Intelligent variable mode control method for variable altitude turbocharging system of diesel engine

An intelligent variable mode control method for variable altitude turbocharging system of diesel engine, characterized by comprising:

• step one:
• querying a turbocharging pressure MAP chart according to demand signals of speed and torque of the diesel engine, and reading a target turbocharging pressure, and correcting the target turbocharging pressure based on an intake environmental temperature under altitude environmental conditions measured by a temperature sensor upstream of a low-pressure stage compressor, to obtain a corrected target turbocharging pressure;
• and,
• step two:
• using a fuzzy control algorithm to carry out intelligent selection of variable turbocharging mode based on adaptive ranges of flow and pressure ratio in single-stage and two-stage turbocharging modes, to obtain an adjustment range of high- and low-pressure stage bypass valve opening degrees;
• and,
• step three:
• under actual environmental conditions at different altitudes, initializing the opening degrees of the high- and low-pressure stage bypass valves within the adjustment range, and using a neural network rapid prediction model of diesel engine turbocharging pressure to carry out turbocharging pressure prediction for the variable altitude turbocharging system;
• and,
• step four:
• using a genetic algorithm to fast search opening degrees of the high- and low-pressure stage bypass valves in a corresponding turbocharging mode until the turbocharging pressure reaches a corresponding target value, to output to a valve control unit of the bypass valves;
• wherein
• in step two, the fuzzy control algorithm comprises:
• fuzzifying the real-time signals of speed, throttle, and altitude;
• and using the fuzzified signals as inputs of variable mode control fuzzy rules of the turbocharging system to complete a fuzzy inference process;
• and selecting a turbocharging mode based on an output obtained by the inference process;
• and obtaining the adjustment range of the high- and low-pressure stage bypass valves;
• wherein, the speed is divided into seven levels according to the original speed:
• a critical low speed XS of 800-1000 rpm, a low speed S of 1000-1200 rpm, a medium low speed MS of 1200-1400 rpm, a medium speed M of 1400-1600 rpm, a medium high speed MB of 1600-1800 rpm, a high speed B of 1800-2000 rpm, and a critical high speed XB of 2000-2300 rpm;
• and, according to changes in altitude, the altitude is divided into:
• a low altitude L of 0-1500 m, a medium high altitude O of 1500-3000 m, a high altitude H of 3000-4500 m;
• and,
• a symmetrical triangle is selected as a membership function;
• and the fuzzy rule for the intelligent variable mode control method for variable altitude turbocharging system of diesel engine are establish based on adaptable ranges of flow and pressure ratio in different turbocharging modes;
• and a COG anti-fuzzy method is used to obtain a turbocharging mode for current operating conditions and altitude, and the adjustment range of the high- and low-pressure stage bypass valves is determined according to the turbocharging mode;
• wherein the different turbocharging modes are one-state mode and a two-stage mode;
• and,
• in step three, the neural network rapid prediction model of diesel engine turbocharging pressure adopts a BP neural network model;
• wherein, according to engine speed and torque operating conditions, inlet temperatures and pressures of high- and low-pressure stage compressors, and opening degrees of the high- and low-pressure stage bypass valves are used as input layer neurons to predict a diesel engine turbocharging pressure under current operating conditions;
• and wherein the cost function for neural network training is:
• J
• =
• 1
• 2
• ⁢
• (
• P
• test
• -
• P
• pre
• )
• 2
• +
• φ
• ⁢
• ∑
• n
• i
• =
• 1
• θ
• i
• 2
• ;
• wherein, Ptest is a measure turbocharging pressure in Pa;
• and Ppre is a turbocharging pressure predicted by the neural network in Pa;
• and φ is a regularization parameter used to minimize a weight matrix during an optimization search press to obtain a most stable fast prediction model of the turbocharging pressure;
• and n in the number of input neurons of the BP neural network;
• and θi is a weight coefficient;
• and,
• in step four, the genetic algorithm fast searches opening degrees of the high- and low-pressure stage bypass valves in the corresponding turbocharging mode, and binary rules are used to encode the opening degrees of the bypass valves, and a population number is initialized to 50, and differences between the target turbocharging pressure and predicted turbocharging pressures are used as a fitness function for individual evaluations;
• wherein a calculation formula of the fitness function is:
• F(t)=lg(|Pobj−Ppre|)+1
• wherein, t is an evolutionary algebra, and Pobj is the corrected target turbocharging pressure in Pa;
• and Ppre is the predicted turbocharging pressure in Pa;
• and,
• the genetic algorithm selects excellent individuals to inherit to a next generation population based on individual fitness, wherein a crossover probability is 0.8, and a mutation probability is 0.1, and a termination condition is:
• if a predetermined number of evolutionary generations is reached, or an average fitness of the population has no significant improvement in several consecutive generations, then the individual with a maximum fitness obtained during an evolutionary search process is selected as an optimal solution output, to obtain optimal opening degrees of the high- and low-pressure stage bypass valves which meet air intake requirements at varying altitudes, and meet needs for variable mode adaptation of the diesel engine in all working conditions at varying altitudes.