Intelligent starting control method for diesel engine set of diesel locomotive
By constructing a diesel unit start control model, dynamically calculate the fuel injection volume and perform self-learning and optimization, the carbon soot emission problem of diesel locomotives during low-temperature start-up is solved, and stable start-up and environmental performance improvement in different environments are achieved.
Patent Information
- Application Number
- CN202510822005.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-19
- Publication Date
- 2025-08-12
AI Technical Summary
Existing diesel locomotives are prone to excessive carbon soot emissions when starting in low temperature environments, and traditional control strategies are difficult to adapt to dynamic environmental variables, resulting in a contradiction between environmental performance and startup success rate.
Build a start control model of the diesel unit, obtain multiple operating environment parameters through sensors, dynamically calculate the optimal fuel injection volume, and adjust the fuel supply in real time based on the self-learning optimization model to optimize the combustion process and inhibit the generation of carbon soot.
Without changing the structure of the diesel unit, soot emission control is achieved under various environmental conditions, ensuring the reliability and environmental protection performance, and adapting to stable start-up under complex geographical and climatic conditions.
Smart Images
Figure CN120466094A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of diesel engine set control, and in particular to an intelligent starting control method for an internal combustion locomotive diesel engine set. Background Art
[0002] With the continued expansion of my country's railway network, the number of diesel locomotives reached 11,000 by 2020. As a vital component of the railway transportation system, traditional diesel locomotives, while providing flexible traction, are increasingly attracting attention for their pollutant emissions. In particular, in the area of non-road mobile source pollution control, the emission of black smoke from diesel locomotives has become a key factor hindering air quality improvements. However, the current definition of "black smoke" remains at the qualitative level, lacking a quantifiable technical evaluation system.
[0003] Black smoke from diesel engine exhaust is essentially incompletely burned soot particles, the formation mechanism of which is closely related to the physical and chemical processes within the combustion chamber. During the diffusion combustion phase, the local high temperature and oxygen deficiency trigger thermal cracking of the hydrocarbon fuel, generating 2-nanometer soot cores. These then undergo dehydrogenation and polymerization to form 20-30-nanometer soot elements, which ultimately aggregate to form submicron particles. This pollutant is particularly prominent during locomotive starting conditions. The root cause is the combined effects of multiple combustion constraints during cold starts: the low temperature prevents the compression end temperature from reaching the natural threshold of 350°C, reduced fuel atomization quality leads to an imbalanced mixture concentration distribution, and insufficient cylinder sealing causes compression pressure decay. These physical processes combine to reduce fuel combustion efficiency and exacerbate particulate matter formation.
[0004] To ensure starting reliability in low-temperature environments, existing diesel engine calibration strategies generally adopt a rich-fuel injection scheme. Although the minimum ignition energy requirement can be maintained by increasing the baseline fuel injection volume at the start time, the excessive fuel prolongs the ignition delay period under poor combustion conditions, which in turn exacerbates soot formation. This static parameter setting method is difficult to adapt to the influence of dynamic variables such as ambient temperature, oil viscosity, and accumulator pressure, resulting in an inherent contradiction between environmental performance and starting success rate. In particular, when the ambient temperature is below -10°C, the adaptability defects of conventional control strategies are more obvious, and the exhaust smoke value exceeds the visible pollutant control limit. Therefore, it is urgent to establish a dynamic control mechanism that can coordinate the key parameters of the combustion process in real time to reduce the emission of unconventional pollutants while ensuring starting performance. Summary of the Invention
[0005] Based on the current state of the art, the present invention aims to ensure effective control of soot emissions even under normal locomotive startup conditions through intelligent starting control of a locomotive's electronically controlled single-pump diesel generator set, without changing the locomotive or diesel generator set structure. Therefore, a method for intelligent starting control of a diesel locomotive diesel generator set is proposed. This method can optimize smoke emission control during locomotive startup under various environmental conditions, effectively reducing the "black smoke" phenomenon during diesel generator set startup.
[0006] The present invention adopts the following technical solutions to achieve the purpose:
[0007] An intelligent starting control method for a diesel locomotive diesel generator set comprises the following steps:
[0008] S1. Construct a diesel generator start-up control model. The input of the model is multiple operating environment parameters, and the output is the target smoke value K of the diesel generator. goat and starting fuel injection amount W f ;
[0009] S2, entering the diesel locomotive starting process, the locomotive control system of the diesel locomotive obtains multiple types of operating environment parameters corresponding to the starting moment through sensors and inputs them into the starting control model;
[0010] S3, the engine control model calls the corresponding planned injection amount W according to the current input of multiple operating environment parameters p and target smoke density K goat ; Set the target smoke value K goat Output, as the comparison basis for this startup process;
[0011] S4, the engine start control model is based on the currently acquired multiple operating environment parameters and the called planned injection amount W p , calculate the starting fuel injection amount W f And output; locomotive control system is based on the starting fuel injection amount W f To start the diesel generator set;
[0012] S5. After the start process is completed, the locomotive control system obtains the actual smoke value K of this start through the sensor. i , the actual smoke value K i and target smoke density K goat Make a comparison and, based on the comparison results, perform self-learning optimization of the starting control model.
[0013] Preferably, in step S1, the multiple operating environment parameters include the ambient temperature T e , oil temperature T o , cooling water temperature T w and atmospheric pressure P.
[0014] Specifically, in step S3, the startup control model is based on the ambient temperature T e , call the planned injection amount W corresponding to this starting process p .
[0015] Furthermore, in step S4, the engine start control model is based on the ambient temperature T corresponding to the engine start time. e , oil temperature T o , cooling water temperature T w and atmospheric pressure P, and the corresponding planned injection amount W p , calculate the starting fuel injection amount W f , as follows:
[0016] W f =W p +f(T e ,T o ,T w ,P)
[0017] Where, f(T e ,T o ,T w ,P) represents the ambient temperature T e , oil temperature T o , cooling water temperature T w and atmospheric pressure P as input variables, and the initial functional relationship of the correction function is obtained by fitting the test results of the diesel engine bench calibration test.
[0018] Preferably, in step S4, if the currently calculated starting fuel injection amount W f The starting requirements of the diesel generator set cannot be met, resulting in the failure of the diesel generator set to start. The starting control model is based on the current starting fuel injection amount W f , calculate the restart injection amount W r The locomotive control system is based on the restart injection amount W r to start the diesel generator set; if it still fails to start, repeat the above process, and cycle until the diesel generator set starts successfully.
[0019] Specifically, the starting control model is based on the current starting fuel injection amount W f , calculate the restart injection amount W r , as follows:
[0020] W r =W f +α
[0021] Where α is the preset experimental experience value.
[0022] Preferably, in step S5, if the actual smoke density value Ki Less than or equal to the target smoke value K goat , then the actual smoke density value K corresponding to this starting process i , Starting fuel injection amount W f 、Ambient temperature T e , oil temperature T o , cooling water temperature T w The atmospheric pressure P is stored and solidified into the starting control model to form the planned fuel injection amount W for the next starting process. p and target smoke density K goat Index relationship.
[0023] Preferably, in step S5, if the actual smoke density value K i Greater than the target smoke value K goat , then the actual smoke density value K corresponding to this starting process i , Starting fuel injection amount W f 、Ambient temperature T e , oil temperature T o , cooling water temperature T w and atmospheric pressure P are used to train the correction function f(T e ,T o ,T w ,P), so that the correction function applied in the next startup process can be optimized.
[0024] Specifically, a neural network algorithm is used to correct the function f(T e ,T o ,T w ,P) for training optimization, based on the actual smoke value K corresponding to this startup process i , Starting fuel injection amount W f 、Ambient temperature T e , oil temperature T o , cooling water temperature T w and atmospheric pressure P, the correction function f(T e ,T o ,T w , new functional relationship in P).
[0025] Specifically, a new functional relationship is obtained and the updated correction function f(T e ,T o ,T w ,P), the correction function is called by the startup control model in subsequent startup processes with similar multi-type operating environment parameters.
[0026] In summary, due to the adoption of this technical solution, the beneficial effects of the present invention are as follows:
[0027] Without changing the structure of the diesel engine itself, the method of the present invention establishes a multi-parameter coupled dynamic injection control model as a starting control model, thereby achieving coordinated optimization of soot emissions and starting reliability during the starting process. This method dynamically calculates the optimal starting fuel injection quantity based on real-time data such as ambient temperature, oil temperature, cooling water temperature and atmospheric pressure, breaking through the control limitations of traditional fixed calibration values. By matching the thermal state of the diesel engine with environmental conditions in real time, the fuel supply is precisely controlled to ensure that the mixture concentration in the combustion chamber is always in the high-efficiency combustion range, effectively suppressing the generation of soot particles. Especially under low-temperature starting conditions, the dynamic adjustment of the fuel supply by this method can reduce the smoke peak caused by excessive fuel while ensuring reliable ignition.
[0028] Based on the environmental parameters involved, this invention can address the impact of special operating conditions such as low pressure on plateaus and cold regions on the combustion process. By establishing a dynamic mapping relationship between various parameters and injection quantity, the fuel supply deviation problem of traditional control strategies with altitude changes is overcome. Furthermore, to address the impact of temperature-dependent changes in oil viscosity on the actuator's response characteristics, a corresponding design ensures the consistent operation of the electronically controlled unit pump under different operating conditions. This multi-dimensional parameter coordinated control mode enables the diesel engine set to maintain stable starting performance over a wide temperature range and under complex geographical conditions.
[0029] The advantage of the method of the present invention also lies in the construction of a self-learning architecture with continuous evolution capabilities. By collecting key operating data of each startup process, the parameter combination of successful startup is automatically stored and used to iteratively optimize the initial startup control model. With the continuous accumulation of startup data, the adaptability of this method to different environmental conditions and diesel engine status is continuously enhanced, forming a positive feedback regulation mechanism. By dynamically updating the injection quantity mapping table, the control model is always synchronized with the actual status of the diesel engine, thereby achieving an intelligent control effect that becomes more accurate with use. This method has low implementation cost and strong adaptability, providing a feasible technical path for the environmental protection upgrade of traditional internal combustion power units. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] The present invention further illustrates its implementation and technical solutions in detail through the following drawings, which specifically include two drawings as follows:
[0031] Figure 1 A schematic diagram briefly describing the overall process steps of the method of the present invention;
[0032] Figure 2 The figure is a schematic diagram of a specific implementation process of the method of the present invention. DETAILED DESCRIPTION
[0033] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.
[0034] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.
[0035] An intelligent starting control method for a diesel locomotive diesel generator set, Figure 1 The overall process of the method is briefly described in the following figure, which can be viewed simultaneously. The key steps of the method can be summarized as follows:
[0036] S1. Construct a diesel generator start-up control model. The input of the model is multiple operating environment parameters, and the output is the target smoke value K of the diesel generator. goat and starting fuel injection amount W f ;
[0037] S2, entering the diesel locomotive starting process, the locomotive control system of the diesel locomotive obtains multiple types of operating environment parameters corresponding to the starting moment through sensors and inputs them into the starting control model;
[0038] S3, the engine control model calls the corresponding planned injection amount W according to the current input of multiple operating environment parameters p and target smoke density K goat ; Set the target smoke value K goat Output, as the comparison basis for this startup process;
[0039] S4, the engine start control model is based on the currently acquired multiple operating environment parameters and the called planned injection amount W p , calculate the starting fuel injection amount W f And output; locomotive control system is based on the starting fuel injection amount W f To start the diesel generator set;
[0040] S5. After the start process is completed, the locomotive control system obtains the actual smoke value K of this start through the sensor. i , the actual smoke value K i and target smoke density K goat Make a comparison and, based on the comparison results, perform self-learning optimization of the starting control model.
[0041] This embodiment will provide a detailed introduction to the details of the method steps.
[0042] In step S1, the preferred multiple operating environment parameters of this embodiment include the ambient temperature T e , oil temperature T o , cooling water temperature T w and atmospheric pressure P. In the diesel generator set start-up control process, these four parameters are selected as the core input variables for modeling. The selection is based on the analysis and summary of diesel combustion mechanism and environmental adaptability.
[0043] Ambient temperature T e It directly affects the atomization quality of diesel and the efficiency of mixture formation. Low temperature environment will increase the viscosity of fuel and reduce its volatility, leading to combustion delay and incomplete combustion. o By affecting the lubrication state of the diesel engine's moving parts, it indirectly affects the starting performance. When the oil viscosity increases with decreasing temperature, the crankshaft's rotational resistance increases, which will lead to insufficient compression end temperature, thereby affecting the ignition conditions. w As a key indicator of the diesel engine's thermal state, its value is directly related to the combustion chamber wall temperature. Increased heat dissipation from the wall at low temperatures significantly reduces the combustion reaction rate. Atmospheric pressure (P) is used to adjust the actual charge volume by altering the intake air density. In environments such as those with low air pressure on high plateaus, a reduced air coefficient can easily lead to overly rich mixture combustion, resulting in a surge in soot generation.
[0044] The above parameters show dynamic coupling characteristics under different working conditions, and together constitute the basic physical field that affects the starting smoke density. Therefore, this embodiment can achieve adaptive compensation control of complex environmental combinations by establishing a mapping relationship between four-dimensional parameters and injection quantity and smoke density target values, thereby effectively suppressing abnormal emissions during the starting process.
[0045] The detailed process of this embodiment can be found in the attached Figure 2 's hint.
[0046] In step S2, before the diesel locomotive starts, the fuel, oil, and cooling water may be preheated and pre-supplied. Then, as a preferred embodiment of this embodiment, after the acquisition of multiple types of operating environment parameters in step S2 is completed and input into the starting control model, in step S3, the starting control model determines the ambient temperature T in the multiple types of operating environment parameters currently input. e , call the planned injection amount W corresponding to this starting process p . Ambient temperature T eAs a key factor affecting the cold start combustion characteristics of diesel engines, its value directly determines the fuel atomization quality and the efficiency of mixture formation. When the ambient temperature is low, the increased viscosity of diesel leads to increased injection resistance. At the same time, insufficient end-of-compression temperature in the cylinder will aggravate ignition delay. At this time, it is necessary to increase the basic injection amount to maintain the concentration threshold of the combustible mixture; conversely, in high temperature environments, the injection amount needs to be reduced to avoid over-rich combustion. In this embodiment, the ambient temperature T e As the planned injection amount W p A single benchmark parameter can not only establish the basic framework of environmental adaptability control through the simplest path, but also quickly predict the injection amount through the dominant influence mechanism of this parameter on the combustion system. The coupling effect of other operating parameters on the combustion process will be applied in subsequent calculation links. This strategy not only ensures the real-time responsiveness of the starting control, but also reserves dynamic adjustment space for the coordinated optimization of multiple parameters.
[0047] In step S4, the engine start control model is based on the ambient temperature T corresponding to the engine start time. e , oil temperature T o , cooling water temperature T w and atmospheric pressure P, and the corresponding planned injection amount W p , calculate the starting fuel injection amount W f , as follows:
[0048] W f =W p +f(T e ,T o ,T w ,P)
[0049] Where, f(T e ,T o ,T w ,P) represents the ambient temperature T e , oil temperature T o , cooling water temperature T w The correction function takes atmospheric pressure P as input variables. The initial functional relationship of the correction function is obtained by fitting the test results of the diesel engine bench calibration test. The diesel engine bench calibration test simulates the actual working conditions in a controllable experimental environment and systematically collects combustion characteristic data under different environmental parameter combinations. The ambient temperature T can be set on the test bench. e , oil temperature T o , cooling water temperature T w The typical working condition matrix is constructed by adjusting the temperature control system, air pressure simulation device and other equipment. Multiple start-up tests are carried out under each set of working conditions, and the start-up fuel injection volume W is recorded simultaneously. f and actual smoke density K iThe nonlinear relationship between parameters is fitted by the least square method or neural network algorithm, and finally a correction function f(T e ,T o ,T w ,P). This correction function reflects the sensitivity weight of environmental parameters to the optimal injection quantity under the coupling of multiple physical fields.
[0050] The correction function can be expressed in many different forms. In this embodiment, the correction function f(T e ,T o ,T w ,P) is specifically expressed as follows:
[0051]
[0052] Where ω1 to ω4 are the values corresponding to the ambient temperature T e , oil temperature T o , cooling water temperature T w and the calibration factor of atmospheric pressure P, T e0 、T o0 and P0 corresponds to the ambient temperature T e , oil temperature T o and the reference value of atmospheric pressure P. The correction function f(T e ,T o ,T w ,P) includes three typical correction mechanisms: The first term on the right side of the equation represents the ambient temperature T e For the linear compensation term of fuel atomization delay, when the actual temperature is lower than the reference value, the basic correction amount needs to be increased; the second term reflects the cooling water temperature T through the exponential decay term. w The nonlinear coupling effect on the thermal state of the combustion chamber is also introduced, and the oil temperature T o The difference is used as a viscosity correction factor; the third term uses an inverse square relationship to simulate the compensatory effect of changes in atmospheric pressure P on intake density, ensuring a stable air-fuel ratio in low-pressure environments on the plateau. The functional structure of this correction function ensures physical interpretability while enhancing the model's adaptability to complex operating conditions through nonlinear terms. Its specific parameter values are determined through least-squares optimization from bench testing.
[0053] In step S4 of this embodiment, if the currently calculated starting fuel injection amount W f The starting requirements of the diesel generator set cannot be met, resulting in the failure of the diesel generator set to start. The starting control model is based on the current starting fuel injection amount W f , calculate the restart injection amount W r The locomotive control system is based on the restart injection amount W rTo start the diesel generator set; if the start still fails, repeat the above process until the diesel generator set starts successfully. The starting control model is based on the current starting fuel injection amount W f , calculate the restart injection amount W r , as follows:
[0054] W r =W f +α
[0055] Where α is the preset experimental experience value. The situation here is that when the diesel engine fails to start due to insufficient injection volume during the first start, the starting control model will use the incremental compensation mechanism to restart the strategy process. This strategy is to use the original calculated starting injection volume W f On this basis, a fixed correction value α is added to form the restart injection amount W r , thereby achieving a dynamic adjustment path for successful startup with minimal disturbance. The preset experimental experience value α is determined based on the physical constraints of the diesel engine combustion system and the emission control boundary conditions. It is recommended that the following requirements be met in determining the value range:
[0056] First, it is necessary to ensure that sufficient fuel increments can be provided under extreme operating conditions such as low-temperature starting at -30°C to overcome the ignition delay caused by insufficient compression temperature; secondly, it is necessary to avoid the risk of combustion deterioration caused by excessive fuel injection and avoid the aggravation of soot formation due to local over-rich mixture; finally, it is necessary to match the minimum adjustable resolution of the electronically controlled unit pump to ensure that the correction step size is within the controllable range of the actuator. p When the fuel injection rate is increased by 8%-12%, a successful secondary start can be achieved in more than 95% of starting scenarios, while the increase in smoke density is kept within the allowable range. This method, through its open-loop compensation form, avoids the response delay caused by complex iterative calculations while gradually approaching the optimal injection rate for actual operating conditions through the cumulative effect of fixed step sizes.
[0057] Through the above steps, this implementation achieves a tiered, progressive optimization of fuel injection quantity during diesel engine startup. Its multi-dimensional compensation of environmental parameters ensures that the mixture concentration remains within the optimal combustion range under varying operating conditions. A physical constraint framework for the correction function, combined with a dynamic adjustment strategy based on empirical experimental data, effectively balances the conflict between startup success rate and soot generation. This method uses real-time feedback from actual smoke density values to iteratively update model parameters, enabling the control system to adapt to environmental conditions and self-correct for long-term operating condition drift. Ultimately, the goal of precise, "zero black smoke" startup control is achieved under complex environmental conditions without requiring mechanical modification.
[0058] As a preferred embodiment, the following describes the detailed process of how the startup control model in step S5 realizes self-learning optimization. Figure 2 's hint.
[0059] In step S5, if the actual smoke value K i Less than or equal to the target smoke value K goat , then the actual smoke density value K corresponding to this starting process i , Starting fuel injection amount W f 、Ambient temperature T e , oil temperature T o , cooling water temperature T w The atmospheric pressure P is stored and solidified into the starting control model to form the planned fuel injection amount W for the next starting process. p and target smoke density K goat Index relationship.
[0060] In this embodiment, the actual smoke value K i To meet the target smoke density value K goat Under the constraints of , the system stores the complete operating condition characteristic data of this start-up process as characteristic operating condition points in the database and solidifies them into the basic index of the table lookup method. This solidification process is achieved by establishing a mapping relationship between multi-dimensional parameter space and injection quantity-smoke density: under the typical operating condition dimension, the historical successful data is divided into intervals according to the environmental parameter range, and each interval stores the minimum injection quantity and its corresponding smoke density value verified to be effective under the operating condition. During subsequent starts, the control system matches the characteristic operating condition intervals in the database with the environmental parameters collected in real time, and directly calls the historical optimal injection quantity as the planned injection quantity W p , and set the corresponding smoke density value as the target smoke density value K goat This table-based solidification mechanism not only retains best practice data from bench tests and actual operation, but also expands the coverage of the control strategy through the continuous accumulation of characteristic operating points, ensuring that verified and reliable parameters are prioritized under similar environmental conditions, thereby quickly achieving low-smoke starts and avoiding emission fluctuations caused by repetitive trial-and-error adjustments to the injection amount.
[0061] In addition, in step S5, if the actual smoke value K i Greater than the target smoke value K goat , then the actual smoke density value K corresponding to this starting process i , Starting fuel injection amount W f 、Ambient temperature T e , oil temperature T o , cooling water temperature T w and atmospheric pressure P are used to train the correction function f(T e ,T o ,T w ,P), so that the correction function applied in the next startup process is optimized. This embodiment uses a neural network algorithm to optimize the correction function f(Te ,T o ,T w ,P) for training optimization, based on the actual smoke value K corresponding to this startup process i , Starting fuel injection amount W f 、Ambient temperature T e , oil temperature T o , cooling water temperature T w and atmospheric pressure P, the correction function f(T e ,T o ,T w ,P) in the new functional relationship; after storing the updated correction function, the correction function is called by the startup control model in subsequent startup processes with similar multi-type operating environment parameters.
[0062] In this embodiment, the actual smoke value K i Exceeding target smoke density value K goat When the system starts the neural network training mechanism to dynamically optimize the correction function, it is the main means for the starting control model to achieve self-learning optimization. f And the actual smoke value K i As the training sample input, the parameter weights of the correction function are adjusted through the back propagation algorithm. This embodiment provides a suitable neural network model to implement this process. In actual application, the remaining model structures can also be designed as needed. The neural network adopts a three-layer fully connected structure. The input layer contains 4 nodes corresponding to the four-dimensional environmental parameters. The hidden layer is set with 8-12 nodes and uses the ReLU activation function. The output layer gives the correction value through linear regression. During the training process, the output value of the historical correction function is used as the initial prediction value, and the actual smoke value K is used as the initial prediction value. i and target smoke density K goat The deviation is the gradient direction of the loss function, and the network parameters are iteratively updated through the Adam optimizer so that the correction function gradually approaches the optimal compensation relationship under the actual working conditions. The training process finds the actual smoke value K after each startup. i Exceeding target smoke density value K goat The self-learning mechanism enables the starting control model to continuously adapt to the long-term operating condition drift phenomena such as diesel engine performance degradation and fuel quality changes.
[0063] In a preferred embodiment of neural network training, a sliding window data augmentation strategy is employed to enhance model generalization. The preferred approach is to randomly select approximately 30 sets of historical operating condition data for each training session, along with the current sample, to form a mixed training set. Input parameters are then normalized to eliminate dimensionality differences. The loss function is designed using a weighted combination of Huber loss and the absolute value of smoke density deviation, ensuring sensitivity to samples with large errors while avoiding outlier interference. After training, the system uses cross-validation to evaluate the output accuracy of the starting control model after updating the correction function. Formal parameter updates are only performed when the error of the new correction function is lower than that of the original model. The updated correction function complements the table lookup method for data consolidation: the former captures the complex coupling relationships between parameters through nonlinear mapping, while the latter retains the deterministic control strategy for characteristic operating points. This combination enables the control system to maintain rapid response to sudden environmental changes while achieving a spiral improvement in model accuracy through long-term data accumulation.
Claims
1. An intelligent starting control method for a diesel locomotive diesel generator set, characterized in that: The steps include: S1. Construct a diesel generator start-up control model. The input of the model is multiple operating environment parameters, and the output is the target smoke value K of the diesel generator. goat and starting fuel injection amount W f ; S2, entering the diesel locomotive starting process, the locomotive control system of the diesel locomotive obtains multiple types of operating environment parameters corresponding to the starting moment through sensors and inputs them into the starting control model; S3, the engine control model calls the corresponding planned injection amount W according to the current input of multiple operating environment parameters p and target smoke density K goat ; The target smoke value K goat Output, as the comparison basis for this startup process; S4, the engine start control model is based on the currently acquired multiple operating environment parameters and the called planned injection amount W p , calculate the starting fuel injection amount W f And output; locomotive control system is based on the starting fuel injection amount W f To start the diesel generator set; S5. After the start process is completed, the locomotive control system obtains the actual smoke value K of this start through the sensor. i , the actual smoke value K i and target smoke density K goat Make a comparison and, based on the comparison results, perform self-learning optimization of the starting control model.
2. The intelligent starting control method according to claim 1, characterized in that: In step S1, multiple operating environment parameters include ambient temperature T e , oil temperature T o , cooling water temperature T w and atmospheric pressure P.
3. The intelligent starting control method according to claim 2, characterized in that: In step S3, the engine start control model is based on the ambient temperature T e , call the planned injection amount W corresponding to this starting process p .
4. The intelligent starting control method according to claim 2, characterized in that: In step S4, the engine start control model is based on the ambient temperature T corresponding to the engine start time. e , oil temperature T o , cooling water temperature T w and atmospheric pressure P, and the corresponding planned injection amount W p , calculate the starting fuel injection amount W f , as follows: W f =W p +f(T e ,T o ,T w ,P) Where, f(T e ,T o ,T w ,P) represents the ambient temperature T e , oil temperature T o , cooling water temperature T w and atmospheric pressure P as input variables, and the initial functional relationship of the correction function is obtained by fitting the test results of the diesel engine bench calibration test.
5. The intelligent starting control method according to claim 1, characterized in that: In step S4, if the current calculated starting fuel injection amount W f The starting requirements of the diesel generator set cannot be met, resulting in the failure of the diesel generator set to start. The starting control model is based on the current starting fuel injection amount W f , calculate the restart injection amount W r The locomotive control system is based on the restart injection amount W r To start the diesel generator set; If the startup still fails, repeat the above process until the diesel generator set is successfully started.
6. The intelligent starting control method according to claim 5, characterized in that: The starting control model is based on the current starting fuel injection amount W f , calculate the restart injection amount W r , as follows: W r =W f +a Where α is the preset experimental experience value.
7. The intelligent starting control method according to claim 2, characterized in that: In step S5, if the actual smoke value K i Less than or equal to the target smoke value K goat , then the actual smoke density value K corresponding to this starting process i , Starting fuel injection amount W f 、Ambient temperature T e , oil temperature T o , cooling water temperature T w The atmospheric pressure P is stored and solidified into the starting control model to form the planned fuel injection amount W for the next starting process. p and target smoke density K goat Index relationship.
8. The intelligent starting control method according to claim 4, characterized in that: In step S5, if the actual smoke value K i Greater than the target smoke value K goat , then the actual smoke density value K corresponding to this starting process i , Starting fuel injection amount W f 、Ambient temperature T e , oil temperature T o , cooling water temperature T w and atmospheric pressure P are used to train the correction function f(T e ,T o ,T w ,P), so that the correction function applied in the next startup process can be optimized.
9. The intelligent starting control method according to claim 8, characterized in that: The neural network algorithm is used to correct the function f(T e ,T o ,T w ,P) for training optimization, based on the actual smoke value K corresponding to this startup process i , Starting fuel injection amount W f 、Ambient temperature T e , oil temperature T o , cooling water temperature T w and atmospheric pressure P, the correction function f(T e ,T o ,T w , new functional relationship in P).
10. The intelligent starting control method according to claim 9, characterized in that: Derive a new functional relationship and store the updated correction function f(T e ,T o ,T w ,P), the correction function is called by the startup control model in subsequent startup processes with similar multi-type operating environment parameters.