Optimization fuel calculation device using waste oil

Abstract

The present invention is an optimization fuel calculation device 1 for automatically calculating an optimum biofuel which can be produced from a waste oil that is a biomass resource, the optimized fuel calculation device 1 being provided with a storage unit 12 which retains data about the waste oil, a machine learning model 112 into which information about the waste oil which is to be stored in the storage unit 12 is input, and an optimization calculation execution unit 102 which executes an optimization calculation for producing a biofuel from the waste oil using the machine learning model 112. The data about the waste oil which is to be stored in the storage unit 12 includes the type of the waste oil, the amount of the waste oil, and the acid value of the waste oil, and the optimization calculation execution unit 102 identifies the type of a producible optimum biofuel and determines the mixing ratio of the waste oil on the basis of at least one of combustion properties such as the acid value of the waste oil, the cetane value of the produced biofuel and the amount of heat generation. In the present invention, according to this configuration, it becomes possible to perform an automatic calculation for producing an optimum and high-quality biofuel depending on the intended use of a user by using, as a raw material, a waste oil such as waste water containing oils and fats.

Description

Technical Field

[0001] The present invention relates to an optimization fuel calculation device used for producing optimal and high-quality biofuel according to user applications when producing fuel by using waste oil from restaurants, food factories, and the like.

Background Art

[0002] In recent years, the problem of global warming has been intensifying, and the reduction of greenhouse gases such as carbon dioxide and methane gas has been attempted. For example, the Paris Agreement is an agreement that defines an international framework for global warming countermeasures after 2020. All countries, regardless of whether they are developed or developing countries, participate in global warming countermeasures and aim to substantially reduce greenhouse gas emissions to zero in the latter half of the 21st century.

[0003] In addition, Sustainable Development Goals (SDGs) have been proposed. For example, in the SDGs Action Plan 2021 announced by the Japanese government, the achievement of carbon neutrality, which substantially reduces greenhouse gas emissions by 2050, is described (see, for example, Non-Patent Document 1).

[0004] There are various methods for creating greenhouse gas reduction, such as the introduction of renewable energy facilities such as solar power and wind power generation, the introduction of energy-saving products such as heat pumps, and various activities such as forest management. One of them is the utilization of biomass (reuse of waste).

[0005] For example, the waste water fats and oils, which are biomass resources discharged throughout Japan, amount to more than 300,000 tons per year in relation to restaurants and more than 800,000 tons per year in relation to food factories, with a total of more than 1.1 million tons. Considering the whole world, the amount is even more enormous. In recent years, a system has been developed that produces unique biomass fuel from waste water fats and oils that had to be disposed of as industrial waste sludge and performs biomass power generation with a diesel generator (see, for example, Patent Documents 1 to 3). By providing this biomass power generation system, CO2 reduction, recycling, and water purification can be achieved. [Prior art documents] [Patent Documents]

[0006] [Patent Document 1] Japanese Patent Publication No. 2014-217804 [Patent Document 2] Patent No. 5452814 [Patent Document 3] Utility Model Registration No. 3216173 Gazette [Non-patent literature]

[0007] [Non-Patent Document 1] "What are SDGs? JAPAN SDGs Action Platform, Ministry of Foreign Affairs" [Accessed March 28, 2021], Internet<URL:https: / / www.mofa.go.jp / mofaj / gaiko / oda / sdgs / index.html> [Overview of the project] [Problems that the invention aims to solve]

[0008] However, producing optimal and high-quality biofuel from waste oil is no easy task. The reasons for this are explained below.

[0009] The first reason is that the waste oil separated and recovered during the wastewater treatment process at restaurants and food factories includes various types of oils, such as those derived from wastewater treatment (beef tallow, lard, palm oil, etc.) and household waste oil (tempura oil). Furthermore, these waste oils vary in acid value from low to high, have high water content, are solid or liquid at room temperature, and contain many impurities. Therefore, it is not easy to determine how much of these various waste oils should be used and what kind of treatment is necessary to produce optimized biofuel.

[0010] The second problem is that the user applications for biofuels refined from waste oil vary widely, such as using them as fuel for power generation, as a substitute for heavy fuel oil A in boilers and incinerators, or as a clean energy source like hydrogen. Therefore, it is desirable to produce biofuels that are optimally suited to user applications from existing waste oil inventory, but this has not yet been realized.

[0011] This invention has been made in view of the above problems, and aims to provide an optimized fuel calculation device that automatically calculates how to obtain the optimal and high-quality biofuel according to the user's application, using waste oil such as wastewater and oil as raw materials. [Means for solving the problem]

[0012] To achieve the above objective, the present invention provides an optimization fuel calculation device for automatically calculating a biofuel having a desired combustion performance that can be produced from oils and fats, comprising a storage unit for holding data related to oils and fats, Based on the type and quantity of oil and fat, combustion performance is predicted, including (a) ignition, main combustion start time, and cetane number, which evaluate the ease of ignition during combustion of the mixed oil and fat; (b) maximum ROHR (Rate of Heat Release) time, which evaluates engine efficiency; (c) total combustion time, afterburning period 95%, afterburning period 99%, and afterburning period 100%, which evaluate the unburned portion; (d) SD' standard deviation, which evaluates multi-cylinder variation; and (e) ROHR (Rate of Heat Release) index, which evaluates the calorific value. Machine learning models and, An input unit for inputting information regarding the combustion performance of a desired biofuel, and a learning processing execution unit for learning the setting values ​​in the machine learning model based on the training data, The system includes an optimization calculation execution unit that performs optimization calculations for producing biofuels from oils and fats using the aforementioned machine learning model, and the data relating to oils and fats stored in the storage unit includes The aforementioned training data The oils and fats include beef tallow, lard, palm oil, and tempura oil, and the optimization calculation execution unit is The desired biofuel input via the input unit combustion Based on performance, the machine learning model is used The above optimization calculation is performed to determine the optimal mixing ratio of oils and fats, thereby calculating a biofuel having the desired combustion performance, and the type of biofuel is a liquid fuel.

[0017] In this optimized fuel calculation device, The aforementioned learning data includes the type and amount of oil and fat, and combustion performance, which is (a) evaluation of the ease of ignition during combustion of the mixed oil and fat, such as ignition, main combustion start time, and cetane number, (b) evaluation of engine efficiency, such as the maximum ROHR (Rate of Heat Release) time, (c) evaluation of the unburned portion, such as the total combustion period, afterburn period 95%, afterburn period 99%, and afterburn period 100%, (d) evaluation of multi-cylinder variation, such as the SD' standard deviation, and (e) evaluation of the calorific value, such as the ROHR (Rate of Heat Release) index, and the learning processing execution unit Obtained through the learning process Set value The above is stored in the memory unit. ru, It is characterized by the following:

[0019] To solve the above problems, the present invention provides an optimized fuel calculation method for automatically calculating a biofuel having desired combustion performance that can be produced from fats and oils, comprising a storage step of holding data related to fats and oils, and An input step of inputting information regarding the combustion performance of a desired biofuel, and a learning process execution step of learning setting values ​​in a machine learning model based on training data, and the an optimization calculation execution step of performing an optimization calculation for producing a biofuel from fats and oils using a machine learning model, The aforementioned machine learning model predicts combustion performance based on the type and amount of oil and fat, including (a) ignition, main combustion start time, and cetane number, which are evaluations of the ease of ignition during combustion of the mixed oil and fat; (b) maximum ROHR (Rate of Heat Release) time, which is an evaluation of engine efficiency; (c) total combustion time, afterburning period 95%, afterburning period 99%, and afterburning period 100%, which are evaluations of the unburned portion; (d) SD' standard deviation, which is an evaluation of multi-cylinder variation; and (e) ROHR (Rate of Heat Release) index, which is an evaluation of calorific value. The data related to fats and oils stored in the storage step includes The aforementioned training data which includes the types of fats and oils being beef tallow, lard, palm oil, and tempura oil, and in the optimization calculation execution step, The desired biofuel input in the above input step combustion Based on performance, the machine learning model is used perform the optimization calculation, calculate a biofuel having desired combustion performance by determining an optimal mixing ratio of fats and oils, and the type of the biofuel is a liquid fuel.

[0020] To solve the above problems, the present invention provides a program used in an optimized fuel calculation device for automatically calculating a biofuel having desired combustion performance that can be produced from fats and oils, comprising a storage step of holding data related to fats and oils, and An input step of inputting information regarding the combustion performance of a desired biofuel, and a learning process execution step of learning setting values ​​in a machine learning model based on training data, and the an optimization calculation execution step of performing an optimization calculation for producing a biofuel from fats and oils using a machine learning model, The aforementioned machine learning model predicts combustion performance based on the type and amount of oil and fat, including (a) ignition, main combustion start time, and cetane number, which are evaluations of the ease of ignition during combustion of the mixed oil and fat; (b) maximum ROHR (Rate of Heat Release) time, which is an evaluation of engine efficiency; (c) total combustion time, afterburning period 95%, afterburning period 99%, and afterburning period 100%, which are evaluations of the unburned portion; (d) SD' standard deviation, which is an evaluation of multi-cylinder variation; and (e) ROHR (Rate of Heat Release) index, which is an evaluation of calorific value. The data related to fats and oils stored in the storage step includes The aforementioned training data which includes the types of fats and oils being beef tallow, lard, palm oil, and tempura oil, and in the optimization calculation execution step, The desired biofuel input in the above input step combustion Based on performance, the machine learning model is used perform the optimization calculation, calculate a biofuel having desired combustion performance by determining an optimal mixing ratio of fats and oils, and the type of the biofuel is a liquid fuel.

Advantages of the Invention

[0021] The present invention is an optimization fuel calculation device that automatically calculates an optimal biofuel that can be produced from waste oil, which is a biomass resource. The device includes a storage unit that holds data related to waste oil, a machine learning model into which information related to the waste oil stored in the storage unit is input, and an optimization calculation execution unit that executes an optimization calculation for producing biofuel from the waste oil. The data related to the waste oil stored in the storage unit includes the type, amount, and acid value of the waste oil. The optimization calculation execution unit determines the type of optimal biofuel that can be produced and the mixing ratio of the waste oil based on at least one of the combustion characteristics such as the acid value of the waste oil, the cetane number, and the calorific value of the produced biofuel. With this configuration, the optimization fuel calculation device according to the present invention can perform automatic calculation to obtain an optimal and high-quality biofuel according to the user's application using waste oil such as waste water grease as a raw material.

Brief Description of the Drawings

[0022] [Figure 1] It is an overall process diagram for the use of biofuel including an optimization fuel calculation device according to an embodiment of the present invention. [Figure 2] It is a diagram showing an example of an oil sludge separation device that can be used in the same process. [Figure 3] It is an overall configuration diagram of the same optimization fuel calculation device. [Figure 4] It is a functional block diagram of the same optimization fuel calculation device. [Figure 5] It is a functional block diagram related to the optimization calculation function of the same optimization fuel calculation device. [Figure 6] It is a diagram showing an example of the overall flow when using the same optimization fuel calculation device. [Figure 7] It is a diagram showing an example of the flow when converting an oil-containing by-product discharged in biofuel production using the same optimization fuel calculation device into clean energy. [Figure 8] It is an explanatory diagram of the fuel performance calculated by the same optimization fuel calculation device. [Figure 9] It is a diagram showing training data (library data) according to an experimental example of the present invention. [Figure 10] This is a diagram illustrating the machine learning model using a multivariate normal distribution in the same experimental example. [Figure 11] This is an explanatory diagram of the deep learning model using an autoencoder in the same experimental example. [Figure 12] This figure shows the prediction results when using a machine learning model in the same experimental example. [Figure 13] This figure shows the prediction results when using a deep learning model in the same experimental example. [Figure 14] This is an overall diagram of a system equipped with an optimization fuel calculation device according to a modified example of the above embodiment. [Modes for carrying out the invention]

[0023] (Embodiment) An optimized fuel calculation device utilizing waste oil according to an embodiment of the present invention will be described with reference to the drawings. This optimized fuel calculation device is mainly used to produce optimal and high-quality biofuels using waste oil such as recovered wastewater, oil, and oil sludge.

[0024] <Overall Process Diagram> The overall process for biofuel utilization, including the optimized fuel calculation device according to this embodiment, will be described with reference to Figure 1. This process includes a biomass recovery process, a biofuel manufacturing process, and a biofuel utilization process.

[0025] First, in the biomass recovery process, waste oils such as wastewater and oil discharged from restaurants and food factories, and used cooking oil collected by local governments from the general public, are recovered. These waste oils are still unused or have a low rate of effective utilization as local resources, and are therefore biomass. Biomass refers to "renewable, organic resources of biological origin, excluding fossil resources." Waste biomass includes discarded paper, livestock manure, food waste, and sewage sludge. The CO2 released by burning biomass is the same CO2 that organisms absorb from the atmosphere through photosynthesis during their growth process. By replacing fossil-derived energy and products with biomass, we can make a significant contribution to reducing CO2 emissions, one of the greenhouse gases that cause global warming.

[0026] Generally, wastewater from restaurants, fast food establishments, hotels, and food processing plants contains various water pollutants. If such wastewater is discharged without any treatment, oil and other substances in the wastewater can adhere to and solidify in drainpipes, causing blockages. Furthermore, it can make water purification difficult in combined treatment tanks and sewage treatment plants, and have adverse effects on the environment.

[0027] Therefore, businesses that discharge wastewater containing solids such as oil, sediment, and suspended matter are equipped with treatment equipment (grease traps, oil-water separators, and raw water tanks) on a business-by-business basis to temporarily pool the wastewater, allow the solids to settle or the oil to float and separate, and remove them physically by periodically drawing them out and treating them.

[0028] In this biomass recovery process, for example, a field agent (a specialist company that the applicant has provided its unique know-how to guide, train, and certify) will perform cleaning and management of grease traps in restaurants and recover oily wastewater, which is a biomass resource, from the grease traps. A grease trap is a storage tank that contains oily wastewater, which, if discharged as is, would contaminate the drainage system (sewer pipes) and pollute rivers and the sea. For the technology to recover biomass from grease traps, for example, the devices shown in Patent No. 4401007, Patent No. 4420750, and Utility Model No. 3216173, all patented technologies of the applicant, can be used.

[0029] Other field agents manage oil-water separators in food processing plants and recover biomass from these separators. Specifically, they separate the oil sludge obtained from the oil-water separators in food processing plants into oil, water, and sludge using an oil sludge separation device, and recover the oil (i.e., biomass). The field agents also manage the oil sludge separation devices.

[0030] The oil-sludge separation device uses, for example, the technology owned by the applicant as shown in Figure 2. Specifically, oil-sludge is recovered from the oil-water separation tank 21 of the food factory via the oil-sludge heating and transfer device 22 and transferred into the oil-sludge separation device 23. In the heated oil-sludge separation device 23, the water and sludge, which have a higher specific gravity, move to the lower layer, and the oil, which has a lower specific gravity, moves to the upper layer, so the oil in the upper layer is recovered as biomass. This oil is collected in a recovery tank 24, for example, installed on a truck. For heating the oil-sludge heating and transfer device 22 and the oil-sludge separation device 23, a mobile heat source that is kept warm using a latent heat storage material (PCM: Phase Change Material) container, for example, on a truck, is used. The biomass recovered at the food factory is then transported to the biofuel manufacturing process.

[0031] In the biofuel manufacturing process, biofuel is produced from the biomass brought in, based on the calculation results of the optimized fuel calculation device 1. Normally, the waste oil recovered is low-quality biomass containing a lot of water and foreign matter and solidifying at room temperature. However, in the biofuel manufacturing process, it is refined and modified into fuel for power generation (e.g., SMO (Straight Mixed Oil), boiler fuel (heavy oil substitute fuel), clean energy fuel (e.g., hydrogen, biogas), biomass additives (e.g., manufactured using oily by-products discharged during the refining process) as raw materials, and then biofuels are manufactured and sold.

[0032] For the production of biofuels, a biofuel manufacturing machine equipped with a mixing and blending function may be used. Alternatively, for example, a private power generator (diesel generator) can be used, allowing the biofuel production to be powered by green electricity generated from the biofuel produced. Regarding the production of biofuels from biomass, methods such as those described in the applicant's patents No. 5269552 and No. 5452814 can be used.

[0033] In the next biofuel utilization process, SMO, hydrogen, and carbon-free fuels produced in the biofuel manufacturing process will be used as fuel. SMO will be used, for example, as a fuel source for on-site diesel power generation equipment. Specifically, this will target businesses (such as food factories) that have introduced dedicated SMO tanks or dedicated SMO power generation equipment. SMO refers to oil that is produced by reforming wastewater oil containing a large amount of beef tallow and lard without chemical synthesis. On the other hand, carbon-free fuels can be used in existing boilers and incinerators in factories, etc., and will be utilized as an alternative fuel to mineral oil such as heavy fuel oil A.

[0034] Furthermore, applications of biofuels for energy use include (a) diesel power generation, (b) fuel for boilers and incinerators, (c) methane power generation using oily by-products such as residual sludge and high-acid value oils discharged during the biofuel refining process (methane power generation using pig manure and food waste is also possible), and maintaining highly efficient energy in clean centers through biomass additives, and (d) hydrogen supply.

[0035] <Operation Procedure of Optimized Fuel Calculation Device 1> Next, the operating procedure of the optimized fuel calculation device 1 used in the biofuel manufacturing process will be explained with reference to Figure 3. First, wastewater oil, oil sludge, and waste cooking oil are collected from various locations such as industrial waste treatment companies (floating oil in facilities), waste oil recovery companies (high acid value oils, grease trap oil sludge), wastewater oils (restaurants, food factories), and household waste oils (local governments, citizens) (S301).

[0036] When waste oil is collected, the amount, type, and acid value of the collected waste oil are input and stored in the optimized fuel calculation device 1 (S302). This input may be made not only by direct input but also via the internet by field agents and other remote industrial waste treatment companies and waste oil collection companies.

[0037] Here, we will explain the acid value of oils and fats. Acid value (AV) is one of the numerical indicators of the formation and deterioration of oils and fats. Acid value is defined as the number of milligrams of potassium hydroxide, etc., required to neutralize the free fatty acids contained in 1 gram of oil, measured by titration. Generally, the older the oil, the higher the acid value. Peroxide value (POV) may also be used as an indicator of oxidation of oils and fats. Acid value can be measured using dedicated measuring equipment or by using a simple test kit. Alternatively, the acid value of various types of waste oil can be predetermined for each type of waste oil, rather than measuring it each time.

[0038] Next, the optimized fuel calculation device 1 performs calculations (inferences) for the production of an optimized biofuel from the stock of waste oil (S303). Specifically, (1) it automatically determines the biofuel that can be produced at the present time, the amount of biofuel to be produced, its acid value, and quality characteristics based on the acid value of the stock of waste oil. Or (2) it can automatically determine the production process using the stock of waste oil, the amount to be produced, and its quality characteristics (combustion characteristics, etc.) depending on the intended use for carbon neutrality. When (2) the user inputs the type of biofuel they desire into the optimized fuel calculation device 1, they may select from (1) bio-liquid fuel (fuel for power generation, heavy oil substitute fuel, or co-firing (combustion aid)), (2) methane gas (power generation and mixed use with city gas), (3) hydrogen (response to a hydrogen society (hydrogen is produced from methane or SMO)), etc.

[0039] Then, based on the manufacturing method calculated by the optimized fuel calculation device 1, the desired biofuel is actually manufactured at each fuel manufacturing facility (S304). This biofuel includes, for example, fuel for power generation, carbon-free fuel, biogas, hydrogen, high value-added fuel, and bioplastics.

[0040] Next, the processing units provided in the optimized fuel calculation device 1 according to this embodiment will be described with reference to Figure 4. The optimized fuel calculation device 1 is a server computer and includes, for example, a control unit 10, an optimization calculation unit 11, a storage unit 12, a communication unit 13, a display unit 14, an operation unit 15, and a reading unit 16, as shown in Figure 4. These may be cloud-based.

[0041] The control unit 10 uses a processor such as a CPU and memory to control the components of the device and realize various functions. The optimization calculation unit 11 uses a processor and memory to perform optimization calculations for fuel production in response to control instructions from the control unit 10. The storage unit 12 stores the optimization calculation program 1P and the machine learning library 1L that enables the machine learning model to function. The storage unit 12 also stores definition data that defines the machine learning model, parameters including setting values ​​in the trained machine learning model, etc.

[0042] The communication unit 13 is a communication module that enables communication connection to a communication network such as the Internet. The display unit 14 uses an LCD panel or the like. The operation unit 15 includes a user interface such as a keyboard or mouse. The operation unit 15 notifies the control unit 10 of user operation information. The reading unit 16 can read optimization calculation programs and machine learning libraries stored on a recording medium 2, such as an optical disc, using, for example, a disk drive.

[0043] Next, the optimization calculation function of the optimized fuel calculation device 1 will be explained with reference to Figure 5. The control unit 10 of the optimized fuel calculation device 1 includes a learning processing execution unit 101 and an optimization calculation execution unit 102 that utilize AI (Artificial Intelligence).

[0044] The learning processing execution unit 101 functions as a machine learning model (machine learning engine) based on the machine learning library 1L, definition data, and parameter information stored in the memory unit 12. That is, the learning processing execution unit 101 uses the machine learning model to be learned to perform a process of learning the settings (parameters, etc.) of the machine learning model to be learned based on the learning data. During this learning, for example, the results of acid value adjustment of various blending pattern samples based on the acid values ​​of waste oil, combustion characteristics such as cetane number and calorific value of the main component oil type (lard, beef tallow, palm oil, tempura oil, etc.), component characteristics such as sulfur content, and calcium hydroxide based on the acid value of the waste oil are used. Mu The input amounts of potassium hydroxide or sodium hydroxide, the oil separation ratio, and the results of applying various biofuel production technologies are added as library data. For example, the learning processing execution unit 101 inputs the learning data into the entire machine learning model 112 and performs a process to minimize the error between the output data (data on the biofuel produced) and the known learning data, thereby updating the parameters (weights). The parameters obtained through this learning process are stored in the storage unit 12.

[0045] The optimization calculation execution unit 102 performs optimization calculations based on the optimization calculation program 1P stored in the memory unit 12. That is, the optimization calculation execution unit 102 uses a machine learning model to perform predetermined optimization calculations on the input inventory oils and fats. The optimization calculation execution unit 102 also performs the function of inputting the input data, such as inventory oils and fats and the biofuels to be manufactured, into the input unit 111 based on the user's operation using the operation unit 15.

[0046] The input unit 111 of the optimization calculation unit 11 receives data on the inventory of oils and fats and the desired biofuel from the storage unit 12 and inputs it into the machine learning model.

[0047] When using a pre-trained model, the machine learning model 112 performs optimization calculations based on already trained parameters, such as the selection of waste oil to use (e.g., stock waste oil to be used for acid value adjustment) and the manufacturing method (e.g., application of various biofuel manufacturing technologies). Specifically, the machine learning model 112 extracts feature quantities such as the acid value of stock oils and fats, the cetane number and other combustion characteristics of the biofuel to be manufactured, and other component characteristics such as sulfur content (at least one feature quantity of combustion characteristics such as the acid value of waste oil, the cetane number and calorific value of the biofuel to be manufactured), and performs optimization calculations based on the extracted feature quantities, such as the selection of waste oil, the mixing ratio, and the selection of the manufacturing method. The calculation results include the type of biofuel, the amount to be manufactured, the manufacturing process, and the acid value adjustment. In addition, (1) calcium hydroxide according to the amount of oil and acid value of the biofuel to be manufactured Mu (1) Automatic calculation of the effective input amount of potassium hydroxide or sodium hydroxide, etc., (2) Reflection of the acid value that does not cause metal corrosion in the above, (3) Automatic calculation of the combustion characteristics of the biofuel produced (fuel evaluation), and (4) Applicability determination for each biofuel production technology may also be included.

[0048] Furthermore, the optimization calculation execution unit 102 performs the following predictive calculations during methane production: (1) automatic mixing calculation to maximize methane production, (2) automatic calculation of methane production amount and component characteristics, and (3) automatic calculation of hydrogen production amount (by hydrogen production technology).

[0049] The optimization data processed by the machine learning model 112 is input to the output unit 113. In this way, the optimization fuel calculation device 1 prepares a machine learning model that selects the optimal stock oils and manufacturing methods to be used to produce the desired biofuel. As a result, the optimization fuel calculation device 1 according to this embodiment 1 can learn using the AI-based machine learning model 112, and can achieve efficient processing of stock oils in the optimization calculation by machine learning. With this configuration, it is possible to accurately convert all waste oils, including wastewater oils, into fuel using the type of waste oil and acid value of the raw material as basic information, thereby contributing to carbon neutrality.

[0050] Next, the overall flow when using the optimized fuel calculation device 1 will be explained with reference to Figure 6. First, oils and fats recovered from industrial waste treatment companies 601 (floating oil in facilities), waste oil recovery companies 602 (high acid value oils and fats, grease trap oil sludge), wastewater oils and fats / oil sludge 603 (restaurants, food factories), and household waste oil 604 (local governments, citizens) are sorted into high acid value oils and fats (605) or low acid value oils and fats (606) based on their acid value.

[0051] For example, when SMO is produced as a biofuel (607), the optimized fuel calculation device 1 selects not to perform acid value adjustment (608) when using low-acid value oils (609), and automatically calculates the amount of SMO produced, its acid value, combustion performance, etc. This automatic calculation can be performed after training using the machine learning model described above. SMO can be used as fuel (power generation, heavy oil substitute) or in mixed combustion (combustion aid).

[0052] For example, when SMO is produced as a biofuel (611), the optimized fuel calculation device 1 selects to adjust the acid value (608) when it is a high-acid value oil (612), and when various oils are mixed (613), the acid value and mixing ratio after mixing are automatically calculated. At this time, the applicability of various biofuel production technologies is determined (e.g., BDF). (Registered trademark) Alkaline catalyst method, cathode ion contact method, etc. are also selected. Alternatively, depending on the type of oil in stock, mixing may not be performed (614). Next, calcium hydroxide for adjusting the acid value of the oil MuThe system automatically determines whether or not potassium hydroxide or sodium hydroxide needs to be added, and automatically calculates the required amount and processing time (615). At this time, from the viewpoint of preventing corrosion inside the piping, it may be possible to select an acid value that does not cause metal corrosion (616). As a result, the optimized desired SMO production volume, acid value, combustion performance, etc. are automatically calculated. SMO can be used as fuel (power generation, heavy oil substitute) or in mixed combustion (combustion aid).

[0053] For example, when hydrogen is produced as a biofuel (610), the optimized fuel calculation device 1 selects a high-acid value oil (605) for producing methane gas from stocked oils, predicts the amount of methane gas produced and its component characteristics, determines the applicability of various hydrogen production technologies, and automatically calculates the amount of hydrogen produced as a result. Hydrogen is used in vehicles and other applications as a clean energy source that does not produce carbon dioxide.

[0054] Furthermore, for example, when biomass additives are manufactured as biofuels (618), the optimized fuel calculation device 1 automatically calculates the amount of oily by-products such as refined sludge (617) discharged, the total calorific value, etc., and determines the mixing ratio of the oily by-products. Also, when hydrogen is produced from oily by-products such as refined sludge (617) (621), the device predicts the amount of methane gas produced and its component characteristics (619), determines the applicability of various hydrogen production technologies (620), and automatically calculates the amount of hydrogen produced as a result. In this way, the optimized fuel calculation device 1 can optimize the production of the desired biofuel from the waste oil on hand, which is the raw material in stock, and oily by-products such as refined sludge discharged in biofuel production. It is also possible to predict and calculate the amount of hydrogen that can be produced from SMO (607, 611) (622a, 622b), the amount of high value-added fuel produced (623a, 623b), and the amount of bioplastics produced (624a, 624b).

[0055] Next, the overall flow of the optimized fuel calculation device 1 when utilizing methane will be explained with reference to Figure 7. For example, when hydrogen is produced (704) using oily by-products such as refined sludge (701), pig manure (702), and food waste (703), the optimized fuel calculation device 1 predicts the amount of methane gas produced and its component characteristics, determines the applicability of various hydrogen production technologies, and automatically calculates the amount of hydrogen produced as a result. Hydrogen can be produced by reforming biomass (methanol or methane gas) with a catalyst.

[0056] Next, the combustion performance of the biofuel automatically calculated by the optimized fuel calculation device 1 will be explained with reference to Figure 8. The optimized fuel calculation device 1 can estimate the combustion performance of the biofuel to be produced from the mixing ratio of the original waste oils by acquiring data on the combustion performance of various waste oils in advance. For example, the combustion performance here is an FIA-100 analysis item and is used as an indicator to evaluate the combustion performance of bio-liquid fuel. The optimization calculation device can evaluate the following aspects of biofuels produced based on the composition information of the manufactured fuel: (1) Ignitionability (ignition, main combustion start, cetane number): Determining the quality of ignition; smaller values ​​for ignition and main combustion start indicate easier ignition; (2) Calorific value, etc.: Determining the calorific value of the fuel; larger values ​​indicate higher calorific value; (3) Evaluation of unburned portion: Evaluating the generation of graphite, carbon deposition risk, and knox risk; (4) Multi-cylinder variation evaluation: Determining the variation risk in each cylinder of the mixed fuel; an appropriate value of 0-2 (less than 1 is good); (5) Engine efficiency evaluation: Indicators such as the maximum time the piston is pushed down after ignition and an appropriate value for good engine efficiency can be used.

[0057] <Example of experiment> Next, an experimental example of machine learning in the optimized fuel calculation device 1 according to this embodiment will be described with reference to Figures 9 to 13. The purpose of this experiment is to develop an AI model that optimizes the production of biofuels using wastewater oil and the like as the main raw materials.

[0058] The AI ​​model created in this experiment predicts combustion data based on the mixing ratio of oils and fats. In developing the AI ​​model, data on the types of oils and fats (beef tallow, lard, palm oil, tempura oil) and measured values ​​during combustion (evaluation of ignition properties, engine efficiency, unburned portion, multi-cylinder variation, calorific value, etc.) were used to create the model. Note that the acid value of the oils and fats used in this model is a fixed value, unlike that of recovered waste oil, and therefore was not used in the machine learning data. However, it is certainly conceivable to perform machine learning processing including the acid value.

[0059] In this experiment, we selected "Mathematica" (manufactured by Wolfram Inc.), which allows for symbolic computation, possesses the latest neural networks and a wide range of integrated machine learning functions as an AI model development program, and does not require massive amounts of training data or complex programming. Other options such as "Python" or "MATLAB" (manufactured by Mathworks Inc.) may also be used.

[0060] Next, the specific AI model developed in this experiment will be explained with reference to the diagram. This AI model was trained using a total of 14 data points: the mixing ratios of four types of oils {beef tallow, lard, palm oil, tempura oil} (4 data points) and measured values ​​during the combustion of the mixed oils {ignition, start of main combustion, FIA cetane number, maximum RoHR time, total combustion period, afterburn period 95%, afterburn period 99%, afterburn period 100%, SD' standard deviation, RoHR index} (10 data points). A comparison was then made between the measured values ​​and the predicted values. For details of the training data, the measured data shown in Table 90 in Figure 9 was used.

[0061] Furthermore, we created two types of AI models: a machine learning model trained on a high-dimensional data set as a multivariate normal distribution (see Figure 10), and a deep learning model using an autoencoder (a type of deep learning algorithm) (see Figure 11). We then compared these models with actual measured values.

[0062] Next, the calculation results in this experimental example will be explained with reference to the diagram. In the machine learning model trained as a multivariate normal distribution, all the measured values ​​in Table 90 shown in Figure 9 were treated as multivariate quantities to create the learning model. In practice, a single group of multivariate quantities was formed in the order of {beef tallow, lard, palm oil, tempura oil, ignition, start of main combustion, FIA cetane number, ROHR maximum time, total combustion period, afterburning period 95%, afterburning period 99%, afterburning period 100%, SD' standard deviation, ROHR index}, and a machine learning model was created using this group as a multivariate normal distribution. Specifically, the mixing ratio of {beef tallow, lard, palm oil, tempura oil} was determined, and the values ​​of the remaining items of the multivariate group {beef tallow, ..., ROHR index} mentioned above, namely {ignition, main combustion start, FIA cetane number, ROHR maximum time, total combustion period, afterburn period 95%, afterburn period 99%, afterburn period 100%, SD' standard deviation, ROHR index}, were predicted using a machine learning model (see Figure 10). The prediction results obtained by the machine learning model are shown in Table 120 in Figure 12.

[0063] On the other hand, in the deep learning model using an autoencoder, a single feature was created using the autoencoder from the measured values ​​in Table 90: {ignition, start of main combustion, FIA cetane number, maximum RoHR time, total combustion period, afterburn period 95%, afterburn period 99%, afterburn period 100%, SD' standard deviation, RoHR index}. A learning model was created by linking the created feature with the corresponding mixing ratios of oils {beef tallow, lard, palm oil, tempura oil}. Then, the feature was predicted from the learning model based on the determined oil mixing ratios {beef tallow, lard, palm oil, tempura oil}, and the {ignition, start of main combustion, FIA cetane number, maximum RoHR time, total combustion period, afterburn period 95%, afterburn period 99%, afterburn period 100%, SD' standard deviation, RoHR index} was predicted using a decoder (decoding) (see Figure 11). The predicted values ​​obtained by the deep learning model are shown in Table 130 of Figure 13.

[0064] The results of this experiment show that the prediction values ​​of the deep learning model using an autoencoder tended to be lower than the measured values, making it essential to improve prediction accuracy by enriching the library data in the future. On the other hand, the AI ​​model using a multivariate normal distribution obtained values ​​close to the measured values. Referring to the mean error in Figure 12, it was found that the error between the predicted values ​​and the measured values ​​during combustion was relatively low. Furthermore, for samples without actual measured values, such as "beef tallow: 50ml, lard: 50ml, palm oil: 50ml, tempura oil: 50ml," the predicted values ​​were calculated as follows: "Ignition: 10.118, Main combustion start: 10.1525, FIA cetane number: 26.972, ROHR maximum time: 13.6966, Total combustion period: 18.6158, Afterburning period 95%: 2.21944, Afterburning period 99%: 7.94487, Afterburning period 100%: 23.3996, SD' standard deviation: 0.403133, ROHR index: 168.877."

[0065] In this experimental example, the set of 25 actual data points used as training data (library data) was small, and more training data (library data) is needed to improve accuracy. In all machine learning models, increasing the number of actual data points and using them as training data (library data) for the learning model is important to improve prediction accuracy. Furthermore, it goes without saying that improvements considering operation are necessary, such as creating a program to reflect additional actual data as training data (library data) in the learning model, and adopting a user-friendly interface for easy learning model creation.

[0066] As described above, the present invention is an optimized fuel calculation device 1 that automatically calculates the optimal biofuel that can be produced from waste oil, which is a biomass resource, and comprises a storage unit 12 that holds data on waste oil, a machine learning model 112 into which the waste oil information (such as oil and fat related data) stored in the storage unit 12 is input, and an optimized calculation execution unit 102 that performs an optimized calculation for producing biofuel from waste oil using the machine learning model 112. The waste oil data stored in the storage unit 12 includes the type, amount and acid value of the waste oil, and the optimized calculation execution unit 102 determines the optimal type of biofuel that can be produced and the mixing ratio of the waste oil based on at least one of the combustion characteristics of the acid value of the waste oil, the cetane number and calorific value of the biofuel to be produced. Based on the said determination, the optimized calculation execution unit 102 outputs at least one result of the type, amount, acid value, combustion characteristics and component characteristics of the biofuel to be produced. With this configuration, the optimized fuel calculation device 1 can perform automatic calculations to obtain the optimal and high-quality biofuel according to the user's application using waste oil such as wastewater and oil as raw materials. In other words, by using the acid value of the waste oil used as raw material as basic information, we can convert all waste oils, including wastewater and oils, into optimal biofuels, thereby contributing to the realization of carbon neutrality.

[0067] More specifically, by using the optimized fuel calculation device 1, (1) it becomes possible to convert all waste oils, from low-acid value oils to ultra-high-acid value oils, into fuel, enabling carbon-neutral use toward achieving decarbonization. (2) The acid value adjustment of oils can be calculated instantly and automatically. (3) The manufacturing process can be automatically determined for each use of the biofuel. (4) The optimality of various fuel manufacturing technologies can be automatically determined according to the purpose of use of the biofuel. (5) The amount produced, component characteristics, and combustion characteristics can be automatically calculated for each type of biofuel produced. (6) By-products generated in the biofuel manufacturing process (oil-containing by-products such as refined sludge) can be utilized to contribute to carbon neutrality. For example, if the amount of wastewater oil nationwide is 1.1 million tons, the fuel yield is 60%, and the biofuel price is 97 yen / liter (actual market price of heavy oil A), the market size could exceed 60 billion yen (660,000 tons of biofuel).

[0068] (modified version) Next, the overall system including the optimized fuel calculation device 1 according to a modified example of this embodiment will be described with reference to Figure 14. This system comprises terminal devices 91 owned by field agent A, which manages wastewater from stores; terminal devices 92 owned by field agent B, which manages wastewater from food factories; terminal devices 93 owned by fuel manufacturing plants; terminal devices 94 owned by on-site power generation equipment users; terminal devices 95 owned by boiler and other equipment users; and the optimized fuel calculation device 1, all of which are connected via a wide-area network such as the Internet.

[0069] Terminal device 91 is a terminal such as a personal computer or smartphone installed in the store, and establishes a communication session with the optimized fuel calculation device 1 based on a predetermined protocol. Terminal device 91 sends and receives data (acid value, quantity, type, etc.) regarding wastewater oil and grease recovered in the grease trap. Terminal device 92 is a terminal installed in the food factory, and runs a dedicated application and displays the results on the screen. Terminal device 92 sends and receives data (acid value, quantity, type, etc.) regarding wastewater oil and grease data from the oil-water separator tank to and from the optimized fuel calculation device 1.

[0070] Terminal device 93 is a terminal in the fuel manufacturing plant that runs a dedicated application and displays the results on a screen. Terminal device 93 can send and receive data regarding the desired biofuel with the optimized fuel calculation device 1 based on a predetermined protocol. Terminal device 94 is a terminal installed in the on-site power generation unit that runs a dedicated application and displays the results on a screen. Terminal device 94 sends and receives data regarding the desired biofuel with the optimized fuel calculation device 1 based on a predetermined protocol.

[0071] The terminal device 95 is a terminal installed in a boiler or other equipment user's facility. It installs and runs a dedicated application and displays the results on its screen. The terminal device 95 can send and receive data related to the desired biofuel with the optimized fuel calculation device 1.

[0072] The optimized fuel calculation device 1 is connected to terminal devices 91, 92, 93, 94, and 95 via a network, and transmits and receives data, as well as performs calculations to optimize the biofuels being produced. Data such as access means, metadata, format, units of measurement, and names are standardized and defined between each terminal device 91-95 and the optimized fuel calculation device 1.

[0073] Furthermore, the present invention can be implemented not only as such an optimized fuel calculation device, but also as an optimized fuel calculation method using the characteristic means of such an optimized fuel calculation device as steps, or as a program that causes a computer to execute those steps. Needless to say, such a program can be distributed via recording media such as USB or transmission media such as the Internet. [Explanation of symbols]

[0074] 1. Optimization Fuel Calculation System 10 Control Unit 11 Optimization Calculation Unit 12 Storage section 13 Communications Department 102 Optimization Calculation Execution Unit 111 Input Section 112 Machine Learning Models 113 Output section

Claims

1. An optimization fuel calculation device that automatically calculates a biofuel with desired combustion performance that can be manufactured from oils and fats, A memory unit that holds data related to oils and fats, A machine learning model predicts combustion performance based on the type and amount of oil and fat, including (a) ignition, main combustion start time, and cetane number, which evaluate the ease of ignition during combustion of the mixed oil and fat; (b) maximum ROHR (Rate of Heat Release) time, which evaluates engine efficiency; (c) total combustion time, 95% afterburning period, 99% afterburning period, and 100% afterburning period, which evaluate the unburned portion; (d) SD' standard deviation, which evaluates multi-cylinder variation; and (e) ROHR (Rate of Heat Release) index, which evaluates the calorific value. An input unit for inputting information regarding the combustion performance of the desired biofuel, A learning processing execution unit that learns the settings in the machine learning model based on the training data, The system includes an optimization calculation execution unit that performs optimization calculations for producing biofuels from oils and fats using the aforementioned machine learning model, The data relating to oils and fats stored in the memory unit includes the learning data, The aforementioned types of fats and oils are beef tallow, lard, palm oil, and tempura oil. The optimization calculation execution unit performs the optimization calculation using the machine learning model based on the combustion performance of the desired biofuel input via the input unit, and calculates a biofuel having the desired combustion performance by determining the optimal oil and fat mixing ratio. An optimization fuel calculation device characterized in that the type of biofuel is a liquid fuel.

2. The learning data includes the type and amount of oil and fat, and combustion performance, which includes (a) ignition, main combustion start time, and cetane number, which are evaluations of the ease of ignition when the mixed oil and fat is burned, (b) ROHR (Rate of Heat Release) maximum time, which is an evaluation of engine efficiency, (c) total combustion period, afterburn period 95%, afterburn period 99%, and afterburn period 100%, which are evaluations of the unburned portion, (d) SD' standard deviation, which is an evaluation of multi-cylinder variation, and (e) ROHR (Rate of Heat Release) index, which is an evaluation of the calorific value. The optimization fuel calculation device according to claim 1, characterized in that the setting values ​​obtained by the learning process in the learning process execution unit are stored in the storage unit.

3. An optimized fuel calculation method for automatically calculating a biofuel with desired combustion performance that can be produced from oils and fats, A storage step for storing data related to oils and fats, An input step in which information regarding the combustion performance of the desired biofuel is entered, A learning process execution step that learns the settings in the machine learning model based on the training data, The process includes an optimization calculation execution step which performs an optimization calculation for producing biofuel from oils and fats using the aforementioned machine learning model, The aforementioned machine learning model predicts combustion performance based on the type and amount of oil and fat, including (a) ignition, main combustion start time, and cetane number, which are evaluations of the ease of ignition during combustion of the mixed oil and fat; (b) maximum ROHR (Rate of Heat Release) time, which is an evaluation of engine efficiency; (c) total combustion time, afterburning period 95%, afterburning period 99%, and afterburning period 100%, which are evaluations of the unburned portion; (d) SD' standard deviation, which is an evaluation of multi-cylinder variation; and (e) ROHR (Rate of Heat Release) index, which is an evaluation of calorific value. The data relating to fats and oils stored in the memory step includes the learning data, The aforementioned types of fats and oils are beef tallow, lard, palm oil, and tempura oil. In the optimization calculation execution step, based on the combustion performance of the desired biofuel input in the input step, the optimization calculation is performed using the machine learning model to determine the optimal oil and fat mixing ratio, thereby calculating a biofuel having the desired combustion performance. The optimization fuel calculation method is characterized in that the type of biofuel is a liquid fuel.

4. A program used in an optimization fuel calculation device that automatically calculates biofuels with desired combustion performance that can be produced from oils and fats, A storage step for storing data related to oils and fats, An input step in which information regarding the combustion performance of the desired biofuel is entered, A learning process execution step that learns the settings in the machine learning model based on the training data, The process includes an optimization calculation execution step which performs an optimization calculation for producing biofuel from oils and fats using the aforementioned machine learning model, The aforementioned machine learning model predicts combustion performance based on the type and amount of oil and fat, including (a) ignition, main combustion start time, and cetane number, which are evaluations of the ease of ignition during combustion of the mixed oil and fat; (b) maximum ROHR (Rate of Heat Release) time, which is an evaluation of engine efficiency; (c) total combustion time, afterburning period 95%, afterburning period 99%, and afterburning period 100%, which are evaluations of the unburned portion; (d) SD' standard deviation, which is an evaluation of multi-cylinder variation; and (e) ROHR (Rate of Heat Release) index, which is an evaluation of calorific value. The data relating to fats and oils stored in the memory step includes the learning data, The aforementioned types of fats and oils are beef tallow, lard, palm oil, and tempura oil. In the optimization calculation execution step, based on the combustion performance of the desired biofuel input in the input step, the optimization calculation is performed using the machine learning model to determine the optimal oil and fat mixing ratio, thereby calculating a biofuel having the desired combustion performance. The program is characterized in that the type of biofuel is a liquid fuel.

Images