A method and system for controlling exhaust gas temperature in a CNG engine

Abstract

The application provides a CNG engine exhaust temperature control method and system, which comprises the following steps: based on the component pre-detection data of the fuel supply pipeline and the historical working condition time sequence characteristics of the engine, a corresponding working condition change prediction model is constructed in combination with a preset neural network; the opening change rate of the throttle valve, the rotating speed of the turbocharger and the intake air temperature are collected and input into the working condition change prediction model to output corresponding CNG component prediction results and working condition foresight levels; the corresponding intake air preheating reference is determined according to the CNG component prediction results, the preheating power is dynamically adjusted according to the working condition foresight level, and the vortex adjustment parameters are dynamically corrected in combination with the actual exhaust temperature; the intake air preheating reference, the preheating power and the vortex adjustment parameters are corrected through a rolling optimization algorithm to generate corresponding target control data sets, and the actual working state of the engine is dynamically adjusted according to the target control data sets so that the exhaust temperature is kept in a stable interval. The application can improve the temperature control efficiency.

Description

Technical Field

[0001] This invention relates to the field of automotive technology, and in particular to a method and system for controlling the exhaust temperature of a CNG engine. Background Technology

[0002] As the commercial vehicle industry places increasing demands on economy and environmental protection, compressed natural gas (CNG) engines are widely used in the commercial vehicle sector due to their advantages of low fuel cost and low emissions, becoming an important direction for powertrain upgrades. However, in practical applications, the exhaust temperature control problem of CNG engines is a core bottleneck, especially the sudden rise in exhaust temperature under transient operating conditions, which seriously affects the reliability of engine operation and the lifespan of key components such as exhaust aftertreatment systems.

[0003] CNG's main component is methane (CH4), whose stable molecular structure results in a flame propagation speed much lower than that of gasoline. This leads to inherent problems in CNG engines, such as slow combustion and a long afterburning period, which can easily cause abnormally high exhaust temperatures. Existing exhaust temperature control strategies are mostly derived from traditional internal combustion engines and cannot be accurately adapted to the combustion characteristics of CNG engines, making it difficult to fundamentally solve the problem of high exhaust temperatures.

[0004] Furthermore, this control challenge is even more pronounced under transient conditions (especially rapid deceleration and acceleration): during rapid deceleration, combustion is unstable, and unburned gas escapes into the exhaust system, triggering afterburning and causing exhaust temperature to spike; during rapid acceleration, incremental gas cannot burn completely in time, and continuous combustion in the exhaust causes a rapid temperature increase. Existing control methods and systems are not designed for CNG combustion characteristics and transient conditions, making it impossible to achieve rapid and precise regulation, exacerbating component wear, reducing aftertreatment efficiency, and even causing safety hazards, thus hindering the improvement of CNG engine performance and reliable operation. Summary of the Invention

[0005] Based on this, the purpose of the present invention is to provide a CNG engine exhaust temperature control method and system to solve the problem that the existing control methods are not designed for CNG combustion characteristics and transient operating conditions, resulting in unstable exhaust temperature control.

[0006] The first aspect of the present invention proposes:

[0007] A method for controlling the exhaust temperature of a CNG engine specifically includes the following steps:

[0008] Based on the component pre-detection data of the fuel supply pipeline and the historical operating condition time series characteristics of the engine, a corresponding operating condition change prediction model is constructed by combining the preset neural network.

[0009] The throttle opening change rate, turbocharger speed and intake air temperature are collected and input into the operating condition change prediction model to output the corresponding CNG component prediction results and operating condition forecast level.

[0010] Based on the CNG component prediction results, the corresponding intake preheating benchmark is determined, and the preheating power is dynamically adjusted according to the operating condition forecast level. The vortex adjustment parameters are dynamically corrected in combination with the actual exhaust temperature.

[0011] The intake preheating reference, the preheating power, and the vortex adjustment parameters are corrected by a rolling optimization algorithm to generate a corresponding target control dataset. The actual operating state of the engine is then dynamically adjusted according to the target control dataset to keep the exhaust temperature within a stable range.

[0012] The beneficial effects of this invention are as follows: This technical solution precisely matches the combustion characteristics and transient operating condition control requirements of CNG. By combining component pre-detection data and historical operating condition time series characteristics with a neural network model, it achieves advance prediction of CNG components and operating condition levels, avoiding transient response lag. Based on the prediction results, it dynamically matches the intake preheating benchmark and power, and simultaneously combines exhaust temperature to correct swirl parameters, improving combustion uniformity and suppressing local overburning. Furthermore, the rolling optimization algorithm dynamically corrects the control parameters, ensuring real-time and accurate adaptation of the engine's operating state. This effectively solves the problem of unstable exhaust temperature control in existing technologies, ensuring that the exhaust temperature remains stable within the target range under all operating conditions, while improving combustion efficiency and operational reliability.

[0013] Furthermore, the step of inputting the data into the operating condition change prediction model to output the corresponding CNG component prediction results and operating condition look-ahead level includes:

[0014] The opening change rate, the rotational speed and the intake temperature are decoupled by multi-scale features. The corresponding cross-domain correlation features are extracted by combining the component pre-detection data. At the same time, several similar segments are matched in the historical operating condition time series features according to the cross-domain correlation features to generate the corresponding multi-source decoupling feature vector.

[0015] The historical working condition time series features are clustered according to working condition, and combined with the working condition type to which the multi-source decoupling feature vector belongs, the appropriate inference sub-model is called up to output the corresponding initial component prediction results and initial working condition look-ahead level.

[0016] The initial component prediction results and the initial operating condition prospective level are corrected to generate the corresponding CNG component prediction results and the operating condition prospective level.

[0017] Furthermore, the step of correcting the initial component prediction result and the initial operating condition look-ahead level to generate the corresponding CNG component prediction result and operating condition look-ahead level includes:

[0018] The initial component prediction deviation and the initial operating condition prospective level deviation are used as result variables. Simultaneously, the initial component prediction result and the initial operating condition prospective level are used as candidate cause variables. The core causal factors are located by the propensity score matching algorithm.

[0019] Based on the temporal variation pattern of the core causal factors, a corresponding factor influence intensity sequence is generated, and an MPC compensation model is constructed simultaneously based on the factor influence intensity sequence.

[0020] The engine hardware operating limit parameters are used as constraints of the MPC compensation model. Compensation coupling coefficients are introduced simultaneously. The initial component prediction results and the initial operating condition look-ahead level are corrected by the MPC compensation model to generate the CNG component prediction results and the operating condition look-ahead level accordingly.

[0021] Furthermore, the step of correcting the intake preheating reference, the preheating power, and the eddy current adjustment parameters using a rolling optimization algorithm to generate the corresponding target control dataset includes:

[0022] Redundant features in the intake preheating benchmark, the preheating power and the eddy current adjustment parameters are removed by multi-source data fusion algorithm in order to extract the core feature set containing operating condition features, component fluctuation features and environmental impact features.

[0023] Based on the aforementioned working condition prospective level and the aforementioned core feature set, the engine working conditions are divided into different clusters, and rolling optimization factors adapted to each cluster are set synchronously.

[0024] The core feature set is transformed based on the rolling optimization factor to generate the target control dataset.

[0025] Furthermore, the step of transforming the core feature set based on the rolling optimization factor to generate the target control dataset includes:

[0026] The core feature set is divided into a main feature layer, a secondary feature layer and an auxiliary feature layer. Simultaneously, the rolling optimization factor is decomposed by dimension to generate the corresponding time dimension factor, intensity dimension factor and compensation dimension factor.

[0027] The main feature layer is transformed to correspond to the time dimension factor, the secondary feature layer is transformed to correspond to the intensity dimension factor, and the auxiliary feature layer is transformed to correspond to the compensation dimension factor to generate the corresponding initial control dataset.

[0028] The initial control dataset is iteratively corrected to generate the target control dataset.

[0029] Furthermore, the step of dynamically adjusting the actual operating state of the engine based on the target control dataset to maintain the exhaust temperature within a stable range includes:

[0030] The core parameters of intake preheating, material flow regulation and fuel injection in the target control dataset are analyzed to construct the corresponding multi-parameter collaborative timing adjustment map;

[0031] Based on the multi-parameter coordinated timing adjustment spectrum, an air-fuel ratio adapted to CNG components is matched in a preset parameter database;

[0032] The engine's fuel injection parameters are adjusted according to the air-fuel ratio, and the engine's actual operating state is dynamically adjusted simultaneously based on the fuel injection parameters.

[0033] Furthermore, the step of dynamically adjusting the actual operating state of the engine based on the fuel injection parameters includes:

[0034] Based on the fuel injection parameters and the CNG component prediction results, corresponding injection target models are constructed according to engine cylinder groups;

[0035] With injection uniformity as the goal, the injection adjustment command for each group of cylinders is output through the injection target model, and the exhaust manifold temperature and exhaust composition of each group of cylinders are collected simultaneously to construct a corresponding temperature-composition dual-dimensional feedback verification mechanism.

[0036] The actual operating state of the engine is monitored and adjusted through the temperature-component dual-dimensional feedback verification mechanism.

[0037] The second aspect of the present invention proposes:

[0038] A CNG engine exhaust temperature control system, wherein the system comprises:

[0039] The module is used to construct a corresponding operating condition change prediction model based on the component pre-detection data of the fuel supply pipeline and the historical operating condition time series characteristics of the engine, combined with a preset neural network.

[0040] The data acquisition module is used to collect the throttle opening change rate, turbocharger speed and intake air temperature, and input them into the operating condition change prediction model to output the corresponding CNG component prediction results and operating condition forecast level.

[0041] The processing module is used to determine the corresponding intake preheating benchmark based on the CNG component prediction result, and simultaneously dynamically adjust the preheating power according to the operating condition look-ahead level, and dynamically correct the vortex adjustment parameters in combination with the actual exhaust temperature.

[0042] The adjustment module is used to correct the intake preheating reference, the preheating power and the vortex adjustment parameters through a rolling optimization algorithm to generate a corresponding target control dataset, and simultaneously dynamically adjust the actual operating state of the engine according to the target control dataset so that the exhaust temperature is kept in a stable range.

[0043] Furthermore, the acquisition module is specifically used for:

[0044] The opening change rate, the rotational speed and the intake temperature are decoupled by multi-scale features. The corresponding cross-domain correlation features are extracted by combining the component pre-detection data. At the same time, several similar segments are matched in the historical operating condition time series features according to the cross-domain correlation features to generate the corresponding multi-source decoupling feature vector.

[0045] The historical working condition time series features are clustered according to working condition, and combined with the working condition type to which the multi-source decoupling feature vector belongs, the appropriate inference sub-model is called up to output the corresponding initial component prediction results and initial working condition look-ahead level.

[0046] The initial component prediction results and the initial operating condition prospective level are corrected to generate the corresponding CNG component prediction results and the operating condition prospective level.

[0047] Furthermore, the acquisition module is specifically used for:

[0048] The initial component prediction deviation and the initial operating condition prospective level deviation are used as result variables. Simultaneously, the initial component prediction result and the initial operating condition prospective level are used as candidate cause variables. The core causal factors are located by the propensity score matching algorithm.

[0049] Based on the temporal variation pattern of the core causal factors, a corresponding factor influence intensity sequence is generated, and an MPC compensation model is constructed simultaneously based on the factor influence intensity sequence.

[0050] The engine hardware operating limit parameters are used as constraints of the MPC compensation model. Compensation coupling coefficients are introduced simultaneously. The initial component prediction results and the initial operating condition look-ahead level are corrected by the MPC compensation model to generate the CNG component prediction results and the operating condition look-ahead level accordingly.

[0051] Furthermore, the adjustment module is specifically used for:

[0052] Redundant features in the intake preheating benchmark, the preheating power and the eddy current adjustment parameters are removed by multi-source data fusion algorithm in order to extract the core feature set containing operating condition features, component fluctuation features and environmental impact features.

[0053] Based on the aforementioned working condition prospective level and the aforementioned core feature set, the engine working conditions are divided into different clusters, and rolling optimization factors adapted to each cluster are set synchronously.

[0054] The core feature set is transformed based on the rolling optimization factor to generate the target control dataset.

[0055] Furthermore, the adjustment module is specifically used for:

[0056] The core feature set is divided into a main feature layer, a secondary feature layer and an auxiliary feature layer. Simultaneously, the rolling optimization factor is decomposed by dimension to generate the corresponding time dimension factor, intensity dimension factor and compensation dimension factor.

[0057] The main feature layer is transformed to correspond to the time dimension factor, the secondary feature layer is transformed to correspond to the intensity dimension factor, and the auxiliary feature layer is transformed to correspond to the compensation dimension factor to generate the corresponding initial control dataset.

[0058] The initial control dataset is iteratively corrected to generate the target control dataset.

[0059] Furthermore, the adjustment module is specifically used for:

[0060] The core parameters of intake preheating, material flow regulation and fuel injection in the target control dataset are analyzed to construct the corresponding multi-parameter collaborative timing adjustment map;

[0061] Based on the multi-parameter coordinated timing adjustment spectrum, an air-fuel ratio adapted to CNG components is matched in a preset parameter database;

[0062] The engine's fuel injection parameters are adjusted according to the air-fuel ratio, and the engine's actual operating state is dynamically adjusted simultaneously based on the fuel injection parameters.

[0063] Furthermore, the adjustment module is specifically used for:

[0064] Based on the fuel injection parameters and the CNG component prediction results, corresponding injection target models are constructed according to engine cylinder groups;

[0065] With injection uniformity as the goal, the injection adjustment command for each group of cylinders is output through the injection target model, and the exhaust manifold temperature and exhaust composition of each group of cylinders are collected simultaneously to construct a corresponding temperature-composition dual-dimensional feedback verification mechanism.

[0066] The actual operating state of the engine is monitored and adjusted through the temperature-component dual-dimensional feedback verification mechanism.

[0067] The third aspect of the present invention proposes:

[0068] A computer includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the CNG engine exhaust temperature control method as described above.

[0069] The fourth aspect of the present invention proposes:

[0070] A readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the CNG engine exhaust temperature control method as described above.

[0071] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0072] Figure 1 A flowchart of the CNG engine exhaust temperature control method provided in the first embodiment of the present invention;

[0073] Figure 2 This is a structural block diagram of the CNG engine exhaust temperature control system provided in the third embodiment of the present invention.

[0074] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0075] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0076] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0077] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0078] Please see Figure 1 The figure shows a CNG engine exhaust temperature control method provided in the first embodiment of the present invention. The CNG engine exhaust temperature control method provided in this embodiment can continuously maintain the engine exhaust temperature within a reasonable range, thereby improving the temperature control efficiency.

[0079] Specifically, this embodiment provides:

[0080] A method for controlling the exhaust temperature of a CNG engine specifically includes the following steps:

[0081] Step S10: Based on the pre-detection data of components in the fuel supply pipeline and the historical operating condition time series characteristics of the engine, a corresponding operating condition change prediction model is constructed by combining the preset neural network.

[0082] It should be noted that, firstly, addressing the issue of large fluctuations in CNG fuel composition (methane, ethane, etc.) and the difficulty in accurately predicting exhaust temperature due to dynamic changes in engine operating conditions (idling, acceleration, full load), a predictive model for operating condition changes is constructed based on pre-detection data of fuel supply pipeline components (such as parameters like methane volume fraction and calorific value detected by gas chromatography) and historical operating condition time-series characteristics of the engine (such as exhaust temperature change curves under different components and loads, and throttle opening change patterns). This is combined with a pre-set neural network (such as a Long Short-Term Memory network, which excels at handling long-term dependencies in time-series data). Specifically, this model can uncover hidden correlations between "components-operating conditions-temperature," providing a model foundation for subsequent look-ahead control and avoiding the shortcomings of traditional control that is "based on real-time feedback and has strong lag."

[0083] Step S20: Collect the throttle opening change rate, turbocharger speed and intake air temperature, and input them into the operating condition change prediction model to output the corresponding CNG component prediction results and operating condition forecast level.

[0084] It should be noted that, secondly, three core real-time parameters are collected during engine operation: throttle opening change rate (reflecting the trend of engine load change; a large opening change rate corresponds to acceleration conditions), turbocharger speed (affecting intake pressure and exhaust energy recovery efficiency), and intake temperature (directly affecting combustion efficiency and exhaust temperature). These parameters are input into the operating condition change prediction model, which outputs two key results: CNG component prediction results (predicting the trend of fuel component changes at the next moment, such as methane content increasing / decreasing) and operating condition look-ahead level (classified according to the rate and intensity of load change, such as stable operating condition, low-speed dynamic operating condition, and high-speed dynamic operating condition). Specifically, the look-ahead level classification enables differentiated adaptation of control strategies, improving control accuracy.

[0085] Step S30: Determine the corresponding intake preheating benchmark based on the CNG component prediction result, and dynamically adjust the preheating power according to the operating condition forward-looking level, and dynamically correct the vortex adjustment parameters in combination with the actual exhaust temperature.

[0086] It should be noted that, next, the intake preheating benchmark is determined based on the CNG composition prediction results: different CNG compositions have different ignition points and combustion rates (e.g., CNG with high methane content has a high ignition point and requires a higher intake preheating temperature; CNG with high ethane content has a fast combustion rate and requires a lower preheating temperature to avoid detonation). The intake preheating benchmark can ensure that the fuel burns completely in the cylinder and reduce exhaust temperature fluctuations caused by incomplete combustion. Simultaneously, the preheating power is dynamically adjusted according to the operating condition forecast level. For example, the preheating power is increased under high-speed dynamic conditions to respond quickly to load changes, and the preheating power is reduced under stable conditions to save energy. At the same time, combined with the actual exhaust temperature collected by the exhaust temperature sensor, the swirl adjustment parameters are dynamically corrected (e.g., adjusting the opening of the swirl control valve to change the airflow intensity in the cylinder, thereby adjusting the combustion rate and temperature distribution). Specifically, this step achieves preliminary coordinated control of "composition-operating condition-temperature", laying the foundation for exhaust temperature stability.

[0087] Step S40: The intake preheating reference, the preheating power, and the vortex adjustment parameters are corrected by a rolling optimization algorithm to generate a corresponding target control dataset. The actual operating state of the engine is dynamically adjusted according to the target control dataset to keep the exhaust temperature in a stable range.

[0088] It should be noted that, finally, to address the issues of "not considering multi-factor coupling interference and insufficient parameter adaptability" in the initial control parameters, a rolling optimization algorithm (such as the rolling time-domain optimization strategy of Model Predictive Control (MPC)) is used to iteratively correct the intake preheating benchmark, preheating power, and vortex adjustment parameters. The rolling optimization algorithm can continuously collect new real-time data (such as actual exhaust temperature and turbine speed changes) within a finite time domain, update the optimization objective function (such as minimizing the deviation between exhaust temperature and target value, and minimizing preheating power energy consumption), and generate a target control dataset that balances temperature stability and energy consumption optimization. The target control dataset is then distributed to the engine's actuators (such as the preheater, vortex control valve, and fuel injector) to dynamically adjust the engine's actual operating state, ultimately stabilizing the exhaust temperature within a preset range (such as 450℃~650℃, to avoid damage to the turbocharger due to excessive temperature and excessive NOx emissions due to excessive temperature), forming a closed-loop control of "prediction-adjustment-optimization-execution".

[0089] Second Embodiment

[0090] Furthermore, the step of inputting the data into the operating condition change prediction model to output the corresponding CNG component prediction results and operating condition look-ahead level includes:

[0091] The opening change rate, the rotational speed and the intake temperature are decoupled by multi-scale features. The corresponding cross-domain correlation features are extracted by combining the component pre-detection data. At the same time, several similar segments are matched in the historical operating condition time series features according to the cross-domain correlation features to generate the corresponding multi-source decoupling feature vector.

[0092] The historical working condition time series features are clustered according to working condition, and combined with the working condition type to which the multi-source decoupling feature vector belongs, the appropriate inference sub-model is called up to output the corresponding initial component prediction results and initial working condition look-ahead level.

[0093] The initial component prediction results and the initial operating condition prospective level are corrected to generate the corresponding CNG component prediction results and the operating condition prospective level.

[0094] It should be noted that, firstly, for the multi-scale characteristics of three types of parameters—throttle opening change rate, turbine speed, and intake air temperature (e.g., the throttle opening change rate includes instantaneous abrupt changes and long-term trend characteristics)—multi-scale feature decoupling is performed: instantaneous features (e.g., peak values ​​of opening abrupt changes), short-term features (e.g., average change rate within 10 seconds), and long-term features (e.g., trends within 1 minute) of the parameters are extracted using algorithms such as wavelet decomposition; cross-domain correlation features are extracted by combining fuel component pre-detection data (e.g., the correlation feature of "increased intake air temperature + decreased methane content" corresponding to increased exhaust temperature); simultaneously, based on the cross-domain correlation features, several similar operating condition segments are matched in the historical operating condition time-series feature library (e.g., the current parameter combination has a parameter similarity of 90% with a certain historical acceleration operating condition); the decoupled multi-scale features are then fused with the similar operating condition features to generate a multi-source decoupled feature vector. Specifically, this vector contains both the dynamic characteristics of real-time parameters and incorporates empirical data from historical operating conditions, providing high-quality feature input for accurate prediction.

[0095] Secondly, the historical operating conditions of the engine are clustered: K-means and other clustering algorithms are used to divide the historical operating conditions into several clusters (such as the "high methane content + stable load" cluster and the "low methane content + acceleration load" cluster) according to the dimensions of load size, rate of change, and component type. Based on the feature distribution of the multi-source decoupling feature vector, the operating condition cluster to which it belongs is matched, and the inference sub-model corresponding to the cluster is retrieved (the parameters of the sub-models for different operating condition clusters have been trained and optimized through historical data). The initial component prediction result (such as predicting that the methane content will rise from 85% to 90% in the next moment) and the initial operating condition look-ahead level (such as judging it as a high-speed dynamic operating condition) are calculated and output through the inference sub-model. Specifically, the method of operating condition clustering and sub-model matching avoids the insufficient adaptability of a single model to all operating conditions and improves the accuracy of prediction.

[0096] Finally, considering that the initial prediction results may be affected by factors such as sensor noise and data matching deviation, the initial component prediction results and the initial operating condition look-ahead level are corrected: historical prediction error data (such as the deviation between the previous prediction value and the actual value) are introduced as the basis for correction to compensate for the initial results; at the same time, a secondary calibration is performed by combining current environmental parameters (such as atmospheric pressure and humidity; high humidity will reduce the oxygen content of the intake air and affect the combustion temperature), and finally, accurate CNG component prediction results and operating condition look-ahead level are generated. Specifically, the correction process can effectively reduce prediction errors and provide a reliable basis for the determination of subsequent control parameters.

[0097] Furthermore, the step of correcting the initial component prediction result and the initial operating condition look-ahead level to generate the corresponding CNG component prediction result and operating condition look-ahead level includes:

[0098] The initial component prediction deviation and the initial operating condition prospective level deviation are used as result variables. Simultaneously, the initial component prediction result and the initial operating condition prospective level are used as candidate cause variables. The core causal factors are located by the propensity score matching algorithm.

[0099] Based on the temporal variation pattern of the core causal factors, a corresponding factor influence intensity sequence is generated, and an MPC compensation model is constructed simultaneously based on the factor influence intensity sequence.

[0100] The engine hardware operating limit parameters are used as constraints of the MPC compensation model. Compensation coupling coefficients are introduced simultaneously. The initial component prediction results and the initial operating condition look-ahead level are corrected by the MPC compensation model to generate the CNG component prediction results and the operating condition look-ahead level accordingly.

[0101] It should be noted that, firstly, to pinpoint the core factors affecting the accuracy of the initial prediction results, the initial component prediction deviation (the difference between the initial prediction value and the historical actual value) and the initial working condition look-ahead level deviation (the difference between the initial level and the actual working condition) are used as outcome variables. The initial component prediction results, the initial working condition look-ahead level, and related influencing parameters (such as sensor measurement errors and historical working condition matching degree) are used as candidate causal variables. A propensity score matching algorithm is used for causal association analysis: propensity score matching can eliminate the interference of confounding variables and accurately identify the core causal factors (such as "component pre-detection data error" and "low matching degree of similar segments in historical working conditions" are the core factors causing prediction deviation). Specifically, the identification of the core factors provides a clear direction for subsequent targeted corrections.

[0102] Secondly, based on the temporal variation patterns of core causal factors (such as the daily variation patterns of component pre-detection data errors and the fluctuation trends of matching degree), a factor influence intensity sequence is generated (quantifying the degree of influence of core factors on the prediction results at different times; for example, when the matching degree is below 80%, the influence intensity reaches 0.8; when it is above 90%, the influence intensity drops to 0.2). An MPC compensation model is constructed based on the influence intensity sequence: the MPC compensation model has multivariate constraint optimization capabilities and can dynamically adjust the compensation strategy according to the changes in factor influence intensity, avoiding the limitations of traditional fixed compensation methods.

[0103] Finally, the engine hardware operating limits are used as constraints for the MPC compensation model (e.g., the upper limit of intake preheating temperature is 120℃ to avoid damage to the intake manifold due to excessive temperature; the upper limit of turbine speed is 15000r / min to prevent overspeed damage). At the same time, a compensation coupling coefficient is introduced (to quantify the coupling effect between component prediction deviation and operating condition look-ahead level deviation; for example, when the component deviation is large, the compensation amount of the operating condition level needs to be adjusted synchronously). The initial component prediction results and the initial operating condition look-ahead level are iteratively compensated and corrected by the MPC compensation model, ultimately generating CNG component prediction results and operating condition look-ahead levels with higher accuracy and closer fit to actual operating conditions. Specifically, the introduction of hardware constraints ensures that the corrected results will not exceed the safe operating range of the engine, improving the safety of control.

[0104] Furthermore, the step of correcting the intake preheating reference, the preheating power, and the eddy current adjustment parameters using a rolling optimization algorithm to generate the corresponding target control dataset includes:

[0105] Redundant features in the intake preheating benchmark, the preheating power and the eddy current adjustment parameters are removed by multi-source data fusion algorithm in order to extract the core feature set containing operating condition features, component fluctuation features and environmental impact features.

[0106] Based on the aforementioned working condition prospective level and the aforementioned core feature set, the engine working conditions are divided into different clusters, and rolling optimization factors adapted to each cluster are set synchronously.

[0107] The core feature set is transformed based on the rolling optimization factor to generate the target control dataset.

[0108] It should be noted that, firstly, for the redundant features of the three types of control parameters—intake preheating reference, preheating power, and swirl adjustment parameters (such as the repetitive correlation between preheating power and intake preheating reference, and the weak correlation between swirl parameters and ambient temperature), a multi-source data fusion algorithm (such as principal component analysis, PCA) is used for feature dimensionality reduction and redundancy removal. By calculating the correlation coefficients between parameters, features strongly correlated with exhaust temperature control (such as the correlation between preheating reference and components, and the correlation between swirl parameters and combustion rate) are retained. A core feature set containing operating condition features, component fluctuation features, and environmental impact features is extracted. Specifically, the construction of the core feature set can reduce the computational power consumption of subsequent optimization calculations and improve optimization efficiency.

[0109] Secondly, based on the classification criteria of the forward-looking operating conditions (such as stable operating conditions, low-speed dynamic operating conditions, and high-speed dynamic operating conditions), and combined with the feature distribution of the core feature set, the current and future operating conditions of the engine are divided into different clusters: for example, the high-speed dynamic operating condition cluster includes sub-clusters such as "acceleration load + high methane content" and "acceleration load + low methane content"; and appropriate rolling optimization factors are set for different clusters. The optimization factors include time-domain factors (optimization cycle length, such as 10 seconds for stable operating conditions and 2 seconds for high-speed dynamic operating conditions) and weight factors (weight allocation of temperature deviation and energy consumption, such as 0.8 for temperature deviation and 0.2 for energy consumption in high-speed operating conditions; the opposite is true for stable operating conditions). Specifically, the differentiated rolling optimization factors can achieve the optimization of control objectives under different operating conditions, taking into account both temperature stability and energy economy.

[0110] Finally, the core feature set is transformed based on the rolling optimization factor: the parameters in the core feature set are mapped to the objective function of rolling optimization, and the optimal solution of the objective function is solved through iterative calculation (such as solving the combination of preheating power and eddy current parameters that minimizes exhaust temperature deviation under high-speed dynamic conditions); the parameters corresponding to the optimal solution are integrated to generate a target control dataset. Specifically, this dataset contains the optimal control parameter combinations under different operating conditions, providing a basis for the precise control of the engine actuator.

[0111] Furthermore, the step of transforming the core feature set based on the rolling optimization factor to generate the target control dataset includes:

[0112] The core feature set is divided into a main feature layer, a secondary feature layer and an auxiliary feature layer. Simultaneously, the rolling optimization factor is decomposed by dimension to generate the corresponding time dimension factor, intensity dimension factor and compensation dimension factor.

[0113] The main feature layer is transformed to correspond to the time dimension factor, the secondary feature layer is transformed to correspond to the intensity dimension factor, and the auxiliary feature layer is transformed to correspond to the compensation dimension factor to generate the corresponding initial control dataset.

[0114] The initial control dataset is iteratively corrected to generate the target control dataset.

[0115] It should be noted that, firstly, based on the degree of influence of the core feature set on exhaust temperature control, it is divided into three levels: the main feature layer (including the operating condition look-ahead level and CNG component prediction results, which play a decisive role in temperature control), the secondary feature layer (including the intake preheating benchmark and preheating power, which are the core execution parameters of temperature control), and the auxiliary feature layer (including ambient temperature and atmospheric pressure, which are the interference compensation parameters for temperature control). Simultaneously, the rolling optimization factor is decomposed according to the dimension of action, generating three types of factors: time dimension factor (adjustment cycle and frequency of control parameters), intensity dimension factor (adjustment amplitude and rate of parameters), and compensation dimension factor (parameter compensation amount to cope with environmental interference). Specifically, the correspondence between feature layering and factor decomposition enables precise targeted optimization of parameters.

[0116] Secondly, a hierarchical mapping strategy is adopted to match and transform the core feature layer with the rolling optimization factors: the main feature layer corresponds to the time dimension factor transformation, such as the time dimension factor of 2 seconds for high-speed dynamic conditions, so the adjustment cycle of the main feature layer parameters is 2 seconds; the secondary feature layer corresponds to the intensity dimension factor transformation, such as the intensity dimension factor of 1.2 for high methane content conditions, so the preheating power adjustment range of the secondary feature layer is increased by 20%; the auxiliary feature layer corresponds to the compensation dimension factor transformation, such as the compensation dimension factor of 0.1 for high ambient humidity conditions, so the auxiliary feature layer parameters need to increase the compensation amount by 10%; through hierarchical transformation, an initial control dataset is generated. Specifically, this dataset achieves accurate matching between features at different levels and optimization factors, avoiding the blindness of parameter adjustment.

[0117] Finally, the initial control dataset is iteratively corrected: the initial control parameters are input into the engine dynamics simulation model (such as the GT-Power simulation model), the exhaust temperature change corresponding to the simulated parameters is simulated, and the deviation between the simulated value and the target value is calculated; based on the deviation, the parameters are adjusted using the gradient descent algorithm, and iterative optimization is performed until the deviation is less than a preset threshold (such as ±5℃); the iteratively optimized parameters are integrated to generate the final target control dataset. Specifically, iterative correction can effectively eliminate the deviation between simulation and actual operating conditions and improve the practicality of control parameters.

[0118] Furthermore, the step of dynamically adjusting the actual operating state of the engine based on the target control dataset to maintain the exhaust temperature within a stable range includes:

[0119] The core parameters of intake preheating, material flow regulation and fuel injection in the target control dataset are analyzed to construct the corresponding multi-parameter collaborative timing adjustment map;

[0120] Based on the multi-parameter coordinated timing adjustment spectrum, an air-fuel ratio adapted to CNG components is matched in a preset parameter database;

[0121] The engine's fuel injection parameters are adjusted according to the air-fuel ratio, and the engine's actual operating state is dynamically adjusted simultaneously based on the fuel injection parameters.

[0122] It should be noted that, firstly, three types of core parameters in the target control dataset are analyzed: intake preheating parameters (preheating temperature, preheating power), material flow regulation parameters (vortex control valve opening, intake pressure), and fuel injection parameters (injection pressure, injection timing). Based on the temporal correlation of these parameters (e.g., after the intake preheating temperature increases, the injection timing needs to be delayed to avoid detonation), a multi-parameter collaborative temporal adjustment map is constructed. Specifically, the map intuitively presents the adjustment sequence, magnitude, and timing of each parameter, ensuring coordinated adjustment of multiple parameters and avoiding mutual interference between parameters that could lead to exhaust temperature fluctuations.

[0123] Secondly, changes in CNG composition directly affect the theoretical air-fuel ratio (e.g., the theoretical air-fuel ratio for pure methane is 17.2:1, and for ethane it is 15.7:1). A deviation from the theoretical value in the air-fuel ratio leads to incomplete combustion, which in turn causes fluctuations in exhaust temperature. Based on the composition prediction results in the multi-parameter coordinated time-series adjustment spectrum, the optimal air-fuel ratio adapted to the current CNG composition is matched in a preset parameter database. Specifically, the preset parameter database stores the optimal air-fuel ratio data corresponding to different CNG compositions, such as matching an air-fuel ratio of 17.0:1 when methane content is 90% and 16.5:1 when methane content is 80%. Matching the optimal air-fuel ratio ensures complete fuel combustion and stabilizes exhaust temperature.

[0124] Finally, the engine's fuel injection parameters are adjusted according to the optimal air-fuel ratio: if the air-fuel ratio is too high, the injection pressure is increased to improve the fuel supply; if the air-fuel ratio is too low, the injection pressure is reduced or the injection timing is delayed. By adjusting the injection parameters, the fuel supply and mixture ratio in the cylinder are changed, thereby dynamically adjusting the engine's actual operating state (such as combustion rate and cylinder temperature distribution), ultimately achieving stable control of exhaust temperature. Specifically, the adjustment of fuel injection parameters is the core execution link of exhaust temperature control, which directly determines the stability of the combustion process.

[0125] Furthermore, the step of dynamically adjusting the actual operating state of the engine based on the fuel injection parameters includes:

[0126] Based on the fuel injection parameters and the CNG component prediction results, corresponding injection target models are constructed according to engine cylinder groups;

[0127] With injection uniformity as the goal, the injection adjustment command for each group of cylinders is output through the injection target model, and the exhaust manifold temperature and exhaust composition of each group of cylinders are collected simultaneously to construct a corresponding temperature-composition dual-dimensional feedback verification mechanism.

[0128] The actual operating state of the engine is monitored and adjusted through the temperature-component dual-dimensional feedback verification mechanism.

[0129] It should be noted that, firstly, considering the combustion differences between cylinders in multi-cylinder CNG engines (such as uneven intake volume and different injector wear, leading to differences in combustion efficiency), an injection target model is constructed based on fuel injection parameters and CNG component prediction results, according to engine cylinder groups (e.g., a 6-cylinder engine is divided into 3 groups, with 2 cylinders in each group). The injection target model for each group of cylinders is optimized based on its historical combustion data (such as exhaust temperature and combustion pressure) to ensure that the injection parameters of each group of cylinders are adapted to its combustion characteristics. Specifically, group modeling can realize differentiated injection control between cylinders and improve combustion uniformity.

[0130] Secondly, with the goal of ensuring uniform injection in each cylinder (e.g., the exhaust temperature deviation of each group of cylinders is less than 10℃), injection adjustment commands for each group of cylinders are output through the injection target model (e.g., the injection pressure of one group of cylinders is increased by 5%, and that of the second group by 3%). Simultaneously, the exhaust manifold temperature (directly reflecting the combustion temperature of a single cylinder) and exhaust components (e.g., CO and HC content, reflecting the completeness of combustion) of each group of cylinders are collected to construct a temperature-component dual-dimensional feedback verification mechanism. Specifically, the temperature dimension verifies whether the exhaust temperature is stable within the target range, and the component dimension verifies whether the combustion is complete. The dual-dimensional verification can avoid the problem of "temperature meets the standard but combustion is incomplete" caused by single temperature feedback.

[0131] Finally, the engine status is monitored in real time based on a temperature-component dual-dimensional feedback verification mechanism: if the exhaust manifold temperature of a certain group of cylinders is higher than the target value and the HC content exceeds the standard, it is determined that the combustion of that group of cylinders is incomplete, and the injection pressure needs to be reduced or the intake preheating temperature needs to be increased; if the temperature is lower than the target value and the CO content exceeds the standard, it is determined that the fuel supply is excessive, and the injection pressure needs to be increased or the injection timing needs to be delayed; by adjusting the injection parameters of each group of cylinders in real time, the actual working state of the engine is dynamically optimized, and finally, uniform and stable control of the exhaust temperature of all cylinders is achieved. Specifically, this feedback verification mechanism forms a refined control closed loop of "group control - dual-dimensional monitoring - real-time adjustment" to ensure the stability of the exhaust temperature and emission compliance of the engine under all operating conditions.

[0132] Please see Figure 2 The third embodiment of the present invention provides:

[0133] A CNG engine exhaust temperature control system, wherein the system comprises:

[0134] The module is used to construct a corresponding operating condition change prediction model based on the component pre-detection data of the fuel supply pipeline and the historical operating condition time series characteristics of the engine, combined with a preset neural network.

[0135] The data acquisition module is used to collect the throttle opening change rate, turbocharger speed and intake air temperature, and input them into the operating condition change prediction model to output the corresponding CNG component prediction results and operating condition forecast level.

[0136] The processing module is used to determine the corresponding intake preheating benchmark based on the CNG component prediction result, and simultaneously dynamically adjust the preheating power according to the operating condition look-ahead level, and dynamically correct the vortex adjustment parameters in combination with the actual exhaust temperature.

[0137] The adjustment module is used to correct the intake preheating reference, the preheating power and the vortex adjustment parameters through a rolling optimization algorithm to generate a corresponding target control dataset, and simultaneously dynamically adjust the actual operating state of the engine according to the target control dataset so that the exhaust temperature is kept in a stable range.

[0138] Furthermore, the acquisition module is specifically used for:

[0139] The opening change rate, the rotational speed and the intake temperature are decoupled by multi-scale features. The corresponding cross-domain correlation features are extracted by combining the component pre-detection data. At the same time, several similar segments are matched in the historical operating condition time series features according to the cross-domain correlation features to generate the corresponding multi-source decoupling feature vector.

[0140] The historical working condition time series features are clustered according to working condition, and combined with the working condition type to which the multi-source decoupling feature vector belongs, the appropriate inference sub-model is called up to output the corresponding initial component prediction results and initial working condition look-ahead level.

[0141] The initial component prediction results and the initial operating condition prospective level are corrected to generate the corresponding CNG component prediction results and the operating condition prospective level.

[0142] Furthermore, the acquisition module is specifically used for:

[0143] The initial component prediction deviation and the initial operating condition prospective level deviation are used as result variables. Simultaneously, the initial component prediction result and the initial operating condition prospective level are used as candidate cause variables. The core causal factors are located by the propensity score matching algorithm.

[0144] Based on the temporal variation pattern of the core causal factors, a corresponding factor influence intensity sequence is generated, and an MPC compensation model is constructed simultaneously based on the factor influence intensity sequence.

[0145] The engine hardware operating limit parameters are used as constraints of the MPC compensation model. Compensation coupling coefficients are introduced simultaneously. The initial component prediction results and the initial operating condition look-ahead level are corrected by the MPC compensation model to generate the CNG component prediction results and the operating condition look-ahead level accordingly.

[0146] Furthermore, the adjustment module is specifically used for:

[0147] Redundant features in the intake preheating benchmark, the preheating power and the eddy current adjustment parameters are removed by multi-source data fusion algorithm in order to extract the core feature set containing operating condition features, component fluctuation features and environmental impact features.

[0148] Based on the aforementioned working condition prospective level and the aforementioned core feature set, the engine working conditions are divided into different clusters, and rolling optimization factors adapted to each cluster are set synchronously.

[0149] The core feature set is transformed based on the rolling optimization factor to generate the target control dataset.

[0150] Furthermore, the adjustment module is specifically used for:

[0151] The core feature set is divided into a main feature layer, a secondary feature layer and an auxiliary feature layer. Simultaneously, the rolling optimization factor is decomposed by dimension to generate the corresponding time dimension factor, intensity dimension factor and compensation dimension factor.

[0152] The main feature layer is transformed to correspond to the time dimension factor, the secondary feature layer is transformed to correspond to the intensity dimension factor, and the auxiliary feature layer is transformed to correspond to the compensation dimension factor to generate the corresponding initial control dataset.

[0153] The initial control dataset is iteratively corrected to generate the target control dataset.

[0154] Furthermore, the adjustment module is specifically used for:

[0155] The core parameters of intake preheating, material flow regulation and fuel injection in the target control dataset are analyzed to construct the corresponding multi-parameter collaborative timing adjustment map;

[0156] Based on the multi-parameter coordinated timing adjustment spectrum, an air-fuel ratio adapted to CNG components is matched in a preset parameter database;

[0157] The engine's fuel injection parameters are adjusted according to the air-fuel ratio, and the engine's actual operating state is dynamically adjusted simultaneously based on the fuel injection parameters.

[0158] Furthermore, the adjustment module is specifically used for:

[0159] Based on the fuel injection parameters and the CNG component prediction results, corresponding injection target models are constructed according to engine cylinder groups;

[0160] With injection uniformity as the goal, the injection adjustment command for each group of cylinders is output through the injection target model, and the exhaust manifold temperature and exhaust composition of each group of cylinders are collected simultaneously to construct a corresponding temperature-composition dual-dimensional feedback verification mechanism.

[0161] The actual operating state of the engine is monitored and adjusted through the temperature-component dual-dimensional feedback verification mechanism.

[0162] The fourth embodiment of the present invention provides a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the CNG engine exhaust temperature control method as described above.

[0163] The fifth embodiment of the present invention provides a readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the CNG engine exhaust temperature control method as described above.

[0164] In summary, the CNG engine exhaust temperature control method and system provided in the above embodiments of the present invention can continuously maintain the engine exhaust temperature within a reasonable range, thereby improving the control efficiency of exhaust temperature.

[0165] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0166] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0167] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0168] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0169] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0170] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A method for controlling the exhaust temperature of a CNG engine, characterized in that, The method includes: Based on the component pre-detection data of the fuel supply pipeline and the historical operating condition time series characteristics of the engine, a corresponding operating condition change prediction model is constructed by combining a preset neural network. The operating condition change prediction model is used to discover the hidden correlation between components, operating conditions and temperature. The throttle opening change rate, turbocharger speed and intake air temperature are collected and input into the operating condition change prediction model to output the corresponding CNG component prediction results and operating condition forecast level. Based on the CNG component prediction results, the corresponding intake preheating benchmark is determined, and the preheating power is dynamically adjusted according to the operating condition forecast level. The vortex adjustment parameters are dynamically corrected in combination with the actual exhaust temperature. The intake preheating reference, the preheating power, and the vortex adjustment parameters are corrected by a rolling optimization algorithm to generate a corresponding target control dataset. The actual operating state of the engine is then dynamically adjusted according to the target control dataset to keep the exhaust temperature within a stable range.

2. The CNG engine exhaust temperature control method according to claim 1, characterized in that, The steps of inputting the data into the operating condition change prediction model to output the corresponding CNG component prediction results and operating condition look-ahead level include: The opening change rate, the rotational speed and the intake temperature are decoupled by multi-scale features. The corresponding cross-domain correlation features are extracted by combining the component pre-detection data. At the same time, several similar segments are matched in the historical operating condition time series features according to the cross-domain correlation features to generate the corresponding multi-source decoupling feature vector. The historical working condition time series features are clustered according to working condition, and combined with the working condition type to which the multi-source decoupling feature vector belongs, the appropriate inference sub-model is called up to output the corresponding initial component prediction results and initial working condition look-ahead level. The initial component prediction results and the initial operating condition prospective level are corrected to generate the corresponding CNG component prediction results and the operating condition prospective level.

3. The CNG engine exhaust temperature control method according to claim 2, characterized in that, The step of correcting the initial component prediction results and the initial operating condition prospective level to generate the corresponding CNG component prediction results and operating condition prospective level includes: The initial component prediction deviation and the initial operating condition prospective level deviation are used as result variables. Simultaneously, the initial component prediction result and the initial operating condition prospective level are used as candidate cause variables. The core causal factors are located by the propensity score matching algorithm. Based on the temporal variation pattern of the core causal factors, a corresponding factor influence intensity sequence is generated, and an MPC compensation model is constructed simultaneously based on the factor influence intensity sequence. The engine hardware operating limit parameters are used as constraints of the MPC compensation model. Compensation coupling coefficients are introduced simultaneously. The initial component prediction results and the initial operating condition look-ahead level are corrected by the MPC compensation model to generate the CNG component prediction results and the operating condition look-ahead level accordingly.

4. The CNG engine exhaust temperature control method according to claim 1, characterized in that, The step of correcting the intake preheating reference, the preheating power, and the eddy current adjustment parameters using a rolling optimization algorithm to generate the corresponding target control dataset includes: Redundant features in the intake preheating benchmark, the preheating power and the eddy current adjustment parameters are removed by multi-source data fusion algorithm in order to extract the core feature set containing operating condition features, component fluctuation features and environmental impact features. Based on the aforementioned working condition prospective level and the aforementioned core feature set, the engine working conditions are divided into different clusters, and rolling optimization factors adapted to each cluster are set synchronously. The core feature set is transformed based on the rolling optimization factor to generate the target control dataset.

5. The CNG engine exhaust temperature control method according to claim 4, characterized in that, The step of transforming the core feature set based on the rolling optimization factor to generate the target control dataset includes: The core feature set is divided into a main feature layer, a secondary feature layer and an auxiliary feature layer. Simultaneously, the rolling optimization factor is decomposed by dimension to generate the corresponding time dimension factor, intensity dimension factor and compensation dimension factor. The main feature layer is transformed to correspond to the time dimension factor, the secondary feature layer is transformed to correspond to the intensity dimension factor, and the auxiliary feature layer is transformed to correspond to the compensation dimension factor to generate the corresponding initial control dataset. The initial control dataset is iteratively corrected to generate the target control dataset.

6. The CNG engine exhaust temperature control method according to claim 1, characterized in that, The step of dynamically adjusting the actual operating state of the engine based on the target control dataset to keep the exhaust temperature within a stable range includes: The core parameters of intake preheating, vortex regulation and fuel injection in the target control dataset are analyzed to construct the corresponding multi-parameter collaborative timing adjustment map; Based on the multi-parameter coordinated timing adjustment spectrum, an air-fuel ratio adapted to CNG components is matched in a preset parameter database; The engine's fuel injection parameters are adjusted according to the air-fuel ratio, and the engine's actual operating state is dynamically adjusted simultaneously based on the fuel injection parameters.

7. The CNG engine exhaust temperature control method according to claim 6, characterized in that, The step of dynamically adjusting the actual operating state of the engine based on the fuel injection parameters includes: Based on the fuel injection parameters and the CNG component prediction results, corresponding injection target models are constructed according to engine cylinder groups; With injection uniformity as the goal, the injection adjustment command for each group of cylinders is output through the injection target model, and the exhaust manifold temperature and exhaust composition of each group of cylinders are collected simultaneously to construct a corresponding temperature-composition dual-dimensional feedback verification mechanism. The actual operating state of the engine is monitored and adjusted through the temperature-component dual-dimensional feedback verification mechanism.

8. A CNG engine exhaust temperature control system, characterized in that, The system includes: The module is used to construct a corresponding operating condition change prediction model based on the component pre-detection data of the fuel supply pipeline and the historical operating condition time series characteristics of the engine, combined with a preset neural network. The data acquisition module is used to collect the throttle opening change rate, turbocharger speed and intake air temperature, and input them into the operating condition change prediction model to output the corresponding CNG component prediction results and operating condition forecast level. The processing module is used to determine the corresponding intake preheating benchmark based on the CNG component prediction result, and simultaneously dynamically adjust the preheating power according to the operating condition look-ahead level, and dynamically correct the vortex adjustment parameters in combination with the actual exhaust temperature. The adjustment module is used to correct the intake preheating reference, the preheating power and the vortex adjustment parameters through a rolling optimization algorithm to generate a corresponding target control dataset, and simultaneously dynamically adjust the actual operating state of the engine according to the target control dataset so that the exhaust temperature is kept in a stable range.

9. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the CNG engine exhaust temperature control method as described in any one of claims 1 to 7.

10. A readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the CNG engine exhaust temperature control method as described in any one of claims 1 to 7.

Images