Engine idle speed control method based on dynamic idle deviation compensation

Abstract

This invention discloses an engine idle speed control method based on dynamic idle speed deviation compensation, belonging to the field of engine control technology. The method comprises the following components: establishing an engine nonlinear model, setting an objective function, rolling optimization, and implementing control. By establishing an accurate engine nonlinear model, this invention comprehensively considers multiple subsystems of the engine, including the intake system, fuel injection system, combustion process, and power output. Combined with experimental data to fit parameters, it can accurately simulate the engine's operation under different working conditions. Simultaneously, by employing a rolling optimization strategy based on dynamic idle speed deviation compensation, it predicts the future state based on the current engine state at each sampling time and determines the optimal control sequence by calculating the minimum value of the objective function, continuously compensating for idle speed deviation. This control method significantly improves the accuracy of idle speed control, reduces speed fluctuations, and enhances the stability of engine operation.

Description

Technical Field

[0001] This invention relates to the field of engine control technology, specifically to an engine idle speed control method based on dynamic idle speed deviation compensation. Background Technology

[0002] With the rapid development of the automotive industry and the increasing demands of consumers for vehicle performance, the stability of engine idling speed has become one of the important indicators for measuring the advancement of engine control technology. Maintaining stable engine speed during idling is crucial for reducing emissions, improving fuel economy, and enhancing driving comfort. However, in actual operation, engines are affected by a variety of factors, including but not limited to load changes, ambient temperature fluctuations, altitude differences, and wear of internal engine components. These factors work together to cause deviations in engine idling speed, increasing the difficulty of control.

[0003] Traditional engine idling speed control methods mostly employ PID control or open-loop control strategies based on fixed models. While these methods can achieve basic stability of idling speed to a certain extent, they have significant limitations. First, PID control relies on precise mathematical models and pre-set parameters. Faced with the complex nonlinearity and time-varying nature of engine operation, its control effect is often greatly reduced, making it difficult to maintain optimal control performance under different operating conditions. Second, open-loop control strategies based on fixed models lack the ability to provide real-time feedback and adjustment to the actual operating state. Once the engine state or external conditions change, the control effect will significantly decrease.

[0004] Given the limitations of traditional engine idle speed control methods, it is particularly important to develop an engine idle speed control method based on dynamic idle speed deviation compensation. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide an engine idle speed control method based on dynamic idle speed deviation compensation. It can dynamically sense changes in engine operating status by establishing an accurate engine nonlinear model and combining real-time data acquisition and processing technology. It also uses advanced optimization algorithms to achieve rolling optimization control, which significantly improves the control accuracy of idle speed and effectively reduces speed fluctuations.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an engine idle speed control method based on dynamic idle speed deviation compensation, the specific steps of which are as follows:

[0007] Establish a nonlinear engine model: comprehensively consider multiple subsystems of the engine, including the intake system, fuel injection system, combustion process, and power output, and construct a mathematical model that can accurately characterize the dynamic characteristics of the engine. This model uses engine speed, intake volume, fuel injection volume, throttle opening, etc. as variables reflecting the engine's operating state, and sets the throttle control signal and fuel injection control signal as adjustable control variables. By deeply analyzing the physicochemical reactions in the engine's working process and combining experimentally collected data for parameter fitting, the various parameters in the model are determined, enabling the model to accurately simulate the engine's operation under different working conditions.

[0008] Define the objective function: Construct an objective function to evaluate the engine idle speed control effect. This objective function comprehensively considers the deviation between the actual idle speed and the target idle speed, as well as the changes in the throttle control signal and the fuel injection control signal. By adjusting the relevant weights, the control performance can be optimized. If a fast response is desired, the weight reflecting the change in control quantity can be appropriately reduced; if the smoothness of control is emphasized, the corresponding weight can be increased. This balances the relationship between speed deviation and change in control quantity under different requirements, achieving the ideal control effect.

[0009] Rolling optimization: At each sampling moment, based on the actual operating status information of the current engine, the established nonlinear model is used to predict the engine status for a future period of time. Within this predicted time range, the minimum value of the objective function is calculated to determine the optimal control sequence for a shorter future period. This control sequence includes the optimal values ​​of the throttle control signal and the fuel injection control signal corresponding to each control moment. In the process of solving the optimal control sequence, an optimization algorithm is used, while strictly following the various constraints of the engine system to ensure that the obtained control sequence is operable in the actual engine system.

[0010] Control implementation: The first control variable in the optimal control sequence obtained from the optimization calculation is applied to the engine system to adjust the throttle opening and fuel injection quantity. After entering the next sampling time, the above prediction, optimization and control steps are repeated. Based on the newly acquired engine state information, the calculation and control are re-performed to achieve continuous rolling optimization control of the engine idle speed. Through this continuous cyclical control method, the idle speed deviation is continuously compensated so that the engine idle speed can be stabilized near the set target speed.

[0011] Furthermore, before establishing the nonlinear engine model, a data preprocessing step is included. The collected engine sensor data is preprocessed using a wavelet transform-principal component analysis (PCA)-based method. First, wavelet transform is used to denoise the data, with a decomposition level of 3 layers and the db4 wavelet basis function selected to remove high-frequency noise and outliers. Then, PCA is used to extract features from the denoised data, mapping the high-dimensional data to a low-dimensional space, removing redundant information, and retaining the feature variables that significantly affect engine idle speed control, thereby improving the efficiency and accuracy of subsequent model establishment and control algorithm operation.

[0012] Furthermore, in the step of establishing the nonlinear engine model, an improved multi-layer adaptive fuzzy inference system is used to construct the model. Specifically, the system divides the engine's operating conditions into multiple fuzzy regions, each corresponding to different engine load, ambient temperature, and altitude conditions. Within each fuzzy region, an adaptive neural network algorithm is used to learn and train the input and output data. In addition to engine speed, intake air volume, fuel injection volume, and throttle opening, the input variables also include exhaust gas recirculation rate and intake air temperature correction coefficient. The exhaust gas recirculation rate is obtained in real time through the exhaust gas recirculation valve opening sensor, and the intake air temperature correction coefficient is calculated by comparing the intake air temperature sensor data with the standard temperature. The output variables are the predicted engine speed and the predicted change in intake air volume at the next moment. The model parameters are determined as follows: using a large amount of actual operating data, the membership function parameters and rule base of the fuzzy inference system are optimized through a genetic algorithm. In the genetic algorithm, the population size is set to 50, the crossover probability is 0.8, and the mutation probability is 0.05. After 500 generations of evolution, the optimal model parameters are obtained, thereby constructing a model that can accurately reflect the nonlinear dynamic characteristics of the engine.

[0013] Furthermore, in the step of setting the objective function, an objective function construction method based on entropy weight-grey relational analysis is adopted, and the objective function expression is: ,in Based on the objective function, For the index of the predicted time, To predict the length of the time domain, For the first Speed ​​deviation at any moment For the first The amount of change in the throttle control signal at any given time. For the first The amount of change in the fuel injection control signal at all times. For the first The deviation between the predicted exhaust temperature and the standard exhaust temperature at a given moment is used to measure the stability of engine combustion. , , , The weighting coefficients are determined as follows: First, engine operating data under different operating conditions are collected, including engine speed, throttle opening, fuel injection quantity, and exhaust temperature. The objective weight of each indicator is calculated using the entropy weight method. The entropy weight method calculates the information entropy of the indicator; the smaller the information entropy, the more information the indicator provides, and the greater its weight. Next, grey relational analysis is used to calculate the correlation between each indicator and the ideal target, obtaining the subjective weight. Finally, the fusion coefficient between the objective and subjective weights is determined using the analytic hierarchy process (AHP). The two are then weighted and fused to obtain the final weighting coefficients. , , , This allows the objective function to more comprehensively reflect the overall performance of engine idle speed control.

[0014] Furthermore, in the step of setting the objective function, engine wear factors are also considered. Wear degree indicators for key engine components are introduced. Through vibration signal analysis of the crankshaft and piston components, combined with convolutional neural network algorithms in deep learning, a wear degree prediction model is established. The predicted engine wear degree is used as an influencing factor in the objective function, and a wear compensation term is added to the objective function. ,in The overall objective function taking wear factors into account; This is an indicator of engine wear. The wear factor weighting coefficients are determined using an adaptive weighting adjustment algorithm based on a large amount of experimental data. This allows the control strategy to be adjusted reasonably at different stages of engine wear, extending engine life while ensuring idle speed control performance.

[0015] Furthermore, in the rolling optimization step, the impact of vehicle grid load on engine idle speed is considered. A correlation model between vehicle grid load and engine idle speed is established, the operating status of vehicle electrical equipment is monitored in real time, grid load demand is calculated, and a grid load balancing term is added to the objective function. ,in The objective function is defined by taking the grid load into account. For vehicle electrical grid load power; The power generated by the generator driven by the engine; The grid load weighting coefficient is determined by a hybrid genetic-simulated annealing algorithm. During the rolling optimization process, factors such as idle speed deviation, control quantity changes, and grid load balance are comprehensively considered to obtain a more reasonable optimal control sequence, ensuring that the engine maintains a stable idle speed while meeting the grid load requirements.

[0016] Furthermore, in the step of predicting the engine state in the future prediction time domain based on the engine state information at the current time using the established nonlinear model at each sampling moment, an improved particle swarm optimization-Monte Carlo simulation prediction algorithm is adopted. In the particle swarm optimization algorithm, adaptive inertia weights and dynamic learning factors are introduced. With the number of iterations The formula for change is: ,in The maximum inertia weight is set to 0.9. The minimum inertia weight is set to 0.4. The maximum number of iterations is set to 200, and the dynamic learning factor is... , Based on the dynamic adjustment of particle fitness values, Monte Carlo simulation is used to handle uncertainties in the model. By randomly generating a large number of model parameter samples, the engine operating state under different parameter combinations is simulated to obtain the probability distribution of the engine state in the prediction time domain. By combining particle swarm optimization algorithm with Monte Carlo simulation, the optimal prediction result can be quickly searched while considering model uncertainties, thereby improving the accuracy and reliability of engine state prediction.

[0017] Furthermore, in the step of obtaining the optimal control sequence for the future control time domain by solving for the minimum value of the objective function in the prediction time domain, a hybrid ant colony-tabu search optimization algorithm is adopted. In the ant colony algorithm, the pheromone update formula is as follows when ants are searching for paths: ,in for Time Path The concentration of pheromones on the surface; This is the pheromone evaporation coefficient, with a value of 0.1. Set the number of ants to 20; For the first Only ants on the path The pheromone increment left on the surface, the tabu search algorithm is used to avoid the ant colony algorithm getting stuck in local optima. A tabu table is set up to record recently searched paths, and these paths are prohibited from being searched again within a certain number of iterations. The global search capability of the ant colony algorithm and the local search capability of the tabu search algorithm are combined to quickly and accurately solve the minimum value of the objective function under the constraints of the engine system, and obtain the optimal control sequence, thereby improving the efficiency and accuracy of the control sequence solution.

[0018] Furthermore, after the implementation of the control step, a feedback correction step is also included, specifically: real-time acquisition of engine speed, intake air volume, and fuel injection volume data after actual engine operation, comparison with the data output by the prediction model, calculation of the prediction error, and online correction of the prediction model using an error correction model based on least squares support vector machines. The regression function of the least squares support vector machine is: ,in For predicting output; For Lagrange multipliers; For the kernel function, a radial basis kernel function is used; As the bias, the parameters of the regression function are solved by the least squares method. Based on the calculated prediction error, the parameters of the nonlinear model are adjusted to make the model more consistent with the actual engine operation and further improve the accuracy of idle speed control.

[0019] Compared with existing technologies, this engine idle speed control method based on dynamic idle speed deviation compensation has the following advantages:

[0020] I. This method establishes an accurate nonlinear engine model, comprehensively considering multiple subsystems such as the engine's intake system, fuel injection system, combustion process, and power output. By fitting parameters with experimental data, it can accurately simulate the engine's operation under different working conditions. At the same time, it adopts a rolling optimization strategy based on dynamic idle speed deviation compensation. At each sampling time, it predicts the future state based on the current engine state and determines the optimal control sequence by calculating the minimum value of the objective function, continuously compensating for idle speed deviation. This control method significantly improves the accuracy of idle speed control, enabling the engine idle speed to be stabilized near the set target speed, reducing speed fluctuations, and improving the stability of engine operation.

[0021] Second, this method incorporates wear indicators of key engine components when setting the objective function. By analyzing vibration signals from the crankshaft and piston components and combining them with deep learning algorithms, a wear prediction model is established, and the prediction results are used as a factor influencing the objective function. This design allows the control system to rationally adjust the control strategy at different wear stages, avoiding accelerated wear of engine components due to over-control or under-control, thereby effectively extending the engine's service life. Simultaneously, reduced engine wear extends maintenance intervals, correspondingly lowering maintenance costs and bringing significant economic benefits to users.

[0022] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0024] Figure 1 The flowchart shows the operation of the engine idle speed control method based on dynamic idle speed deviation compensation.

[0025] Figure 2 This is a schematic diagram illustrating the functional implementation of an engine idle speed control method based on dynamic idle speed deviation compensation. Detailed Implementation

[0026] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0027] Example 1

[0028] In the cold start scenario of urban commuter cars in the early morning, the ambient temperature is usually low. This leads to a significant amount of low-temperature interference noise mixed into the raw data collected by the engine sensors. Furthermore, the data contains multiple dimensions such as engine speed, intake air volume, fuel injection quantity, and throttle opening. High-dimensional data increases the complexity and computational cost of subsequent model building. Therefore, the collected engine sensor data is first preprocessed using a wavelet transform-principal component analysis (PCA)-based data preprocessing method. Wavelet transform can accurately identify and filter high-frequency interference noise in the sensor data under low-temperature conditions, preserving as much effective information reflecting the engine's true operating state as possible and avoiding misleading subsequent modeling and control decisions due to noise. The subsequent PCA extracts key features from the denoised high-dimensional data, mapping it to a low-dimensional space. This reduces data redundancy and computational load in subsequent data processing, providing high-quality, low-dimensional effective data support for building an accurate engine nonlinear model and ensuring the reliability and effectiveness of the model input data.

[0029] During the cold start phase of urban commuter cars, the engine load is initially unstable, and the low ambient temperature significantly affects intake efficiency and fuel atomization, resulting in a significant difference from normal driving conditions. A single model cannot accurately depict the engine's dynamic characteristics during this phase. Therefore, this study comprehensively considers multiple subsystems of the engine, including the intake system, fuel injection system, combustion process, and power output. An improved multi-layer adaptive fuzzy inference system is used to construct a nonlinear model. This system divides the engine's operating conditions into multiple fuzzy regions based on the actual cold start conditions. Each fuzzy region precisely corresponds to different initial engine loads, low ambient temperatures, and current altitude conditions, ensuring the model can adapt to diverse operating conditions during the cold start phase. Within the region, the adaptive neural network algorithm learns and trains on the input and output data, enabling it to autonomously optimize model parameters to fit the operating conditions of the region. In addition to conventional input variables such as engine speed, intake air volume, fuel injection quantity, and throttle opening, it also introduces exhaust gas recirculation rate and intake air temperature correction coefficient. This allows for a more comprehensive capture of the complex physicochemical reactions within the engine during cold starts, such as the impact of exhaust gas recirculation on combustion stability at low temperatures and the change in intake air density due to intake air temperature. The output variables are set as predicted values ​​for engine speed and intake air volume changes at the next moment, enabling early prediction of engine state changes and providing accurate model basis for subsequent control strategy formulation, effectively improving the accuracy of engine dynamic characteristic characterization during the cold start phase.

[0030] During cold starts, engine idle speed is prone to fluctuations, and frequent adjustments to the throttle and fuel injection signals may exacerbate component wear. Abnormal exhaust temperature also affects engine lifespan. Therefore, it is necessary to construct an objective function that balances speed control accuracy, control signal stability, exhaust temperature, and component wear. An objective function construction method based on entropy weighted grey relational analysis is adopted, and the objective function expression is as follows: ,in Based on the objective function, For the index of the predicted time, To predict the length of the time domain, For the first Speed ​​deviation at any moment For the first The amount of change in the throttle control signal at any given time. For the first The amount of change in the fuel injection control signal at all times. For the first The deviation between the predicted exhaust temperature and the standard exhaust temperature at each predicted time; , , , Using the weighting coefficients, this method objectively determines the weights of each evaluation index through the entropy weighting method, and combines grey relational analysis to quantify the correlation between each index and the control effect, ensuring the scientific and reasonable construction of the objective function. The objective function first comprehensively considers the deviation between the actual idle speed and the target idle speed, which can directly measure the core accuracy of idle speed control. The smaller the deviation, the closer the speed is to the ideal state, improving driving comfort and reducing fuel consumption. Second, it incorporates the changes in throttle control signal and fuel injection control signal, which can avoid frequent and large fluctuations in control signals, reduce mechanical wear of actuators (such as throttle body and fuel injector), and extend their service life. Simultaneously, it incorporates the k-th prediction time... The deviation between the measured exhaust temperature and the standard exhaust temperature prevents damage to the engine exhaust system caused by excessively high or low exhaust temperatures during cold starts, ensuring the normal operation of the exhaust system. Furthermore, engine lubrication is poor during cold starts, making critical components such as the crankshaft and pistons prone to additional wear. Therefore, a wear degree index for key engine components is introduced. By analyzing the vibration signals of the crankshaft and piston components and combining this with a convolutional neural network algorithm from deep learning, a wear degree prediction model is established. This model can extract wear features from the vibration signals and accurately predict the wear state of components. The predicted wear degree is used as an influencing factor in the objective function, and a wear compensation term is added to the objective function. ,in The overall objective function taking wear factors into account; This is an indicator of engine wear. The wear factor weighting coefficients are used to obtain the overall objective function after considering the wear factor. This function can control the engine speed while protecting the components, reduce the wear on key engine components during cold starts, and extend the overall service life of the engine.

[0031] After a cold start, the engine state of a city commuter car changes rapidly over time, and the external power grid load (such as when the heater or lights are turned on) may suddenly increase. Therefore, the control strategy needs to be adjusted in real time to cope with these dynamic changes. At each sampling moment, based on the current actual engine operating status information (such as real-time speed, intake air volume, and fuel injection quantity), an established nonlinear model is used to predict the engine state over a future period. The prediction process employs an improved particle swarm optimization-Monte Carlo simulation prediction algorithm, where the particle swarm optimization algorithm introduces adaptive inertia weights and a dynamic learning factor. With the number of iterations The formula for change is: ,in The maximum inertia weight; The minimum inertia weight; To maximize the number of iterations, it can autonomously adjust its search capabilities based on the iteration process, avoiding getting trapped in local optima. Combined with Monte Carlo simulation, it can effectively handle uncertainties in the cold start phase (such as the cumulative effect of small sensor errors in low-temperature environments), significantly improving the accuracy of state prediction. Within the predicted time range, the optimal control sequence for a shorter future time period is determined by solving for the minimum value of the objective function. The solution process uses a hybrid ant colony-tabu search optimization algorithm, and the pheromone update formula is: ,in for Time Path The concentration of pheromones on the surface; The pheromone evaporation coefficient; The number of ants; For the first Only ants on the path The pheromone increment left on the engine is used to determine the optimal control scheme. Ant colony optimization has strong global search capabilities and can quickly traverse the feasible solution space. Tabu search avoids repeatedly searching for poor solutions that have already been explored. The combination of the two can efficiently find the optimal control scheme. At the same time, it strictly follows the various constraints during engine cold start (such as maximum fuel injection limit, throttle opening range, minimum speed protection, etc.) to prevent the control quantity from exceeding the safe range and causing engine failure. In addition, turning on electrical equipment such as heater and lights during cold start will increase the load on the vehicle's electrical grid. If the engine's power generation cannot match the load demand, it may lead to battery depletion. Therefore, a correlation model between vehicle electrical grid load and engine idle speed is established to monitor the working status of vehicle electrical equipment in real time, accurately calculate the electrical grid load demand, and add an electrical grid load balancing term to the objective function to obtain an objective function that takes the electrical grid load into account. This ensures that the power generated by the engine driving the generator matches the power grid load, guarantees the stable operation of the vehicle's electrical system, and avoids battery depletion.

[0032] During the cold start phase, the engine needs to respond quickly to control commands to stabilize the idle speed. Therefore, the first control variable in the optimal control sequence obtained from rolling optimization is promptly applied to the engine system to precisely adjust the throttle opening and fuel injection quantity. Adjusting the throttle opening changes the intake air volume, thereby affecting the air-fuel mixture concentration and combustion power in the engine cylinder, achieving macroscopic control of the engine speed. Adjusting the fuel injection quantity precisely controls the air-fuel ratio of the mixture, ensuring complete combustion and reducing pollutant emissions while stabilizing the engine speed. Upon entering the next sampling moment, since the engine state has changed (e.g., the engine speed has been adjusted, or the external load may have changed), the above prediction, optimization, and control steps need to be repeated. Based on the newly acquired engine state information, the optimal control variable is recalculated, forming a closed-loop control process of "prediction-optimization-control-re-prediction". This ensures that the control strategy at each moment can adapt to the current operating conditions, continuously stabilizing the engine idle speed within the target range, effectively solving the problem of large speed fluctuations during the cold start phase, and improving the engine's operational stability.

[0033] Although control strategies are formulated through model prediction and optimization, factors such as model errors and external disturbances (e.g., sudden temperature changes, slight sensor drift) may cause deviations between predicted and actual data during actual operation. Therefore, feedback correction is required after implementing control steps. Real-time data on engine speed, intake air volume, and fuel injection volume are collected and compared with the data output by the prediction model to accurately calculate the prediction error. The magnitude of the error directly reflects the degree of fit between the model and actual operating conditions. An error correction model based on least squares support vector machines is adopted. This model constructs a regression function through Lagrange multipliers, kernel functions, and biases, possessing strong nonlinear fitting and generalization capabilities. It can effectively analyze the variation law of prediction error and perform online correction of the prediction model. By correcting the model parameters in real time, the prediction error can be continuously reduced, ensuring that the prediction model always fits the actual operating state of the engine during the cold start phase. This avoids the failure of the control strategy due to model deviation, further improving the accuracy of subsequent control decisions, forming a complete closed-loop control system, and ensuring the continuous and stable engine idle speed control effect during the cold start phase.

[0034] Example 2

[0035] The low air pressure and thin air in high-altitude areas, coupled with complex electromagnetic interference and bumpy road conditions, result in a significant amount of noise mixed into the data collected by the sensors of freight truck engines (such as signal fluctuations caused by electromagnetic interference and momentary sensor misreadings due to bumps). Furthermore, the data contains multiple dimensions, including engine speed, intake air volume, fuel injection volume, throttle opening, and intake air temperature. This high-dimensional data increases the complexity of subsequent model calculations and may introduce redundant information. Therefore, the collected sensor data is first preprocessed using a wavelet transform-principal component analysis (PCA)-based data preprocessing method. Wavelet transform, through multi-scale decomposition, accurately separates the effective signal from the noise components in the data, effectively filtering out noise caused by high-altitude electromagnetic interference and bumps, preserving as much as possible the key information reflecting the engine's true operating state, and avoiding bias in subsequent model judgments due to noisy data. The subsequent PCA extracts core features from the denoised high-dimensional data, mapping the high-dimensional data to a low-dimensional space. This reduces data dimensionality while eliminating redundant information, decreasing the workload of subsequent model construction and calculations. This provides high-quality, low-redundancy input data for establishing an accurate engine nonlinear model, ensuring the efficiency and accuracy of subsequent modeling.

[0036] The low-pressure, low-oxygen environment of high-altitude regions significantly alters the intake efficiency, fuel atomization, and combustion process of freight truck engines. Furthermore, trucks frequently experience varying loads, such as empty or fully loaded. A single model cannot adapt to such complex operating conditions. Therefore, this study comprehensively considers the engine's intake system (the impact of low oxygen on intake air density), fuel injection system (the change in injection pressure due to low pressure), combustion process (changes in combustion rate caused by oxygen deficiency), and power output subsystem. An improved multi-layer adaptive fuzzy inference system is used to construct a nonlinear model. This system divides the engine's operating state into multiple fuzzy regions based on actual high-altitude conditions. Each region precisely corresponds to different engine loads (e.g., empty transport, fully loaded cargo), high-altitude ambient temperatures (large diurnal temperature variations), and different altitudes (e.g., 3000 meters, 5000 meters), ensuring the model's responsiveness. It adapts to the diverse operating conditions of high-altitude environments. Within each fuzzy region, an adaptive neural network algorithm is used to learn and train the input and output data, allowing the model parameters to be autonomously adjusted to fit the operating characteristics of that region, thus improving the model's fitting accuracy under specific conditions. In addition to conventional input variables such as engine speed, intake air volume, fuel injection volume, and throttle opening, it also introduces exhaust gas recirculation rate (the impact of exhaust gas recirculation on combustion stability under high-altitude hypoxia) and intake air temperature correction coefficient (the change in intake air density due to diurnal temperature variation at high altitudes), enabling a more comprehensive capture of the complex physicochemical reactions inside the engine under high-altitude conditions. The output variables are set as the predicted values ​​of engine speed and intake air volume change at the next moment, which can predict the engine state trend in advance and provide accurate model support for subsequent control strategy formulation, effectively solving the problem of inaccurate characterization of engine dynamic characteristics under high-altitude conditions.

[0037] Idle speed control for freight trucks in high-altitude areas needs to balance speed stability, control signal smoothness, exhaust temperature safety, and component wear protection. Therefore, an objective function construction method based on entropy weighting and grey relational analysis is adopted. This method objectively assigns weights to each indicator through entropy weighting and combines grey relational analysis to quantify the correlation between indicators and control effects, ensuring the scientific and rational nature of the objective function. The objective function first comprehensively considers the deviation between the actual idle speed and the target idle speed. The smaller the deviation, the more precise the speed control, which can avoid excessively high idle speed leading to increased fuel consumption and excessively low idle speed leading to insufficient power (such as inability to drive the air conditioning and hydraulic system), ensuring the normal operation of the truck's auxiliary systems. Secondly, it incorporates the changes in throttle control signal and fuel injection control signal, which can prevent frequent and large fluctuations in control signals, reduce mechanical wear of actuators such as throttle body and fuel injectors, and extend their service life in the harsh environment of high altitudes. By incorporating the deviation between the exhaust temperature at the k-th predicted moment and the standard exhaust temperature, abnormal exhaust temperatures caused by incomplete combustion due to high-altitude hypoxia (excessive heat damages the three-way catalytic converter, while excessive heat leads to carbon buildup) can be avoided, ensuring the safety of the exhaust system. In addition, truck loads change frequently in high-altitude areas (such as loading and unloading cargo), and critical components such as crankshafts and pistons bear greater loads, which can easily accelerate wear. Therefore, a wear degree index for key engine components is introduced. By collecting vibration signals from the crankshaft and pistons and combining them with convolutional neural network algorithms in deep learning, a wear degree prediction model is established. This model can extract wear features from the vibration signals (such as abnormal vibration frequencies corresponding to component wear) and accurately predict the wear state of components. By using the wear degree as an influencing factor of the objective function and adding a wear compensation term to obtain the overall objective function, the wear rate of components can be reduced while controlling the engine speed, thus extending the overhaul cycle of the engine in high-altitude environments.

[0038] In high-altitude areas, the engine state of freight trucks is affected by factors such as altitude changes, load fluctuations (e.g., temporary acceleration / deceleration), and the start / stop of electrical equipment (e.g., vehicle air conditioning, navigation). Therefore, control strategies need to be adjusted in real time. At each sampling moment, based on the current actual engine operating status information (e.g., real-time speed, intake pressure, fuel injection quantity), an established nonlinear model is used to predict the engine state over a future period. An improved particle swarm optimization-Monte Carlo simulation prediction algorithm is employed. The particle swarm optimization algorithm introduces adaptive inertia weights and dynamic learning factors, which can adjust the search capability according to the iteration process, avoiding getting trapped in local optima. Combined with Monte Carlo simulation, it can effectively handle the uncertainties of high-altitude operating conditions (e.g., sudden altitude changes, accumulation of small sensor errors), improving the accuracy of state prediction. Within the prediction time domain, the optimal control sequence is determined by solving for the minimum objective function. The solution process uses a hybrid ant colony approach. - The tabu search optimization algorithm, combined with the ant colony algorithm, possesses strong global search capabilities, enabling rapid traversal of the feasible solution space. The tabu search algorithm avoids repeatedly searching for inferior solutions. The combination of the two can efficiently find the optimal control scheme while strictly adhering to engine constraints (such as maximum fuel injection pressure, throttle opening limit, and minimum idle speed) to prevent control variables from exceeding safe ranges, which could lead to engine stalling or malfunction. In addition, trucks in high-altitude areas have large electrical loads (such as high-power air conditioners and onboard refrigerators). If the generator power cannot match the load demand, it may lead to battery depletion (low temperatures at high altitudes reduce battery capacity). Therefore, a correlation model between vehicle grid load and engine idle speed is established to monitor the operating status of electrical equipment in real time and calculate grid load power. A grid load balancing term is added to the objective function to obtain an objective function that considers grid load, ensuring that the power generated by the engine driving the generator matches the grid load and avoiding starting difficulties caused by battery depletion.

[0039] The first control variable in the optimal control sequence obtained through rolling optimization is promptly applied to the engine system. By adjusting the throttle opening to change the intake air volume (precise control of intake air volume is crucial for combustion stability in high-altitude, low-oxygen environments), the engine power is adjusted. The fuel injection quantity is adjusted to optimize the air-fuel ratio, addressing the incomplete combustion problem caused by high-altitude hypoxia and achieving precise idle speed control. Upon entering the next sampling moment, since the engine state has changed (e.g., increased altitude leading to reduced intake air volume, increased load leading to decreased speed), the prediction, optimization, and control steps must be repeated based on newly acquired state information (e.g., updated speed and intake air temperature) to form a closed-loop control process. This ensures that the control strategy at each moment is adapted to the current high-altitude operating conditions, effectively solving the problem of large idle speed fluctuations in high-altitude environments and guaranteeing the truck's idle stability and the normal operation of auxiliary systems.

[0040] In high-altitude areas, factors such as model errors (e.g., sudden altitude changes causing model adaptation lag) and external disturbances (e.g., strong winds affecting air intake) can lead to deviations between predicted and actual data. Therefore, feedback correction is necessary after control implementation. Real-time data on engine speed, air intake, and fuel injection are collected and compared with the output data of the prediction model to calculate the prediction error. The magnitude of the error directly reflects the fit between the model and actual operating conditions. An error correction model based on least squares support vector machines is adopted. This model constructs a regression function through Lagrange multipliers, kernel functions, and biases, possessing strong nonlinear fitting and generalization capabilities. It can analyze error variation patterns (e.g., lower predicted speed due to altitude increase) and perform online correction of the prediction model. By correcting model parameters in real time, the prediction error can be continuously reduced, ensuring the model always adapts to dynamic high-altitude conditions and preventing control failure due to model deviations (e.g., idling speed runaway), further improving the stability and reliability of idling control.

[0041] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method of engine idle speed control based on dynamic idle deviation compensation, characterized by, The specific steps of this method are as follows: Establish a nonlinear engine model: comprehensively consider multiple subsystems of the engine, including the intake system, fuel injection system, combustion process, and power output, and construct a mathematical model that can accurately characterize the dynamic characteristics of the engine. Through in-depth analysis of the physicochemical reactions in the engine working process, parameter fitting is performed in combination with experimentally collected data. Define the objective function: Construct an objective function to evaluate the engine idle speed control effect. This objective function takes into account the deviation between the actual idle speed and the target idle speed, as well as the changes in the throttle control signal and the fuel injection control signal. By adjusting the relevant weights, the control performance is optimized. Rolling optimization: At each sampling time, based on the actual operating status information of the engine, the established nonlinear model is used to predict the engine status in the future. Within this predicted time range, the minimum value of the objective function is calculated to determine the optimal control sequence in the shorter future time period. In the process of solving the optimal control sequence, an optimization algorithm is used, while strictly following the various constraints of the engine system. Implement control: Apply the first control variable in the optimal control sequence obtained from the optimization calculation to the engine system to adjust the throttle opening and fuel injection quantity. After entering the next sampling time, repeat the above prediction, optimization and control steps, and recalculate and control based on the newly acquired engine state information.

2. The engine idle speed control method based on dynamic idle deviation compensation according to claim 1, characterized in that, Before establishing the nonlinear model of the engine, a data preprocessing step is also included. The collected engine sensor data is preprocessed using a data preprocessing method based on wavelet transform-principal component analysis. First, wavelet transform is used to denoise the data. Then, principal component analysis is used to extract features from the denoised data and map the high-dimensional data to a low-dimensional space.

3. The engine idle speed control method based on dynamic idle deviation compensation according to claim 1, characterized in that, In the step of establishing the nonlinear engine model, an improved multi-layer adaptive fuzzy inference system is used to construct the model. Specifically, the system divides the engine's operating conditions into multiple fuzzy regions. Each fuzzy region corresponds to different engine load, ambient temperature, and altitude conditions. Within each fuzzy region, the input and output data are trained using an adaptive neural network algorithm. In addition to engine speed, intake air volume, fuel injection volume, and throttle opening, the input variables also include exhaust gas recirculation rate and intake air temperature correction coefficient. The output variables are the predicted engine speed and the predicted change in intake air volume at the next moment.

4. The engine idle speed control method based on dynamic idle speed deviation compensation according to claim 1, characterized in that, In the step of setting the objective function, an objective function construction method based on entropy weight-grey relational analysis is adopted, and the objective function expression is: ,in Based on the objective function, For the index of the predicted time, To predict the length of the time domain, For the first Speed ​​deviation at any moment For the first The amount of change in the throttle control signal at any given time. For the first The amount of change in the fuel injection control signal at all times. For the first The deviation between the predicted exhaust temperature and the standard exhaust temperature at each predicted time; , , , These are the weighting coefficients.

5. The engine idle speed control method based on dynamic idle speed deviation compensation according to claim 1, characterized in that, In the step of setting the objective function, a wear degree index for key engine components is introduced. By analyzing the vibration signals of the crankshaft and piston components and combining them with a convolutional neural network algorithm in deep learning, a wear degree prediction model is established. The predicted engine wear degree is used as an influencing factor in the objective function, and a wear compensation term is added to the objective function. ,in The overall objective function taking wear factors into account; The basic objective function; This refers to the engine wear index value; This represents the weighting coefficient for wear and tear factors.

6. The engine idle speed control method based on dynamic idle speed deviation compensation according to claim 1, characterized in that, In the rolling optimization step, a correlation model between vehicle grid load and engine idle speed is established, the operating status of vehicle electrical equipment is monitored in real time, grid load demand is calculated, and a grid load balancing term is added to the objective function. ,in The objective function is defined by taking the grid load into account. The basic objective function; For vehicle electrical grid load power; The power generated by the generator driven by the engine; This is the power grid load weighting coefficient.

7. The engine idle speed control method based on dynamic idle speed deviation compensation according to claim 1, characterized in that, In the step of predicting the engine state in the future time domain based on the engine state information at each sampling time using the established nonlinear model, an improved particle swarm optimization-Monte Carlo simulation prediction algorithm is adopted. In the particle swarm optimization algorithm, adaptive inertia weights and dynamic learning factors are introduced. With the number of iterations The formula for change is: ,in The maximum inertia weight; The minimum inertia weight; This represents the maximum number of iterations.

8. The engine idle speed control method based on dynamic idle speed deviation compensation according to claim 1, characterized in that, In the step of obtaining the optimal control sequence for the future control time domain by solving for the minimum value of the objective function within the prediction time domain, a hybrid ant colony-tabu search optimization algorithm is adopted. In the ant colony algorithm, the pheromone update formula is as follows when ants are searching for paths: ,in for Time Path The concentration of pheromones on the surface; The pheromone evaporation coefficient; The number of ants; For the first Only ants on the path The increase in pheromones left on the surface.

9. The engine idle speed control method based on dynamic idle speed deviation compensation according to claim 1, characterized in that, Following the implementation of the control steps, a feedback correction step is also included, specifically: real-time acquisition of engine speed, intake air volume, and fuel injection volume data after actual engine operation, comparison with the data output by the prediction model, calculation of the prediction error, and online correction of the prediction model using an error correction model based on least squares support vector machines. The regression function of the least squares support vector machine is: ,in For predicting output; For Lagrange multipliers; For kernel functions; This is the bias value.

Images