A control method for a combustion and emission system of an ammonia-hydrogen fuel internal combustion engine

Abstract

The present application relates to a kind of ammonia hydrogen fuel internal combustion engine, specifically to a kind of ammonia hydrogen fuel internal combustion engine combustion and emission system control method, using LSTM neural network to carry out prediction control, including leakage safety detection, combustion and pipeline detection and post-processing detection;Leakage safety detection is used to detect whether the gas storage device exists leakage;Combustion and pipeline detection are used to detect whether the combustion appears knock and whether the pipeline exists leakage;Post-processing detection is used to detect whether the ammonia content and nitrogen oxide content in the treated exhaust gas exceed the standard.Compared with the prior art, the present application solves the problem that the response time difference of various sensors in the prior art is large, which makes it difficult to synchronize when applied to the ammonia hydrogen fuel internal combustion engine emission system.The present application realizes the prediction control of ammonia sensor and the entire emission system through LSTM neural network, greatly weakening the response time difference problem between various sensors, and ensuring the normal operation and control of the emission system.

Description

Technical Field

[0001] This invention relates to an ammonia-hydrogen fuel internal combustion engine, and more specifically to a control method for the combustion and emission system of an ammonia-hydrogen fuel internal combustion engine. Background Technology

[0002] Ammonia-hydrogen fusion fuel, as a zero-carbon clean fuel, has already attracted the attention of researchers in the heavy-duty vehicle, shipbuilding, and chemical industries. It integrates the advantages of both ammonia and hydrogen fuels, overcoming the shortcomings of single fuels, and offers benefits such as convenient storage, rapid combustion, and clean and environmentally friendly characteristics. The application of ammonia-hydrogen fusion fuel has become a research hotspot in the transportation sector.

[0003] Currently, there are related patents and system concepts for ammonia-hydrogen fusion fuel internal combustion engine systems. However, due to the special physical properties of ammonia-hydrogen fuel, it poses fire hazards and is explosive, while ammonia itself is toxic and corrosive. Structurally, the ammonia-hydrogen fusion fuel internal combustion engine system requires a large volume of liquid ammonia, as well as an ammonia cracker and mixture cylinder placed in the exhaust pipe, significantly increasing the risk area of ​​the ammonia-hydrogen internal combustion engine. Furthermore, in terms of control, the in-cylinder and exhaust aftertreatment systems of the ammonia-hydrogen fusion fuel internal combustion engine require the combined use of various types of sensors for monitoring and regulation. The different response times of these sensors (currently, the response time range of ammonia sensors is 10-100 seconds, significantly different from conventional sensors) increase the difficulty of monitoring and controlling the ammonia-hydrogen internal combustion engine system. Moreover, since the engine is constantly running during driving, each operating cycle is interconnected, making detection and control also time-series correlated. Therefore, a predictive method that can predict system combustion and emissions in advance is needed to ensure the normal operation of the system and the safety and functionality of passengers.

[0004] Chinese patent CN114776417A discloses an exhaust aftertreatment system for a hydrogen-ammonia fuel engine. This system is mainly based on the system structure, but it does not provide a suitable solution for the control part. Since no ammonia sensor is set up, the ammonia content in the exhaust gas can only be calculated theoretically, and its practical application capability is weak.

[0005] Chinese patent CN115111031A discloses a pollutant emission treatment system based on an ammonia-hydrogen fuel power system. However, this system cannot ignore the problem of delayed monitoring information caused by the long response time of the ammonia sensor. This results in the failure to detect problems in the ammonia flow section in a timely manner, which may lead to safety accidents.

[0006] Therefore, for ammonia-hydrogen fuel internal combustion engine systems, in addition to a reasonable overall system structure, it is also necessary to establish a good control scheme so that the signals from the various sensors can reflect the same engine operating cycle, or reduce the difference in response time between them. Summary of the Invention

[0007] The purpose of this invention is to provide a control method for the combustion and emission system of an ammonia-hydrogen fuel internal combustion engine to solve at least one of the above-mentioned problems. This method addresses the issue of large differences in the response times of various sensors in the prior art, which makes synchronization difficult when applied to an ammonia-hydrogen fuel internal combustion engine system. Based on CN115111031A, this invention sets up a control scheme with structural improvements. By using a Long Short-Term Memory (LSTM) network, predictive control of the ammonia sensor and the entire emission system is achieved, which greatly reduces the problem of response time differences between various sensors and ensures the normal operation and control of the emission system.

[0008] The objective of this invention is achieved through the following technical solution:

[0009] A control method for the combustion and emission system of an ammonia-hydrogen fuel internal combustion engine, employing an LSTM neural network for predictive regulation, including leak safety detection, combustion and pipeline detection, and aftertreatment detection;

[0010] The aforementioned leak safety detection is used to detect whether there is a leak in the gas storage device;

[0011] The aforementioned combustion and pipeline detection is used to detect whether detonation occurs during combustion and whether there is a leak in the pipeline;

[0012] The post-treatment detection is used to detect whether the ammonia and nitrogen oxide content in the treated exhaust gas exceeds the standard.

[0013] Preferably, the emission system includes a controller, an engine, a cracked gas storage tank, an ammonia fuel tank, an electrically heated catalytic converter, a selective catalytic reduction device, an ammonia oxidation catalyst, and an ammonia cracker;

[0014] The electrically heated catalytic converter, selective catalytic reduction device, and ammonia oxidation catalyst are sequentially connected to the engine's exhaust pipe. The ammonia fuel tank supplies ammonia to the exhaust pipe located upstream of the selective catalytic reduction device for treating the exhaust gas emitted by the engine.

[0015] The ammonia cracker is fitted outside the exhaust pipe of the engine and is connected to an ammonia fuel tank and a cracked gas storage tank respectively. The cracked gas storage tank is connected to the intake pipe of the engine and is used to recover the heat of the exhaust gas to crack ammonia into cracked gas and use the cracked gas as fuel for the engine.

[0016] The controller is used to control the emission system.

[0017] The electrically heated catalytic converter provides reaction energy for subsequent catalytic reactions, the selective catalytic reduction unit reduces nitrogen oxides to nitrogen, and the ammonia oxidation catalyst oxidizes excess ammonia to nitrogen. The treated exhaust gas mainly consists of nitrogen and water vapor.

[0018] Preferably, the emission system further includes:

[0019] The first ammonia sensor is installed inside the ammonia fuel tank.

[0020] The second ammonia sensor and hydrogen sensor are installed inside the cracked gas storage tank.

[0021] The controller is electrically connected to the first ammonia sensor, the second ammonia sensor, and the hydrogen sensor respectively, and is used to realize leakage safety detection.

[0022] Preferably, the leakage safety detection includes the following steps:

[0023] S11: Collect ammonia concentration data in the ammonia fuel tank through the first ammonia sensor, and collect ammonia and hydrogen concentration data in the cracked gas storage tank through the second ammonia sensor and hydrogen sensor, respectively.

[0024] S12: The LSTM neural network predicts whether there is a leak in the ammonia fuel tank and / or cracked gas storage tank in subsequent engine cycles based on the data collected in step S11.

[0025] If a leak is predicted, the controller will sound an alarm and instruct the pipeline connected to the leaking tank to cut off the alarm and control measures before the fault occurs.

[0026] According to actual tests, this LSTM neural network can predict the running state after about 10 cycles, demonstrating good predictive performance and providing sufficient buffer time for alarm, correction and other operations.

[0027] Preferably, the emission system further includes:

[0028] An ion current sensor installed inside the engine,

[0029] The first pressure sensor is installed on the first connecting pipe (cracking gas injection pipe) between the pyrolysis gas storage tank and the inlet pipe.

[0030] The second pressure sensor is installed on the second connecting pipe (ammonia injection pipe) between the ammonia fuel tank and the exhaust pipe.

[0031] The controller is electrically connected to the ion current sensor, the first pressure sensor, and the second pressure sensor, respectively, to realize combustion and pipeline detection.

[0032] Preferably, the first connecting pipe is provided with a first nozzle at its end, through which pyrolysis gas is injected into the intake pipe; the second connecting pipe is provided with a second nozzle at its end, through which ammonia gas is injected into the exhaust pipe.

[0033] Preferably, the detection of knocking during combustion and pipeline testing includes the following steps:

[0034] S21: Collect ion current data reflecting the combustion state in the current engine cylinder through an ion current sensor;

[0035] S22: The LSTM neural network predicts the combustion state in the engine during subsequent cycles based on the data collected in step S21;

[0036] If the engine is predicted to have a tendency to knock in subsequent cycles, the controller will adjust the equivalence ratio of cracked gas and air input to the engine to reduce the tendency of engine knock in subsequent cycles, while also reducing the generation of nitrogen oxides.

[0037] Preferably, the detection of pipeline leaks in the combustion and pipeline inspection includes the following steps:

[0038] S31: Collect pressure data of the first connecting pipe through the first pressure sensor, and collect pressure data of the second connecting pipe through the second pressure sensor;

[0039] S32: The controller determines whether there is a leak in the first connecting pipe and / or the second connecting pipe based on the data collected in step S31.

[0040] Leakage can be determined when: the pressure data collected by the pressure sensor fluctuates significantly and exceeds the reasonable range; or the pressure data collected by the pressure sensor is significantly lower than the design pressure and the normal operating range.

[0041] Preferably, the emission system further includes:

[0042] The first nitrogen oxide sensor and the first temperature sensor are located on the exhaust pipe upstream of the electrically heated catalytic converter.

[0043] The second temperature sensor is installed at the inlet of the electrically heated catalytic converter.

[0044] The third temperature sensor is installed at the outlet of the electrically heated catalytic converter.

[0045] The third ammonia sensor is installed at the inlet of the selective catalytic reduction unit.

[0046] The fourth ammonia sensor and the second nitrogen oxide sensor are located at the end of the exhaust pipe.

[0047] The controller is electrically connected to the first nitrogen oxide sensor, the first temperature sensor, the second temperature sensor, the third temperature sensor, the third ammonia sensor, the fourth ammonia sensor, and the second nitrogen oxide sensor, respectively, for post-processing detection.

[0048] Preferably, the post-processing detection includes the following steps:

[0049] S45: Collect nitrogen oxide concentration data upstream of the exhaust pipe using the first nitrogen oxide sensor;

[0050] S46: The controller calculates the amount of ammonia gas input from the ammonia fuel tank into the exhaust pipe based on the data collected in step S45;

[0051] S47: Collect concentration data of ammonia and nitrogen oxides downstream of the exhaust pipe through the third ammonia sensor, the fourth ammonia sensor, and the second nitrogen oxide sensor.

[0052] S48: The LSTM neural network predicts whether the ammonia and / or nitrogen oxide content in the exhaust gas of the engine exceeds the standard based on the data collected in steps S45 and S47.

[0053] If the predicted content exceeds the standard, the controller will instruct adjustments to the amount of ammonia entering the exhaust pipe, the heating power of the electrically heated catalytic converter, and so on.

[0054] The LSTM neural network determines the engine's NO based on the data collected in step S45. x Whether the emissions are normal and whether the ammonia and / or nitrogen oxide content in the combustion flue gas exceeds the standard. If the detected value exceeds the maximum processing value of the subsequent selective catalytic reduction device and ammonia oxidation catalyst, it means that there is a problem with the combustion in the engine cylinder. There may be special circumstances such as misfire or spark plug damage that lead to unburned combustion. The controller will issue an alarm.

[0055] Preferably, the post-processing detection further includes the following steps:

[0056] S41: Collect temperature data upstream of the exhaust pipe using the first temperature sensor;

[0057] S42: The controller determines whether the emission system is in a cold start based on the data collected in step S41;

[0058] S43: If it is a cold start, continue to step S44; if it is not a cold start, continue to step S45.

[0059] S44: The controller instructs the electrically heated catalytic converter to heat up, and adjusts the heating amount of the electrically heated catalytic converter based on the temperature data collected by the second and third temperature sensors respectively, and continues to step S45.

[0060] Long Short-Term Memory (LSTM) networks are artificial neural networks used in artificial intelligence and deep learning. In actual training, each layer of a recurrent neural network consists of several memory cells. The output of a memory cell at the current time step affects the memory cell at the next time step. In multi-layer recurrent neural networks, the output of the previous layer serves as the input to the next layer. During training, the loss function decreases iteratively, and the training terms are continuously updated, ultimately yielding a neural network model suitable for solving a specific problem. The control method proposed in this paper is based on a trained LSTM model. "Trained" means that the model's training parameter set contains a large number of pressure, temperature, and emission parameters from normal operating cycles, as well as parameters from typical functional failure cycles; all of this data can be derived from known systems and existing technologies. The trained LSTM model is written into the controller for prediction and control. Similarly, if emission regulations change subsequently, the control threshold can be changed by modifying the LSTM model's training set and various parameters.

[0061] The working principle of this invention is as follows:

[0062] Based on the configuration of the ammonia-hydrogen internal combustion engine and aftertreatment system, from the perspectives of storage safety and functional assurance, this paper uses real-time data from sensors for ammonia, hydrogen, nitrogen oxides, temperature, pressure, and ion current to predict and control the engine's operating status and provide safety warnings before a fault occurs, utilizing LSTM temporal deep neural networks.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] This invention proposes a prediction and control method based on LSTM neural networks. LSTM is a mature method whose advantage lies in its control and prediction being related to time series. The LSTM model considers the influence of the previous input on the next, making it highly suitable for predicting continuously operating engines, as engine operation is also periodic with interactive relationships between cycles. The LSTM neural network can predict the next cycle of the engine by forecasting the subsequent output values ​​of all sensors, ensuring that all sensors reflect data from the same cycle.

[0065] This invention adds various types of sensors to detect the status of the emission system, which improves operational safety and provides sufficient data support for the LSTM neural network, ensuring sufficient training for the LSTM neural network and improving the prediction accuracy and reliability of the LSTM neural network.

[0066] This invention further improves the structure and establishes the control method based on the system proposed in CN115111031A, and provides a practical and multifunctional emission system and control method. Attached Figure Description

[0067] Figure 1 This is a schematic diagram of the emission system.

[0068] Figure 2 This is a flowchart illustrating the control method.

[0069] Figure 3 A flowchart illustrating the process of leak safety detection;

[0070] Figure 4 This is a flowchart illustrating the combustion and pipeline inspection process;

[0071] Figure 5 This is a schematic diagram of the post-processing detection process, where EHC is an electrically heated catalytic converter, SCR is a selective catalytic reduction device, and ASC is an ammonia oxidation catalyst.

[0072] In the diagram: 1-Ion current sensor; 2-First nozzle; 3-Third ammonia sensor; 4-Cracked gas storage tank; 5-Fourth ammonia sensor; 6-Controller; 7-Engine; 8-Ammonia fuel tank; 9-Electro-heated catalytic converter; 10-Selective catalytic reduction device; 11-Ammonia oxidation catalyst; 12-Ammonia cracker; 13-Second nozzle; 14-First nitrogen oxide sensor; 15-First pressure sensor; 16-Second ammonia sensor; 17-Hydrogen sensor; 18-Second pressure sensor; 19-First ammonia sensor; 20-First temperature sensor; 21-Second temperature sensor; 22-Third temperature sensor. Detailed Implementation

[0073] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0074] Example 1

[0075] A combustion and emission system for an ammonia-hydrogen fuel internal combustion engine, such as Figure 1 As shown, it mainly includes an engine 7, an ammonia fuel tank 8, a cracked gas storage tank 4, an ammonia cracker 12, an electrically heated catalytic converter 9, a selective catalytic reduction device 10, an ammonia oxidation catalyst 11, and a controller 6.

[0076] An electrically heated catalytic converter 9, a selective catalytic reduction device 10, and an ammonia oxidation catalyst 11 are sequentially connected to the exhaust pipe of the engine 7. An ammonia fuel tank 8 is connected upstream of the exhaust pipe. Ammonia is injected into the exhaust pipe through a first nozzle 2 at the end of a second connecting pipe (the pipe between the ammonia fuel tank 8 and the exhaust pipe) for after-treatment of the exhaust gas produced by the combustion of the engine 7. An ammonia cracker 12 is mounted on the exhaust pipe, located before the electrically heated catalytic converter 9 and after the first nozzle 2. The ammonia cracker 12 connects the ammonia fuel tank 8 to the cracked gas storage tank 4. The ammonia cracker 12 recovers the heat contained in the exhaust gas in the exhaust pipe to crack the ammonia fuel tank 8. The nitrogen, hydrogen, and unreacted ammonia produced by cracking are introduced into the cracked gas storage tank 4 for storage. The cracked gas storage tank 4 is further connected to the intake pipe of the engine 7. Cracked gas, used as fuel, is injected into the intake pipe through a second nozzle 13 at the end of a first connecting pipe (the pipe between the cracked gas storage tank 4 and the intake pipe). The electrically heated catalytic converter 9 heats the exhaust gas (including flue gas from combustion and ammonia from the ammonia fuel tank 8) after heat exchange in the ammonia cracker 12. Then, the nitrogen oxides in the exhaust gas are reduced to nitrogen in the selective catalytic reduction unit 10. Finally, the excess ammonia is further oxidized to nitrogen in the ammonia oxidation catalyst 11, so that the exhaust gas discharged from the emission system after final treatment is mainly nitrogen and water vapor.

[0077] The reactions occurring in the aforementioned engine 7, ammonia cracker 12, selective catalytic reduction unit 10, and ammonia oxidation catalyst 11 are as follows:

[0078] Engine 7: NH3 + H2 + O2 + N2 → N2 + H2O + NO x (Unbalanced);

[0079] Ammonia cracker 12: 2NH3 → N2 + 3H2;

[0080] Selective catalytic reduction unit 10:

[0081] Standard SCR: 4NH3 + 4NO + O2 → 4N2 + 6H2O;

[0082] Rapid SCR: 2NH3 + NO + NO2 → 2N2 + 3H2O;

[0083] Slow SCR: 8NH3 + 6NO2 → 7N2 + 12H2O;

[0084] Ammonia oxidation catalyst 11: 4NH3 + 3O2 → 2N2 + 6H2O.

[0085] The combustion and emission system of this ammonia-hydrogen fuel internal combustion engine is also equipped with various sensors electrically connected to the controller 6, including an ion current sensor 1, an ammonia sensor, a hydrogen sensor 17, a nitrogen oxide sensor, a pressure sensor, and a temperature sensor, which are connected to the controller 6 via I / O interfaces. The controller 6 can be a control device such as a computer. The controller 6 can be an Infineon MCU-AURIX TC397 series, which is highly integrated, reduces complexity, and significantly saves costs; it is widely used in autonomous driving and ADAS system control.

[0086] An ion current sensor 1 is installed inside the engine 7 to detect the value of the ion current in order to monitor whether knocking occurs inside the engine 7.

[0087] Four ammonia sensors are provided: the first ammonia sensor 19 is located in the ammonia fuel tank 8, the second ammonia sensor 16 is located in the cracked gas storage tank 4, the third ammonia sensor 3 is located at the inlet end of the selective catalytic reduction device 10, and the fourth ammonia sensor 5 is located at the tail end of the exhaust pipe (after the ammonia oxidation catalyst 11), which are used to monitor the ammonia concentration at the corresponding locations.

[0088] A hydrogen sensor 17 is installed inside the cracked gas storage tank 4 to monitor the hydrogen concentration inside the cracked gas storage tank 4.

[0089] Two nitrogen oxide sensors are provided. The first nitrogen oxide sensor 14 is located on the exhaust pipe of the engine 7, before the first nozzle 2. The second nitrogen oxide sensor is located at the tail end of the exhaust pipe (after the ammonia oxidation catalyst 11) to monitor the nitrogen oxide concentration upstream of the exhaust pipe, that is, the nitrogen oxide concentration in the flue gas generated by the combustion of the engine 7.

[0090] Two pressure sensors are provided: a first pressure sensor 15 is installed on the first connecting pipe, and a second pressure sensor 18 is installed on the second connecting pipe, which are used to monitor the airflow pressure in the connecting pipe to determine whether there is a leak.

[0091] There are three temperature sensors: the first temperature sensor 20 is located on the exhaust pipe of the engine 7, in front of the first nozzle 2; the second temperature sensor 21 is located at the inlet end of the electrically heated catalytic converter 9; and the third temperature sensor 22 is located at the outlet end of the electrically heated catalytic converter 9, for monitoring the temperature at the corresponding locations.

[0092] The controller 6 uses an LSTM neural network for monitoring, prediction, and control. By coordinating the values ​​acquired by various sensors, it predicts the state of the engine 7 in subsequent cycles, thereby enabling advance adjustments and / or alarms. It fully leverages the combined effect of feedforward and feedback to achieve safe and stable operation and control.

[0093] In addition to the aforementioned sensors, an ammonia sensor and a hydrogen sensor 17 can be further installed in the driver's cab and connected to the controller 6 as data sources.

[0094] LSTM neural networks primarily perform three functions: 1. Prediction and monitoring of in-cylinder combustion state based on ion current. Since an ion current peak is generated during abnormal combustion such as knocking, the ion current value of each cycle is obtained to predict the trend of ion current changes in subsequent cycles. If an ion current peak is detected in a subsequent cycle, an alarm is issued and the fuel injection ratio is adjusted to achieve normal combustion in subsequent cycles. 2. Prediction and monitoring of pipeline leakage state based on pressure sensors. Due to the corrosive and explosive nature of ammonia and hydrogen, leaks of ammonia and hydrogen pose a risk to the use of ammonia-hydrogen combustion internal combustion engine systems. To address and control large-scale leaks of ammonia and hydrogen fuel before they occur, neural networks are used to predict the current pressure state and subsequent pressure fluctuations. If the pressure fluctuation is large or a continuous pressure drop occurs, an alarm is triggered to check the ammonia and hydrogen pipelines. 3. Internal Combustion Engine Emission Prediction and Monitoring Based on Nitrogen Oxide and Ammonia Sensors: Traditionally, external ammonia emissions are simulated and calculated offline using fixed mathematical models. This invention proposes using LSTM neural networks for online prediction and monitoring. The difference lies in that the model can update data in real time during engine operation, making it suitable for a wider range of operating conditions. Secondly, the different response times of various sensors make it difficult to reflect the combustion and emission status of the engine at the same moment. By using LSTM neural networks to record and collect sensor data that has occurred and to predict subsequent cycles, errors caused by different response times can be compensated to a certain extent, thereby achieving accurate prediction and control for each cycle.

[0095] Example 2

[0096] A control method for the combustion and emission system of an ammonia-hydrogen fuel internal combustion engine, such as Figure 2-5 As shown, an LSTM neural network is used for predictive control, comprising three detection components: leak safety detection, combustion and pipeline detection, and post-treatment detection. The detection and control sequence is as follows: Figure 2 As shown, proceed in sequence.

[0097] Leakage safety detection, such as Figure 3As shown, during the operation of the emission system, the first ammonia sensor 19, the second ammonia sensor 16, and the hydrogen sensor 17 collect ammonia / hydrogen concentration data at their respective locations, and the first pressure sensor 15 and the second pressure sensor 18 collect pipeline pressure data at their respective locations. These data are then aggregated into the controller 6. Based on the real-time acquired data and the data from previous cycles stored in the LSTM neural network, the controller predicts whether the hydrogen fuel tank and the cracked gas storage tank 4 are leaking. The LSTM neural network continues to iterate and calculate this data to predict the data for subsequent cycles of the engine 7. If a leak is predicted, the controller 6 issues an alarm signal before the subsequent cycle and simultaneously instructs the associated pipelines to shut off and trigger an alarm.

[0098] Combustion and pipeline inspection, such as Figure 4 As shown, during the operation of the emission system: 1) Knock detection: Ion current sensor 1 collects ion current data (reflecting combustion state) inside engine 7 (cylinder) and transmits it to controller 6. Based on the real-time acquired data and the data from previous cycles stored in the LSTM neural network, it predicts whether knock will occur inside engine 7. The LSTM neural network iteratively calculates this set of data to predict the data for subsequent cycles of engine 7. If knock is predicted or there is a tendency for knock, the ratio of input cracked gas to air is adjusted to ensure normal engine operation, improve engine reliability, reduce the tendency for engine knock, and reduce the generation of nitrogen oxides. 2) Pipeline leakage detection: First pressure sensor 15 and second pressure sensor 18 collect pressure data at their respective locations and transmit it to controller 6. Controller 6 compares the real-time acquired data with the design pressure or calculates the pressure fluctuation over a certain period of time. If either exceeds the set range, it indicates that there is a leak in the connecting pipeline. Controller 6 issues an alarm signal and simultaneously instructs the associated pipeline to be shut off.

[0099] Post-processing inspection, such as Figure 5As shown, during the operation of the emission system, the first temperature sensor 20 and the first nitrogen oxide sensor 14 collect data from their respective locations and feed it back to the controller 6. The controller 6 determines whether it is a cold start based on the temperature data from the first temperature sensor 20, and calculates the amount of ammonia to be input into the exhaust pipe based on the nitrogen oxide concentration data measured by the first nitrogen oxide sensor 14. If it is determined to be a cold start, the controller 6 instructs the electrically heated catalytic converter 9 to operate, raising the temperature of the exhaust gas in the exhaust pipe to reach the reaction temperature of catalytic oxidation, which is usually set above 200°C. The second temperature sensor 21 measures the outlet temperature of the exhaust gas in the electrically heated catalytic converter 9 and feeds the temperature data back to the controller 6 to determine whether the set value / set range has been reached. If the target has not been reached, the heating power of the electrically heated catalytic converter 9 is increased (in other embodiments, an LSTM neural network can be further combined to predict the heating effect and adjust the heating power in advance). If the target has been reached, the heating power is adjusted to maintain a constant temperature inside the electrically heated catalytic converter 9 while maintaining real-time temperature monitoring. The preheated exhaust gas enters the selective catalytic reduction unit 10. The third ammonia sensor 3 monitors the ammonia concentration at the inlet. The controller 6 adjusts the amount of ammonia injected into the exhaust pipe from the ammonia fuel tank 8 based on the difference between real-time data and calculated values. The exhaust gas sequentially passes through the selective catalytic reduction unit 10 and the ammonia oxidation catalyst 11, reducing nitrogen oxides to nitrogen and oxidizing excess ammonia to nitrogen. The treated exhaust gas is then measured by the fourth ammonia sensor 5 and the second nitrogen oxide sensor located at the end of the exhaust pipe to determine whether the exhaust gas exceeds the standard. If the ammonia emission in the exhaust gas is unqualified, it is improved by increasing the efficiency of the ammonia oxidation catalyst 11 or reducing the amount of ammonia injected into the exhaust pipe. If the nitrogen oxide emission in the exhaust gas is unqualified, it is improved by increasing the amount of ammonia injected into the exhaust pipe and / or adjusting the heating power of the electrically heated catalytic converter 9 to regulate the exhaust gas temperature. At the same time, the concentration data measured by the fourth ammonia sensor 5 and the second nitrogen oxide sensor are also incorporated into the calculation by an LSTM neural network to predict whether the exhaust gas emissions in subsequent cycles will exceed the standard, and adjustments are made in advance based on the conclusions. If the detection values ​​of each sensor exceed the maximum processing limit of the selective catalytic reduction device 10 and the ammonia oxidation catalyst 11, it means that there is a problem with combustion in the engine cylinder 7, which may be due to special circumstances such as misfire or spark plug damage, which may lead to unburned combustion. The controller 6 will then issue an alarm.

[0100] If it is determined to be a non-cold start, the controller 6 instructs to maintain a constant temperature inside the electrically heated catalytic converter 9 and continue with subsequent steps.

[0101] This invention addresses the challenges of complex structures in ammonia-hydrogen fusion fuel internal combustion engine systems; complex ammonia fuel pollutant emission control; high safety risks associated with ammonia-hydrogen fuel storage and leakage; and significant differences in response times among various sensors, making synchronization difficult. It proposes a detection, prediction, and control method for ammonia-hydrogen powered internal combustion engines. Based on the configuration of the ammonia-hydrogen internal combustion engine and its aftertreatment system, and considering both storage safety and functional assurance, this invention utilizes real-time data from sensors 1 (ammonia, hydrogen, nitrogen oxides, temperature, pressure, and ion current) and an LSTM temporal deep neural network to predict and control the engine's operating status and provide safety warnings before a fault occurs.

[0102] Currently, the response time of domestic ammonia sensors is generally between 10-100 seconds, and they require stringent conditions such as temperature and humidity. However, the response time of other sensors in the system is much shorter than that of the ammonia sensor. Considering the movement of exhaust gas in the exhaust pipe, different sensors placed in different locations cannot reflect the combustion and emission states at the same moment. Furthermore, due to the toxicity of ammonia, its leakage and emission limits are extremely low. Therefore, a method is needed to enable the signals from multiple sensors to reflect the same cycle, or to reduce the difference in their response times.

[0103] Because ammonia-hydrogen fuel contains hydrogen, knocking in hydrogen engines remains a significant issue affecting engine reliability and emissions. While this patent utilizes ammonia cracking to produce a large amount of hydrogen, the risk of knocking in ammonia-hydrogen engines still exists. This system incorporates an ion current sensor, using data from the sensor and based on LSTM (Laser-Sensitive Timing) to predict and control the likelihood of in-cylinder knocking.

[0104] This system addresses the issues of ammonia sensors having a long response time and a significant time lag compared to other sensors; the toxicity of ammonia gas; and the explosiveness and combustion pollutant (NOx) of ammonia-hydrogen fuel. x (And NH3) emission control. This paper proposes using Long Short-Term Memory (LSTM) for predictive control of an ammonia sensor system and a control strategy for an ammonia-hydrogen fuel internal combustion engine to achieve the system's goals of safe, low-emission, and low-energy consumption.

[0105] It should be noted that the LSTM neural network in this method is applied to the combustion and emission system control of ammonia-hydrogen fuel internal combustion engines, and its implementation method and operation logic are only adapted.

[0106] The system incorporates more sensors to detect the status of the ammonia-hydrogen internal combustion engine. For example, if ammonia leaks, both the pressure sensor and the ammonia sensor will change, improving safety while also providing data acquisition support for the LSTM neural network. Through multiple iterations, prediction accuracy can be improved, enhancing system reliability. Furthermore, the system includes a cracked gas storage tank 4 with corresponding sensors, allowing for independent adjustment of parameters in this cracked gas path.

[0107] In addition to detecting conventional emissions such as nitrogen oxides, the system is also equipped to detect leaks and emissions of flammable and explosive gases such as ammonia and hydrogen, effectively improving the safety of the system during operation. Combined with the prediction of LSTM neural network, it effectively improves the early warning and effectiveness of alarms.

[0108] The above description of the embodiments is provided to enable those skilled in the art to understand and use the invention. It will be apparent to those skilled in the art that various modifications can be made to these embodiments, and the general principles described herein can be applied to other embodiments without inventive effort. Therefore, the present invention is not limited to the above embodiments, and any improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope of the invention should be within the protection scope of the present invention.

Claims

1. A control method for the combustion and emission system of an ammonia-hydrogen fuel internal combustion engine, characterized in that, LSTM neural networks are used for predictive control, including leak safety detection, combustion and pipeline detection, and post-treatment detection. The aforementioned leak safety detection is used to detect whether there is a leak in the gas storage device; The aforementioned combustion and pipeline detection is used to detect whether detonation occurs during combustion and whether there is a leak in the pipeline; The post-treatment detection is used to detect whether the ammonia and nitrogen oxide content in the treated exhaust gas exceeds the standard. The system includes a controller (6), an engine (7), a cracked gas storage tank (4), an ammonia fuel tank (8), an electrically heated catalytic converter (9), a selective catalytic reduction device (10), an ammonia oxidation catalyst (11), and an ammonia cracker (12). The electric heating catalytic converter (9), selective catalytic reduction device (10) and ammonia oxidation catalyst (11) are connected in sequence to the exhaust pipe of the engine (7). The ammonia fuel tank (8) supplies ammonia to the exhaust pipe located upstream of the selective catalytic reduction device (10) for treating the exhaust gas discharged from the engine (7). The ammonia cracker (12) is fitted outside the exhaust pipe of the engine (7) and is connected to the ammonia fuel tank (8) and the cracked gas storage tank (4) respectively. The cracked gas storage tank (4) is connected to the intake pipe of the engine (7) and is used to recover the heat of the exhaust gas to crack ammonia and use the cracked gas as fuel for the engine (7). The controller (6) is used to control the system; The system also includes: The first ammonia sensor (19) is installed inside the ammonia fuel tank (8). The second ammonia sensor (16) and hydrogen sensor (17) are installed in the cracked gas storage tank (4). The first nitrogen oxide sensor (14) is installed on the exhaust pipe upstream of the electrically heated catalytic converter (9). The third ammonia sensor (3) is installed at the inlet of the selective catalytic reduction device (10). The fourth ammonia sensor (5) and the second nitrogen oxide sensor are installed at the end of the exhaust pipe. The controller (6) is connected to the first ammonia sensor (19), the second ammonia sensor (16) and the hydrogen sensor (17) respectively to realize leakage safety detection; and the controller (6) is electrically connected to the first nitrogen oxide sensor (14), the third ammonia sensor (3), the fourth ammonia sensor (5) and the second nitrogen oxide sensor respectively to realize post-processing detection. The aforementioned leak safety detection includes the following steps: S11: The concentration data of ammonia in the ammonia fuel tank (8) is collected by the first ammonia sensor (19), and the concentration data of ammonia and hydrogen in the cracked gas storage tank (4) are collected by the second ammonia sensor (16) and the hydrogen sensor (17), respectively. S12: The LSTM neural network predicts whether there is a leak in the ammonia fuel tank (8) and / or the cracked gas storage tank (4) in the subsequent cycles of the engine (7) based on the data collected in step S11; If a leak is predicted, the controller (6) will sound an alarm and instruct the pipeline connected to the leaking tank to be disconnected; The post-processing detection includes the following steps: S45: Collect nitrogen oxide concentration data upstream of the exhaust pipe through the first nitrogen oxide sensor (14); S46: The controller (6) calculates the amount of ammonia gas input from the ammonia fuel tank (8) into the exhaust pipe based on the data collected in step S45; S47: Collect concentration data of ammonia and nitrogen oxides downstream of the exhaust pipe through the third ammonia sensor (3), the fourth ammonia sensor (5), and the second nitrogen oxide sensor; S48: The LSTM neural network predicts whether the ammonia and / or nitrogen oxide content in the exhaust gas of the engine (7) exceeds the standard based on the data collected in step S47; If the predicted content exceeds the standard, the controller (6) instructs to adjust the amount of ammonia gas entering the exhaust pipe and the heating power of the electrically heated catalytic converter (9).

2. The control method for the combustion and emission system of an ammonia-hydrogen fuel internal combustion engine according to claim 1, characterized in that, The system also includes: An ion current sensor (1) is installed inside the engine (7). The first pressure sensor (15) is installed on the first connecting pipe between the pyrolysis gas storage tank (4) and the inlet pipe. The second pressure sensor (18) is installed on the second connecting pipe between the ammonia fuel tank (8) and the exhaust pipe. The controller (6) is electrically connected to the ion current sensor (1), the first pressure sensor (15), and the second pressure sensor (18) respectively, and is used to realize combustion and pipeline detection.

3. The control method for the combustion and emission system of an ammonia-hydrogen fuel internal combustion engine according to claim 2, characterized in that, The detection of knocking during combustion and pipeline testing includes the following steps: S21: Collect the current ion current data in the current engine (7) cylinder through the ion current sensor (1); S22: The LSTM neural network predicts the combustion state inside the engine (7) in subsequent cycles of the engine (7) based on the data collected in step S21; If knocking is predicted, the controller (6) instructs to adjust the ratio of cracked gas to air input to the engine (7).

4. The control method for the combustion and emission system of an ammonia-hydrogen fuel internal combustion engine according to claim 2, characterized in that, The detection of pipeline leaks in the combustion and pipeline inspection includes the following steps: S31: Collect pressure data of the first connecting pipe through the first pressure sensor (15), and collect pressure data of the second connecting pipe through the second pressure sensor (18); S32: The controller (6) determines whether there is a leak in the first connecting pipe and / or the second connecting pipe based on the data collected in step S31.

5. The control method for the combustion and emission system of an ammonia-hydrogen fuel internal combustion engine according to claim 1, characterized in that, The system also includes: The first temperature sensor (20) is installed on the exhaust pipe upstream of the electrically heated catalytic converter (9). The second temperature sensor (21) is installed at the inlet of the electrically heated catalytic converter (9). The third temperature sensor (22) is installed at the outlet of the electrically heated catalytic converter (9). The controller (6) is electrically connected to the first temperature sensor (20), the second temperature sensor (21) and the third temperature sensor (22) respectively, and is used to realize post-processing detection.

6. The control method for the combustion and emission system of an ammonia-hydrogen fuel internal combustion engine according to claim 5, characterized in that, The post-processing detection also includes the following steps: S41: Collect temperature data upstream of the exhaust pipe through the first temperature sensor (20); S42: The controller (6) determines whether the system is in a cold start based on the data collected in step S41; S43: If it is a cold start, continue to step S44; if it is not a cold start, continue to step S45. S44: The controller (6) instructs the electrically heated catalytic converter (9) to heat up, and adjusts the heating amount of the electrically heated catalytic converter (9) based on the temperature data collected by the second temperature sensor (21) and the third temperature sensor (22) respectively, and continues to step S45.

Images