with high accuracy, it is thought that the use of the algae cultivation apparatus 1 will be promoted and the utilization rate will improve. As a result, effects such as further reduction of the burden on the field and stabilization of the quality of the cultivated algae can be expected.

[0048] Based on this idea, the inventors define a batch as a certain period of time (a fixed period, a predetermined period) from the start of cultivation, and divide the cultivation period into multiple batches in a time-series manner (details will be described later). Furthermore, the inventors perform machine learning using various factors from past batches and utilize measurement data within each batch to predict the harvest time with high accuracy. (Details will be provided later.)

[0049] <<Basic Configuration of Harvest Time Prediction System 100>> Harvest time prediction is possible by combining the algae cultivation apparatus 1 with a harvest time prediction system 100, for example, as schematically shown in Figure 2. The harvest time prediction system 100 comprises a measurement unit 102, a data collection unit 104, an external information acquisition unit 106, an analysis unit 108, and a prediction unit 110. The harvest time prediction system 100 uses AI (artificial intelligence) with these components to predict the harvest time.

[0050] The detailed functions of each component will be described later, but the data acquisition unit 104, the external information acquisition unit 106, the analysis unit 108, and the prediction unit 110 can be configured using one or more information processing devices (for example, a PC or tablet terminal). Furthermore, a portion of the measurement unit 102 may also be configured using an information processing device. As for the information processing device, although not shown in the diagram, a general configuration including a communication unit, storage unit, display unit, input unit, processing unit, etc., can be used.

[0051] Of these components, the communication unit consists of hardware, communication software such as a TCP / IP (Transmission Control Protocol / Internet Protocol) driver, or a combination thereof. The communication unit may also include a network controller for interfacing with the communication network. The network controller may also comply with wireless communication standards such as Bluetooth®, Near Field Communication (NFC), or infrared.

[0052] Next, the memory unit is a semiconductor memory device such as ROM (Read Only Memory) or RAM (Random Access Memory). The memory unit stores operating system programs, driver programs, control programs, and data used in processing in the processing unit. Driver programs stored in the memory unit include output device driver programs that control the display unit and input device driver programs that control the input unit. Control programs stored in the memory unit are application programs for building prediction models (described later) and for using prediction models to predict harvest times.

[0053] Next, the display unit is equipped with a general display device, such as an LCD or OLED display. The display unit shows various data related to harvest time prediction, as well as information such as prediction results.

[0054] Next, the input section consists of input keys, etc. The user can use the input section to input characters, numbers, symbols, or positions on the display screen of the display section. The input section may also be a pointing device such as a touch panel. When the input section is operated by the user, it generates a signal corresponding to that operation. The input section then supplies the generated signal to the processing unit as an instruction from the user.

[0055] Next, the processing unit is a processor that loads the operating system program, driver program, and control program stored in the memory unit into memory and executes the instructions contained in the loaded program. The processing unit is, for example, the CPU (Central Processing Unit). This is an electronic circuit such as a Unit. The processing unit may be a collection of multiple physically separate processors. For example, the processing unit may have multiple processors that operate cooperatively and in parallel to execute instructions.

[0056] The processing unit executes instructions included in the control program, and at least one of the following units shown in Figure 2: the data acquisition unit 104, the external information acquisition unit 106, the analysis unit 108, and the prediction unit 110. Both function as at least one of the analysis unit 108 and the prediction unit 110. The processing unit may also perform some of the functions of the measurement unit 102. Furthermore, the control system 9 of the algae cultivation apparatus 1 (Figure 1) may be used as the processing unit (or an information processing device equipped with a processing unit, etc.). In this case, the control system 9 of the algae cultivation apparatus 1 (Figure 1) will function as at least one of the analysis unit 108 and the prediction unit 110 in the harvest time prediction system 100.

[0057] <<Measurement section 102>> The measurement unit 102 of the harvest time prediction system 100 shown in Figure 2 acquires data on the algae cultivation environment using, for example, multiple types of sensors. The measurement unit 102 consists of multiple types of sensors that measure the algae cultivation state. Examples of sensors that make up the measurement unit 102 include a pH value sensor 112, a dissolved oxygen sensor 114, a culture medium temperature sensor 116, a CO2 (carbon dioxide) flow rate sensor 118, a circulating fluid flow rate sensor 120, a photon flux density sensor 122, an air flow rate sensor 124, and a turbidity sensor (turbidimeter) 126.

[0058] In Figure 2, the sensors in the measurement unit 102 are shown outside the algae cultivation apparatus 1, but these sensors are actually placed inside the algae cultivation apparatus 1, the building 72, etc. The number and arrangement of each type of sensor can be arbitrarily determined to be suitable for acquiring information related to the cultivation state of the algae.

[0059] The sensors described above are used to measure (detect) various factors related to the cultivation environment of microorganisms (in this case, microalgae) in real time. These various factors become parameters used to predict the harvest time. Various known sensors can be used.

[0060] Furthermore, unless there are any particularly hindering circumstances, the sensors provided in the algae cultivation apparatus 1 may be used for at least some of the sensors used in the harvest time prediction system 100. For example, the pH value sensor 112 can be shared with the pH meter 8b (Figure 1), and the dissolved oxygen sensor 114 can be shared with the dissolved oxygen concentration meter 8c (also Figure 1). In addition, the culture medium temperature sensor 116 can be shared with the liquid thermometer (T1) 8a and liquid thermometer (T2, not shown) shown in Figure 1.

[0061] Furthermore, if the algae cultivation apparatus 1 is equipped with a CO2 flow sensor that detects the flow rate of carbon dioxide (CO2) flowing to the carbon dioxide gas supply unit 5c, this CO2 flow sensor can also be used as the CO2 flow sensor 118. The content and meaning of the factors (parameters) related to the various sensors will be described later.

[0062] <<Data Collection Unit 104>> The data acquisition unit 104 collects factor data (factor information data) measured (detected) by various sensors included in the measurement unit 102 and stores it in data storage (storage device, memory unit). Furthermore, the factor data is saved in a database or the like. As will be described later, the harvest time prediction system 100 acquires data for a predetermined period (cultivation period) as one cycle (one batch) and uses it to predict the harvest time. The harvest time prediction system 100 then periodically saves the data.

[0063] <<External information acquisition unit 106>> The external information acquisition unit 106 acquires external information. External information is information acquired from outside the algae cultivation apparatus 1 (or outside the harvest time prediction system 100), and includes information acquired from external resources, such as the weather information (weather observation information and weather forecast information, etc.) mentioned above. From weather information, information such as sunshine hours and temperature can be obtained. This external information The data will be referred to as "external data" below. The external data acquired by the external information acquisition unit 106 is input to the aforementioned data collection unit 104 and integrated with the factor data stored in the data collection unit 104.

[0064] <<Analysis Department 108>> The analysis unit 108 constructs a predictive model by integrating the collected factor data and external data. The predictive model is constructed using a predetermined analysis method. As a predetermined analysis method, for example, it is possible to use machine learning algorithms such as neural networks or LightGBM (Light Gradient Boosting Machine). LightGBM is a type of gradient boosting algorithm for decision trees.

[0065] In constructing the predictive model, data from the data collection unit 104 (factor data, external data) that covers a predetermined number of days (for example, multiple days such as 5 or 10 days) is used. This predetermined number of days of data is used as explanatory variables, and the dependent variable is the "planned harvest date," which indicates how many days later the algae will grow to a state suitable for harvesting. Using these explanatory and dependent variables, the predictive model for the timing of microbial harvesting is trained.

[0066] <<<Learning Stage>>> Figure 3(a) shows the learning stage (also called the training stage, learning phase, or training phase) in which the predictive model is trained. In Figure 3(a), the time axis is shown from left to right, and this time axis represents the overall cultivation period for algae cultivation. Multiple cultivation periods are set from left to right in the figure, and the cultivation period is divided into multiple batches (Batch 1 to Batch N). Each cultivation period is set to a predetermined number of days, and in the example in Figure 3(a), the number of days for the cultivation period is 5 days. The number of days for the cultivation period is the same for all batches.

[0067] In the first batch shown on the far left of Figure 3(a), factor data is acquired by the aforementioned measurement unit 102. Here, the factor data acquired in one batch is referred to as "five-day data." This "five-day data" may also be written as "5-day data."

[0068] Factor data acquired by the measurement unit 102 and external data acquired by the external information acquisition unit 106 are input into the machine learning algorithm. The machine learning algorithm then performs harvest timing prediction, and the target variable "harvest in a day" is obtained. The target variable "harvest in a day" indicates the planned harvest date. More specifically, the target variable "harvest in a day" indicates that a day after the machine learning was performed, the cultured algae will have grown to a state suitable for harvest. One batch (in this case, the first batch) is formed by this combination of culture timing and the calculation of the target variable.

[0069] After the first batch, the process is repeated with the second, third, and so on. In each batch, machine learning is performed using factor data and external data, and the target variable is calculated. In the example in Figure 3(a), the target variables obtained in each batch from the second batch onward are shown as "harvest after b days", "harvest after c days", ..., "harvest after f days", and "harvest after g days". The predictive model may also use the target variable (and / or explanatory variable) obtained in a previous batch (e.g., the batch before that) as part of its parameters.

[0070] In the example shown in Figure 3(a), batches 1 through 7 are shown, but the illustration of batches after batch 7 is omitted. Also, "a" through "g" in "harvest after a day" through "harvest after g days" represent integers greater than or equal to 1. Furthermore, "a" through "g" may contain the same numerical value.

[0071] Alternatively, the timing of the calculation of the target variable may be set towards the end of the culture period, or the timing of the calculation of the target variable may be included in the culture period. Furthermore, the calculation of the target variable may be performed in subsequent batches.

[0072] In this analysis unit 108, if there are missing values ​​in the data (factor data, external data) for a predetermined number of days (in this case, 5 days), data imputation is performed. For data imputation, data from days within the predetermined number of days that do not have missing data (non-missing days) is used. Furthermore, when imputing data, it is possible to calculate the average value of the data related to the non-missing days and use the obtained average value.

[0073] The prescribed number of days can be increased or decreased (changed) depending on the culture species (the type of algae being cultured). For example, when culturing algae of culture species A, the prescribed number of days can be set to 5 days, and when culturing algae of culture species B, which is different from culture species A, the prescribed number of days can be set to 10 days.

[0074] Next, the analysis unit 108 evaluates the prediction accuracy. For evaluating prediction accuracy, it is possible to use metrics such as RMSE (Root Mean Squared Error). Prediction accuracy can be evaluated for each batch, or it can be evaluated for multiple batches, for example.

[0075] Furthermore, transfer learning can be performed to evaluate and improve the prediction model. By evaluating (re-evaluating) the prediction model constructed through transfer learning, it becomes possible to construct a more accurate prediction model. Specifically, for example, a prediction model (trained model) constructed by a harvest timing prediction system (not shown) of another algae cultivation device installed in the same region can be read into the harvest timing prediction system combined with the algae cultivation device 1 of this embodiment, and the read prediction model can be used in the analysis unit 108. The content of the transfer learning can be used for various types of algae. In addition, the content of the transfer learning can be used even if the climate environment is different.

[0076] Furthermore, the installation of the external information acquisition unit 106 and the use of external data are not mandatory for constructing the predictive model; it is also possible to construct a predictive model without using external data.

[0077] <<<Prediction Stage>>> Figure 3(b) shows the prediction stage (also called the inference stage, prediction phase, or inference phase). In the prediction stage, the prediction unit 110 applies the prediction model (trained model) constructed by the analysis unit 108 to predict the harvest time. Predictions are made for each batch. The batch configuration can be the same as that of the training stage described above.

[0078] Each batch consists of a combination of a process for acquiring "five-day data" (factor data, external data) for a predetermined number of days (in this case, five days) (five-day data acquisition process) and a prediction process (prediction process). In the prediction stage, the "five-day data" is used as an explanatory variable to predict when the crop should be harvested. In the example in Figure 3(b), the prediction results are shown as "Harvest after A days" to "Harvest after G days".

[0079] In the example shown in Figure 3(a), batches 1 through 7 are shown, but the illustration of batches after batch 7 is omitted. Also, "A" through "G" in "Harvest after A days" through "Harvest after G days" represent integers greater than or equal to 1. Furthermore, "A" through "G" may contain the same numerical value.

[0080] <<Parameters measured by the measurement unit 102>> Next, the meaning and content of the parameters measured by the measurement unit 102 will be explained. As mentioned above, in the harvest time prediction system 100 of this embodiment, the pH value sensor 112, dissolved oxygen sensor 114, culture medium temperature sensor 116, CO2 flow rate sensor 118, circulating fluid flow rate sensor 120, photon flux density sensor 122, air flow rate sensor 124, turbidity sensor (turbidimeter) 126, etc. are used in 102 for data measurement.

[0081] These sensors acquire the values of parameters necessary for predicting the harvest time. The parameters for which values are acquired by various sensors are pH value, DO (dissolved oxygen concentration), culture solution temperature, CO2 flow rate, recycle liquid flow rate, photosynthetic photon flux density (PPFD), air flow rate, and SS (suspended solids).

[0082] <<<pH value>>> The pH value sensor 112 measures the pH of the culture solution at the installed location. The following matters can be cited as what the measurement results (measurement data) signify and the uses of the measurement results. (1) The measurement results reflect the dissolved carbon dioxide concentration. When the pH is 5.5 or higher, the lower the pH, the higher the concentration of inorganic carbon (dissolved carbon dioxide, bicarbonate ion, carbonate ion). (2) There is an appropriate pH range (appropriate pH range) depending on the culture species. By comparing the measurement results with the appropriate pH range, it can be confirmed whether the culture environment is within the appropriate pH range.

[0083] <<<DO (dissolved oxygen concentration)>>> The dissolved oxygen sensor 114 measures the dissolved oxygen in the culture solution at the installed location. Dissolved oxygen increases as photosynthesis proceeds. Therefore, by referring to the measurement results (measurement data) of dissolved oxygen, it can be determined whether photosynthesis is being properly carried out. Also, if the concentration of dissolved oxygen is too high, the growth of microorganisms will slow down.

[0084] <<<culture solution temperature>>> The culture solution temperature sensor 116 measures the temperature of the culture solution (culture solution temperature) at the installed location. There is an appropriate (proper) temperature range (appropriate temperature range) depending on the culture species. Generally, within the appropriate temperature range, the higher the temperature, the faster the growth of microorganisms.

[0085] <<<CO2 flow rate>>> The CO2 flow sensor 118 measures the amount of CO2 injected into the culture medium at the location where it is installed. When the pH exceeds a set value, it increases the amount of CO2 injected. A higher CO2 flow rate (more frequent CO2 injection) indicates that photosynthesis is actively taking place.

[0086] <<<Circulating fluid flow rate>>> The circulating fluid flow rate sensor 120 measures the flow rate of the culture medium (circulating fluid) circulating in the algae cultivation apparatus 1 at the location where it is installed. The operation of the pump 4 (Figure 1) that circulates the culture medium is performed in such a way that the detection result of the circulating fluid flow rate sensor 120 does not deviate from the set value as much as possible. Furthermore, if the pump 4 stops due to a malfunction, for example, the algae will not only stop growing, but the likelihood of the algae dying will also increase.

[0087] <<<Photon flux density (PPFD)>>> The photosynthetic photon flux density sensor 122 receives sunlight, which serves as culture light, at the location where it is installed, and detects the photosynthetic photon flux density (PPFD) related to the culture light. The photosynthetic photon flux density sensor 122 is, for example, located in the reactor 3 of the algae culture apparatus 1. It is placed inside the roof 74 of the installed building 72 (Figure 1). The roof 74 and walls (not shown) of building 72 are constructed using, for example, transparent materials that transmit culture light (transparent synthetic resin, transparent glass, etc.). Regarding the placement of the photon flux density sensor 122, it is desirable to place the photon flux density sensor 122 near the reactor 3.

[0088] The unit for "photosynthetic photon flux density" is [μmol / m³]. 2 [μE / sec] represents the number of photons with wavelengths of 400-700 [nm] per unit time and per unit area. In relatively older literature, [μE / m 2The unit [ / sec] is sometimes used. In literature related to photosynthesis, it is sometimes simply expressed as "photon flux density." "Photosynthetic photon flux density" and "photon flux density" are often used as indicators of the amount of light irradiated to plants because they indicate the energy of only the wavelengths used for photosynthesis.

[0089] Generally, a higher photosynthesis density leads to faster photosynthesis and increased growth of microorganisms (in this case, algae). However, if the culture concentration (concentration of microalgae) is low, supplying culture light with an excessively high photosynthesis density can cause photoinhibition, slowing down or even stopping microbial growth. Therefore, it is necessary to maintain an appropriate relationship between the culture concentration and the photosynthesis density.

[0090] Furthermore, multiple photon flux density sensors 122 may be placed near the reactor 3. Regarding the placement of the photon flux density sensors 122, they may be placed, for example, on the outside of the roof 74 of the building 72, as long as the information necessary for predicting the harvest time can be obtained. Also, the culture light detected by the photon flux density sensors 122 is not limited to sunlight, but may include artificial light such as LEDs. Furthermore, the culture light may consist solely of artificial light, provided it is sufficient for culturing microorganisms (in this case, algae).

[0091] Here, all sunlight received by the Earth's surface (sunlight from all directions) is called total solar radiation. The value of total solar radiation includes light of wavelengths not used in photosynthesis (ultraviolet and infrared rays). Total solar radiation data is suitable for predicting the rise in culture medium temperature.

[0092] Furthermore, information (data) such as culture light, sunshine duration, and temperature is included in publicly available meteorological data and weather forecast data. This meteorological data and weather forecast data can be obtained, for example, via public communication lines such as the internet.

[0093] In this embodiment, the photosynthetic photon flux density is measured (detected) near the reactor 3. Therefore, it is possible to obtain global solar radiation data using data from an external organization (e.g., the Japan Meteorological Agency) (data from an external resource) without relying on the detection of the photon flux density sensor 122. Furthermore, some global solar radiation data is made public, while others are not, depending on factors such as region. When performing cultivation and harvest timing predictions in areas where global solar radiation data is made public, information obtained from external resources is useful.

[0094] <<<Airflow>>> The air flow sensor 124 measures the air flow rate inside the algae cultivation apparatus 1 at the location where it is installed. The air inside the algae cultivation apparatus 1 is supplied to the gas dissolution pipe 5, for example, through the supply port 5d. Here, the air is mixed with CO2 (CO2 from the cylinder) before the carbon dioxide gas supply section 5c (preliminary stage). The air supply is maintained so that a constant amount of air is always supplied.

[0095] Airflow rate is a parameter that is managed in a way that it is not actively changed. Airflow rate increases when the culture medium temperature rises and decreases when it falls. Also, when blowing in CO2, In a mixed gas, the partial pressure of air decreases, and consequently, the air flow rate also decreases.

[0096] Furthermore, the airflow rate decreases if the air outlet (bubble outlet) in the air supply source (such as an air stone) is blocked by foreign matter (including microorganisms), or if the pressure inside reactor 3 (pressure of the culture medium) increases.

[0097] Furthermore, although not shown in the diagram, in cases where the hydrogen production device directly injects the CO2 emitted when producing hydrogen from city gas into the reactor 3, the airflow rate during operation of the hydrogen production device includes the flow rate of exhaust gas containing CO2, not just air (or only air). The exhaust gas is mixed with CO2 (CO2 from the cylinder) before the carbon dioxide gas supply unit 5c (preceding stage). The exhaust gas is supplied in such a way that a constant amount of exhaust gas is always supplied. Even when exhaust gas is used, only air is supplied to the gas dissolution pipe 5 during times when no exhaust gas is being generated. The exhaust gas is supplied, for example, by drawing it in from near the outlet of the exhaust gas chimney (not shown) using a compressor. If the CO2 in the exhaust gas is sufficient, CO2 is not supplied from the cylinder.

[0098] << <ss>>> The turbidity sensor (turbidimeter) 126 measures the amount of suspended solids (SS) at the location where it is installed. SS is an indicator of the concentration of particles in water. The unit of SS is [mg / L], and SS is expressed as the mass of particles contained per unit volume of liquid. The particles detected in this embodiment are those with a size of 0.6 [μm] or larger. In the culture medium, it is considered that the majority of particles larger than 0.6 μm in the water column are microalgae. Therefore, SS is used as an indicator of the concentration of microalgae (algae concentration).

[0099] The measurement of suspended solids (SS) involves conversion of the measurement results (measurement data). Specifically, the amount of SS in the culture medium is first measured by manual analysis. Furthermore, the correlation between the turbidity (FTU: Formazin Turbidity Unit) measured by the turbidity sensor (turbidimeter) 126 and the SS obtained by manual analysis (e.g., analysis involving filtration with a filter) is determined, and a formula is created by fitting it with a linear function. The formula can be created, for example, by installing application software equipped with various statistical analysis functions (regression analysis and other analytical functions) on a computer.

[0100] Figure 4 shows an example of the generated formula (Y = 0.7509 × X - 112.04) along with a graph. In the graph in Figure 4, the horizontal axis (X axis) represents turbidity, and the vertical axis (Y axis) represents SS. Turbidity is converted to SS using a formula like the example in Figure 4. The accuracy of the analysis in the example in Figure 4 (R 2 The coefficient of determination was 0.9741.

[0101] <<Example of measurement data>> Figure 5 shows an example of measurement data. At the top of the table in Figure 5, "Number", "Date and Time", "pH", "DO (Dissolved Oxygen Concentration)", "Culture Solution Temperature", "CO2 Flow Rate", "Circulation Fluid Flow Rate", "Photosynthetic Photon Flux Density (PPFD)", "Air Flow Rate", and "SS" are shown. Among these, "Number" is the item number of the data (measurement data) obtained by measurement, and "Date and Time" is the date and time (acquisition date and time) when the measurement data was acquired.

[0102] The measurement data with "Number" being "1" has a "Date and Time" of "2024 / 4 / α 17:29", which indicates that the "Date and Time" of the data with "Number" being "1" was acquired at 17:29 on April α, 2024. "α" is a redacted character. Similarly, the measurement data with "Number" being "2" was acquired at 17:30, 1 minute later, and the measurement data with "Number" being "3" was acquired at 17:31, 1 minute further later.

[0103] Also, in the measurement data with "Number" being "1", "pH" is "6.37", "DO (Dissolved Oxygen Concentration)" is 10.46 [mg / L], "Culture Solution Temperature" is 22.93 [℃], "CO2 Flow Rate" is 0.01 [L / min], "Circulation Fluid Flow Rate" is 3.714 [m 3 / h], "Photosynthetic Photon Flux Density (PPFD)" is 44 [μmol / m 2 / sec], "Air Flow Rate" is 1.753 [L / min], and "SS" is approximately 165.50 [mg / L].

[0104] Furthermore, in the measurement data with "Number" being "2", "pH" is "6.35", "DO (Dissolved Oxygen Concentration)" is 10.04 [mg / L], "Culture Solution Temperature" is 22.92 [℃], "CO2 Flow Rate" is 0.007 [L / min], "Circulation Fluid Flow Rate" is 3.718 [m 3 / h], "Photosynthetic Photon Flux Density (PPFD)" is 40 [μmol / m 2 / sec], "Air Flow Rate" is 1.755 [L / min], and "SS" is approximately 118.79 [mg / L].

[0105] As mentioned above, in this embodiment, the "five-day data" is used as an explanatory variable in the prediction model. Therefore, when measurement data with a "number" of "1" or later is used in the first batch, the measurement data up to the same time on the date obtained by adding 5 to "2024 / 4 / α" is used as the "five-day data".

[0106] Furthermore, the dependent variable "harvest after a day" is calculated using the "five-day data" as the explanatory variable. Then, if the measurement data in the example in Figure 5 is, for example, measurement data for a culture species that can be harvested when SS < 920 [mg / L], the number of days (here, a day) until SS exceeds the threshold (here, 920 [mg / L]) is predicted using the "five-day data" from measurement data starting from "number" "1".

[0107] Similarly, in the second batch, the number of days (in this case, b days) until SS ≥ 920 [mg / L] is predicted, using the subsequent "five-day data" as the explanatory variable. Furthermore, in the third batch, similarly, the number of days (in this case, c days) until SS ≥ 920 [mg / L] is predicted, using the "five-day data" as the explanatory variable.

[0108] In the example in Figure 5, for measurement data starting with "Number" "11460", SS ≥ 920 [mg / L], and the "Date and Time" of this measurement data is 16:28 on May β, 2024. "β" is a redacted character.

[0109] For example, in the learning phase (Figure 3(a)) or the prediction phase (Figure 3(b)), if the "five-day data" with a "number" of "1" or later is used as the "five-day data" for the first batch, and the planned harvest date "a days later" for "harvest a day later" or the planned harvest date "A days later" for "harvest A day later" refers to "May β day, 2024" or a date with a predetermined number of days difference close to the number of days until "May β day, 2024", then the prediction accuracy of the prediction model can be said to be high. The "predetermined number of days difference" here could be, for example, +1 to +2 days.

[0110] For example, in the learning phase (Figure 3(a)), if "a days later," which is the planned harvest date for "harvest a day later," does not satisfy the "predetermined number of days difference" condition mentioned above, it is possible to use that difference to evaluate the prediction accuracy in the analysis unit 108 mentioned above (by substituting it into the RMSE formula as the difference between the predicted value and the actual value).

[0111] <Basic effects achieved by the invention according to this embodiment> According to the harvest time prediction system 100 described above, the measurement data from the measurement unit includes the photon flux density of the culture light supplied to the algae culture apparatus 1. The analysis unit 108, in machine learning, divides the algae culture period into multiple batches (first batch, second batch, ...) in a time series, sets a culture period of multiple days (for example, 5 days) for each batch, and uses the measurement data acquired during the culture period. We are constructing a predictive model using constant data as explanatory variables and the planned harvest date as the dependent variable.

[0112] Therefore, after initially predicting the harvest time and several days have passed, it is possible to make further harvest time predictions using measurement data from those multiple days. As a result, compared to simply predicting the harvest time using machine learning based on measurement data, it is possible to make more accurate harvest time predictions by effectively utilizing a certain type of measurement data (parameters) acquired in a time series.

[0113] Specifically, the invention is effective in terms of data consistency, model stability, improved prediction accuracy, and reduced computational costs. Regarding data consistency, using data from multiple days in each batch enables more consistent predictions that are not affected by short-term fluctuations. Regarding model stability, using long-term data stabilizes model learning and prevents overfitting. Regarding improved prediction accuracy, using more historical data makes it easier to capture trends and seasonality, improving prediction accuracy. Regarding computational costs, if predictions are made each time measurement data is acquired, the model needs to be run frequently, resulting in high computational costs. In contrast, as in this embodiment, running the model with data from multiple days at once reduces computational costs.

[0114] Furthermore, since the analysis unit 108 uses external data obtained from sources other than the microbial culture device for predicting the harvest time, it is possible to predict the harvest time using a wide range of data, rather than relying solely on data obtained by the measurement unit 102 (internal data). And by using external data, it is possible to perform harvest time predictions with even higher accuracy.

[0115] Furthermore, according to the harvest time prediction system 100, the analysis unit 108 evaluates the prediction accuracy in batches, for example, using RMSE (Root Mean Squared Error) as an indicator. Therefore, the prediction accuracy can be objectively understood. In addition, by reflecting the evaluation results, it becomes possible to make the harvest time prediction more accurate.

[0116] Furthermore, the prediction unit 110 divides the algae cultivation period into multiple batches in a time series, sets a cultivation period of multiple days for each batch, and uses the measurement data acquired during the cultivation period in the prediction stage in the prediction model to determine the planned harvest date.

[0117] Therefore, not only in the learning phase, but also in the prediction phase, it is possible to predict the harvest time using measurement data from multiple days after the initial prediction of the harvest time has passed. As a result, compared to simply predicting the harvest time using measurement data at that point in time, it is possible to make more accurate harvest time predictions based on, for example, measurement data of the same type (and quantity).

[0118] Furthermore, the harvest time prediction method used in the harvest time prediction system 100, as well as the harvest time prediction program that operates the harvest time prediction system 100, can also achieve the same effects as the invention of the harvest time prediction system 100.

[0119] Furthermore, according to the harvest time prediction method of this embodiment, the prediction model of the analysis unit is a prediction model that has been machine-learned by another harvest time prediction system, and the other harvest time prediction system comprises another measurement unit that measures the culture state in a microbial culture device combined with the other harvest time prediction system, and another analysis unit that uses the measurement data acquired by the measurement unit to perform machine learning in the learning stage using a predetermined prediction model. Therefore, it is possible to efficiently and accurately predict the harvest time by utilizing the prediction model constructed by the other harvest time prediction system. The content of the transfer learning relates to various types of algae. It can be used in various ways. Furthermore, the content of transfer learning can be used even if the climate environment is different.

[0120] <Inventions that can be extracted from the embodiments> From the embodiments described so far, it is possible to extract the following inventions. (1) A measuring unit (measuring unit 102, etc.) for measuring the culture state of microorganisms (microalgae, etc.) in a microbial culture device (algae culture device 1, etc.), An analysis unit (such as analysis unit 108) performs machine learning in the learning phase using measurement data acquired by the measurement unit in a predetermined predictive model, A harvest time prediction system (harvest time prediction system 100, etc.) comprising: a prediction unit (prediction unit 110, etc.) that uses the measurement data in the prediction stage to the prediction model on which machine learning has been performed to predict the harvest time of the microorganism, The measurement data includes the photon flux density of the culture light (such as sunlight) supplied to the microbial culture apparatus. The aforementioned analysis unit is In the aforementioned machine learning process, the culture period of the microorganism is divided into multiple batches in a time series, and a culture period of multiple days is set for each batch. A harvest time prediction system that uses the measurement data (such as five-day data) obtained during the aforementioned cultivation period as an explanatory variable and the planned harvest date as the dependent variable. (2) The harvest time prediction system according to (1) above, wherein the analysis unit evaluates the prediction accuracy in at least one of the batches. (3) The prediction unit is The culture period of the aforementioned microorganism is divided into multiple batches in chronological order, and a culture period of multiple days is set for each batch. A harvest time prediction system according to (1) or (2) above, which uses the measurement data obtained during the cultivation period of the prediction stage to determine the planned harvest date in the prediction model. (4) A measuring unit (measuring unit 102, etc.) for measuring the culture state of microorganisms (microalgae, etc.) in a microbial culture device (algae culture device 1, etc.), An analysis unit (such as analysis unit 108) performs machine learning in the learning phase using measurement data acquired by the measurement unit in a predetermined predictive model, A harvest time prediction method performed in a harvest time prediction system (such as a harvest time prediction system 100) comprising: a prediction unit that uses the measurement data in the prediction stage to the prediction model on which machine learning has been performed to predict the harvest time of the microorganism; The measurement data includes the photon flux density of the culture light (such as sunlight) supplied to the microbial culture apparatus. The aforementioned analysis unit is In the aforementioned machine learning process, the culture period of the microorganism is divided into multiple batches in a time series, and a culture period of multiple days is set for each batch. A method for predicting the harvest time, wherein the measurement data (such as five-day data) obtained during the aforementioned cultivation period is used as the explanatory variable, and the planned harvest date is used as the dependent variable. (5) The harvest time prediction method according to (4) above, wherein the analysis unit evaluates the prediction accuracy in at least one of the batches. (6) The prediction model of the analysis unit is a prediction model that has been machine-learned using another harvest time prediction system and has been transferred learning. The aforementioned other harvest timing prediction systems are: A separate measuring unit for measuring the culture state in a microbial culture apparatus combined with the other harvest time prediction system, The harvest time prediction method according to (4) or (5) above, further comprising: another analysis unit that uses the measurement data acquired by the measurement unit to perform machine learning in the learning stage using a predetermined prediction model. (7) The prediction unit is The culture period of the aforementioned microorganism is divided into multiple batches in chronological order, and a culture period of multiple days is set for each batch. The harvest timing prediction method according to (4) or (5) above, wherein the measurement data obtained during the cultivation period of the prediction stage is used in the prediction model to determine the planned harvest date. (8) A measuring unit (measuring unit 102, etc.) for measuring the culture state of microorganisms (microalgae, etc.) in a microbial culture device (algae culture device 1, etc.), An analysis unit (such as analysis unit 108) performs machine learning in the learning phase using measurement data acquired by the measurement unit in a predetermined predictive model, A harvest time prediction program used in a harvest time prediction system (such as a harvest time prediction system 100), comprising: a prediction unit that uses the measurement data in the prediction stage to the prediction model on which machine learning has been performed to predict the harvest time of the microorganism; The measurement data includes the photon flux density of the culture light (such as sunlight) supplied to the microbial culture apparatus. In the aforementioned analysis unit, In the aforementioned machine learning process, the culture period of the microorganism is divided into multiple batches in a time series, and a culture period of multiple days is set for each batch. A harvest time prediction program that uses the measurement data (such as five-day data) obtained during the aforementioned cultivation period as explanatory variables and the planned harvest date as the dependent variable. (9) The harvest timing prediction program described in (8) above, which causes the analysis unit to perform a process to evaluate the prediction accuracy in at least one of the batches. (10) The prediction unit, The culture period of the aforementioned microorganism is divided into multiple batches in chronological order, and a culture period of multiple days is set for each batch. A harvest timing prediction program according to (8) or (9) above, which uses the measurement data obtained during the cultivation period of the prediction stage to perform a process to determine the planned harvest date using the prediction model.

[0121] <Other> Although this embodiment and its variations have been described in detail above, the present invention is not limited to any particular embodiment. Furthermore, various changes, substitutions, and modifications can be made to the present invention without departing from its scope.

[0122] For example, the harvest time prediction system 100 may be included in the algae cultivation apparatus 1. Furthermore, the harvest time prediction method performed by the harvest time prediction system 100 may be performed in the algae cultivation apparatus 1. Additionally, the harvest time prediction program may be executed in the control unit (e.g., control system 9) of the algae cultivation apparatus 1.

[0123] Furthermore, the parameters to be measured are not limited to the aforementioned "pH," "DO (dissolved oxygen concentration)," "culture medium temperature," "CO2 flow rate," "circulating fluid flow rate," "photon flux density (PPFD)," "air flow rate," and "SS." It is possible to change to various parameters suitable for predicting harvest time, or to add other parameters.

[0124] Furthermore, regarding prediction models, various types of models can be adopted as long as they can predict the harvest time using measurement data. [Explanation of Symbols]

[0125] 1: Algae culture device 2: Circulation tank 3: Reactor 4: Pump 5: Gas dissolution pipe 5a: First piping 5b: Second piping 5c: Carbon Dioxide Gas Supply Unit 5d: Supply port 72: Building 74: Roof 100: Harvest Time Prediction System 102:Measurement section 104: Data Collection Department 106: External information acquisition department 108:Analysis Department 110: Prediction Department 112: pH value sensor 114: Dissolved oxygen sensor 116: Culture medium temperature sensor 118: CO2 flow sensor 120: Circulating fluid flow sensor 122: Photon quantum flux density sensor 124: Airflow sensor 126: Turbidity sensor< / ss>

Claims

1. A measuring unit for measuring the culture state of microorganisms in a microbial culture device, An analysis unit that uses the measurement data acquired by the measurement unit to perform machine learning in the learning phase of a predetermined predictive model, A harvest time prediction system comprising: a prediction unit that uses the measurement data in the prediction stage to the prediction model on which machine learning has been performed to predict the harvest time of the microorganism, The measurement data includes the photon flux density of the culture light supplied to the microbial culture apparatus. The aforementioned analysis unit is In the aforementioned machine learning process, the culture period of the microorganism is divided into multiple batches in a time series, and a culture period of multiple days is set for each batch. A harvest time prediction system in which the measurement data obtained during the aforementioned cultivation period is used as an explanatory variable and the planned harvest date is used as the dependent variable.

2. The harvest time prediction system according to claim 1, wherein the analysis unit evaluates the prediction accuracy in at least one of the batches.

3. The prediction unit, The culture period of the aforementioned microorganism is divided into multiple batches in chronological order, and a culture period of multiple days is set for each batch. The harvest time prediction system according to claim 1 or 2, wherein the measurement data obtained during the cultivation period of the prediction stage is used in the prediction model to determine the planned harvest date.

4. A measuring unit for measuring the culture state of microorganisms in a microbial culture device, An analysis unit that uses the measurement data acquired by the measurement unit to perform machine learning in the learning phase of a predetermined predictive model, A harvest time prediction method performed in a harvest time prediction system comprising: a prediction unit that uses the measurement data in the prediction stage to the prediction model on which machine learning has been performed to predict the harvest time of the microorganism; The measurement data includes the photon flux density of the culture light supplied to the microbial culture apparatus. The aforementioned analysis unit is In the aforementioned machine learning process, the culture period of the microorganism is divided into multiple batches in a time series, and a culture period of multiple days is set for each batch. A method for predicting harvest time, wherein the measurement data obtained during the aforementioned cultivation period is used as the explanatory variable and the planned harvest date is used as the dependent variable.

5. The harvest time prediction method according to claim 4, wherein the analysis unit evaluates the prediction accuracy in at least one of the batches.

6. The prediction model in the analysis unit is a prediction model that has been machine-learned using another harvest time prediction system, and has been transferred and learned from that model. The aforementioned other harvest timing prediction systems are: A separate measuring unit for measuring the culture state in a microbial culture apparatus combined with the other harvest time prediction system, The harvest time prediction method according to claim 4 or 5, further comprising: another analysis unit that uses the measurement data acquired by the measurement unit to perform machine learning in the learning stage using a predetermined prediction model.

7. The prediction unit, The culture period of the aforementioned microorganisms is divided into multiple batches in chronological order, and each batch is given a culture period of several days. Set it, The harvest time prediction method according to claim 4 or 5, wherein the measurement data obtained during the cultivation period of the prediction stage is used in the prediction model to determine the planned harvest date.

8. A measuring unit for measuring the culture state of microorganisms in a microbial culture device, An analysis unit that uses the measurement data acquired by the measurement unit to perform machine learning in the learning phase of a predetermined predictive model, A harvest time prediction program used in a harvest time prediction system comprising: a prediction unit that uses the measurement data in the prediction stage to predict the harvest time of the microorganism using machine learning; The measurement data includes the photon flux density of the culture light supplied to the microbial culture apparatus. In the aforementioned analysis unit, In the aforementioned machine learning process, the culture period of the microorganism is divided into multiple batches in a time series, and a culture period of multiple days is set for each batch. A harvest time prediction program that uses the measurement data obtained during the aforementioned cultivation period as explanatory variables and the planned harvest date as the dependent variable.

9. The harvest timing prediction program according to claim 8, wherein the analysis unit is instructed to perform a process to evaluate the prediction accuracy in at least one of the batches.

10. The prediction unit, The culture period of the aforementioned microorganism is divided into multiple batches in chronological order, and a culture period of multiple days is set for each batch. The harvest time prediction program according to claim 8 or 9, which uses the measurement data obtained during the cultivation period of the prediction stage to perform a process to determine the planned harvest date using the prediction model.

Images