A PID jet flow closed-loop control device based on online sensing for feedback information and a control method thereof

Abstract

The application discloses a PID injection amount closed-loop control device for feedback information of high-pressure natural gas direct injection engine based on online sensing and a control method thereof. Natural gas (1-1) of a natural gas supply system (1) is connected with a gas rail (1-3) through a compressor (1-2), the gas rail (1-3) is connected with a pressure sensor (2) and a rail pressure controller (4) respectively, the pressure sensor (2) is connected with an injector (3), the injector (3) is connected with a PXI trigger device (5), the PXI trigger device (5) and the pressure sensor (2) are connected with a charge amplifier (6) for charge-voltage conversion, and the PXI trigger device (5) and the rail pressure controller (4) are connected with an upper computer. The application is used to solve the problem of reducing the emission of high-pressure natural gas direct injection engine.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power energy, in particular to a HPDI jet quantity PID jet quantity closed-loop control device based on online sensing feedback information and a control method thereof. BACKGROUND

[0002] In recent years, with the energy crisis and the increasingly stringent emission regulations, natural gas as a substitute energy is more and more widely used as an engine alternative fuel due to its good emission to the environment and high thermal efficiency, as well as its rich reserves and greater price advantage, so as to reduce the environmental pollution pressure and relieve the energy problem. With the development of natural gas engines, its use is also extended to heavy-duty and marine engines, and there is a great demand for the power performance of the engine in these engines.

[0003] Currently, there are three modes of operation for natural gas engines. The first mode is a pure natural gas engine, in which natural gas is ignited by a spark plug. In the second mode, the engine operates using a premixed natural gas and direct injection pilot diesel, in which natural gas is introduced into a low-pressure dual-fuel engine during the intake stroke. In this type of engine, a homogeneous natural gas / air mixture is rapidly compressed below its self-ignition condition and is ignited by injecting a certain amount of pilot diesel near the top dead center. Engine load can only be controlled by changing the intake amount of natural gas. At medium and high loads, the emissions of the low-pressure dual-fuel engine are relatively low. However, at light loads, the unburned hydrocarbon emissions increase sharply. In addition, due to the limitation of engine knock, engine power and efficiency are also one of the challenges faced by low-pressure dual-fuel engines. In the third mode, both natural gas and pilot diesel are directly injected into the cylinder near the top dead center, which is commonly referred to as a high-pressure natural gas direct injection (HPDI) dual-fuel engine. In this type of engine, the pilot diesel is injected into the cylinder before the high-pressure natural gas. When the natural gas is injected into the cylinder, the diesel self-ignites. The natural gas is brought into the pilot flame and ignited. Due to the slow flame propagation speed of natural gas, pure natural gas engines are suitable for small engines, while dual-fuel engines are usually used for heavy-duty and marine engines. In recent studies, natural gas dual-fuel marine engines have been extensively studied. The advantages of HPDI natural gas engines are higher thermal efficiency and engine power, as the mixture controls the combustion rate and anti-knock characteristics. However, compared with low-pressure dual-fuel engines, high-pressure direct injection engines face higher nitrogen oxide and soot emissions.

[0004] In order to further improve the thermal efficiency of the high-pressure direct injection natural gas engine (HPDI) and reduce its emissions, scholars at home and abroad have studied its combustion characteristics, combustion model and emission characteristics through software simulation and experimental methods, so as to improve the performance of the engine. SUMMARY

[0005] The application provides a high-pressure natural gas direct injection engine jet quantity PID jet quantity closed-loop control device based on online sensing feedback information and a control method thereof, so as to improve the economic performance and emission performance of the HPDI engine.

[0006] The application is achieved by the following technical solutions:

[0007] A high-pressure natural gas direct injection engine jet quantity PID jet quantity closed-loop control device based on online sensing feedback information, the control device comprising a natural gas supply system 1, a pressure sensor 2, an injector 3, a rail pressure controller 4, a PXI trigger device 5, a charge amplifier 6 for charge-voltage conversion, and an upper computer 7.

[0008] The natural gas 1-1 of the natural gas supply system 1 is connected to the gas rail 1-3 through the air compressor 1-2, the gas rail 1-3 is connected to the pressure sensor 2 and the rail pressure controller 4 respectively, the pressure sensor 2 is connected to the injector 3, the injector 3 is connected to the PXI trigger device 5, the PXI trigger device 5 and the pressure sensor 2 are both connected to the charge amplifier 6 for charge-voltage conversion, and the PXI trigger device 5 and the rail pressure controller 4 are both connected to the upper computer.

[0009] A high-pressure natural gas direct injection engine jet quantity PID jet quantity closed-loop control device based on online sensing feedback information, the injector 3 sprays a preset jet quantity, the pressure change of the injector 3 is sensed by the clamping type high-precision pressure sensor 2 at the gas inlet of the injector 3, the feedback signal sensed by the clamping type high-precision pressure sensor 2 in real time is amplified by the charge amplifier 6 for charge-voltage conversion and input to the PXI trigger device 5, the PXI trigger device 5 transmits the signal to the upper computer 7 for jet quantity neural network calculation model identification, and the calculated jet quantity after identification is compared with the preset jet quantity to realize real-time and accurate control of the next jet quantity.

[0010] A control method of a high-pressure natural gas direct injection engine jet quantity PID jet quantity closed-loop control device based on online sensing feedback information, the control method specifically comprising the following steps:

[0011] Step 1: Assemble the control device and debug it;

[0012] Step 2: Establish a BP neural network fuel injection quantity calculation model based on the jet injection starting point pressure P inj and the maximum drop value ΔP of the inlet pressure measured in the device in step 1, so as to obtain the output quantity of the jet quantity m cyc .

[0013] Step 3: Compare the jet quantity m cycAs a feedback signal input into the PID control algorithm;

[0014] Step 4: iteration of the PID control algorithm to obtain the optimal solution.

[0015] A high-pressure natural gas direct injection engine injection amount PID injection amount closed-loop control method based on online sensing for feedback information, the step 2 specifically comprises the following steps,

[0016] Step 2.1: determine the injection amount neural network structure; the input of the input layer is P inj and ΔP, and the output layer is m cyc ; according to the following empirical formula, the number of hidden layers is determined:

[0017]

[0018] Wherein: l is the number of hidden layers, n is the number of input layer neurons, m is the number of output layer neurons, and a is a constant between 1 and 10; according to the above formula, the number of layers of the neural network is selected as 6 layers;

[0019] Step 2.2: initialize the weight threshold of the network structure;

[0020] Step 2.3: based on the experimental device of step 1, take the feedback signal and the actual injection amount, the feedback signal injection pressure P inj and the maximum drop value of inlet pressure ΔP;

[0021] Step 2.4: normalize the injection pressure P inj and the maximum drop value of inlet pressure ΔP and the actual injection amount to [0, 1], so that the data falls within the sensitive interval of the activation function, and the data normalization formula is:

[0022]

[0023] In the formula: x i is the original data, x’ i is the normalized data, x min is the minimum value in the original data, and x min is the maximum value in the original data;

[0024] Step 2.5: substitute the injection amount obtained in step 1 and the normalized values of injection pressure and inlet pressure drop value ΔP obtained in step 2.4 into the neural network model, and use the gradient descent method with momentum (Traingdx) as the network training function to train the neural network.

[0025] A high-pressure natural gas direct injection engine jet quantity PID jet quantity closed-loop control method based on online perception for feedback information, step 2.1 is specifically, the established BP neural network sets the activation function of the hidden layer and the output layer as logsig function, the formula is:

[0026]

[0027] The selected BP back propagation error function is:

[0028]

[0029] Where, t i is the desired output, O i is the network calculation output;

[0030] A high-pressure natural gas direct injection engine jet quantity PID jet quantity closed-loop control method based on online perception for feedback information, step 2.2 is specifically, the weight matrix from the input layer to the hidden layer is:

[0031]

[0032] The weight matrix from the hidden layer to the output layer is:

[0033] W2=(w 11 ′ w 12 ′ w 13 ′ w 14 ′ w 15 ′ w 16 ′)

[0034] The threshold matrix of the hidden layer activation function is:

[0035] θ=(θ1 θ2 θ3 θ4 θ5 θ6)

[0036] The threshold of the output layer activation function is θ';

[0037] A high-pressure natural gas direct injection engine jet quantity PID jet quantity closed-loop control method based on online perception for feedback information, step 2.5 is specifically, the selected BP back propagation error function is:

[0038]

[0039] Set the appropriate expected error, so that E<0.0001, if the calculation error obtained after training does not meet the accuracy requirement, update the weight threshold, and continuously train the network until the desired accuracy requirement is met, and the network training is completed.

[0040] A PID injection amount closed-loop control method for feedback information of high-pressure natural gas direct injection engine based on online sensing, the PID control algorithm is specifically expressed as formula (1):

[0041]

[0042] Therefore, the increment of the discretized PID algorithm is taken;

[0043] Then e (t) = e (k) - e (k-1) in formula (1),

[0044] The mathematical expression of the output of the increment formula PID control algorithm is shown as formula (2):

[0045] Delta u (k) = K P [e (k) - e (k-1) ] + K i e (k) + K d [e (k) - 2e (k-1) + e (k-2) ] (2)

[0046] Wherein K i is the integral coefficient of the control system; K d is the differential coefficient of the control system; Kp is the proportional coefficient; u (k) is the current output; Delta u (k) is the difference value between the current output and the last output; e (k) is the kth deviation; e (k-2) is the k-2th deviation.

[0047] The beneficial effects of the present application are:

[0048] The present application takes the pressure signal at the injector inlet as the real-time feedback signal, effectively improves the accurate and rapid closed-loop control of the natural gas injection amount in actual working conditions, and the method is convenient and fast.

[0049] The injection amount neural network calculation model proposed in the present application combines the strong adaptive ability of the PID algorithm and the strong anti-interference ability of the neural network, solves the problem that the prior art cannot solve the complex pressure fluctuation in the injection process in real time and efficiently, and compared with the solving mode based on the calculation or simulation model, the combination of the neural network and the PID closed-loop control algorithm is closer to the actual injection working condition.

[0050] The working environment of the additional sensor is stable, the risk of damage to most pressure sensors is avoided, the cost is saved, and the operation is simple. DETAILED DESCRIPTION

[0051] Figure 1 It is a structural schematic diagram of the present application.

[0052] Figure 2It is a jet amount neural network calculation model diagram of the present application.

[0053] Figure 3 It is a neural network structure diagram of the present application.

[0054] Figure 4 It is a PID control block diagram of the present application.

[0055] Figure 5 It is a flow chart of the present application. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0057] A high-pressure natural gas direct injection engine jet amount PID jet amount closed-loop control device based on online sensing for feedback information, the control device comprises a natural gas supply system 1, a pressure sensor 2, an injector 3, a rail pressure controller 4, a PXI trigger device 5, a charge amplifier 6 for charge-voltage conversion and a host computer 7;

[0058] The natural gas 1-1 of the natural gas supply system 1 is connected with the gas rail 1-3 through the compressor 1-2, the gas rail 1-3 is connected with the pressure sensor 2 and the rail pressure controller 4 respectively, the pressure sensor 2 is connected with the injector 3, the injector 3 is connected with the PXI trigger device 5, the PXI trigger device 5 and the pressure sensor 2 are connected with the charge amplifier 6 for charge-voltage conversion, and the PXI trigger device 5 and the rail pressure controller 4 are connected with the host computer.

[0059] A high-pressure natural gas direct injection engine jet amount PID jet amount closed-loop control device based on online sensing for feedback information, the injector 3 sprays a preset jet amount, the pressure change of the injector 3 is sensed by the clamping type high-precision pressure sensor 2 at the gas inlet of the injector 3, the feedback signal sensed by the clamping type high-precision pressure sensor 2 in real time is amplified by the charge amplifier 6 for charge-voltage conversion and input into the PXI trigger device 5, the PXI trigger device 5 transmits the signal to the host computer 7 for jet amount neural network calculation model identification, and the calculated jet amount after identification is compared with the preset jet amount to realize real-time and accurate control of the next jet amount.

[0060] A control method of a high-pressure natural gas direct injection engine jet amount PID jet amount closed-loop control device based on online sensing for feedback information, the control method specifically comprises the following steps:

[0061] Step 1: Assemble the control device and debug;

[0062] Step 2: Establish the jet pressure P based on the jet start point measured in the device in step 1 inj And the BP neural network fuel injection amount calculation model of the maximum drop value ΔP of the inlet pressure, the output quantity is the jet amount m cyc ;

[0063] Step 3: The jet amount m cyc In step 2 is input as a feedback signal into the PID control algorithm;

[0064] Step 4: Iterative PID control algorithm to obtain the optimal solution.

[0065] A high-pressure natural gas direct injection engine jet amount PID jet amount closed-loop control method based on online sensing as feedback information, the step 2 specifically includes the following steps,

[0066] Step 2.1: Determine the jet amount neural network structure; the BP neural network algorithm of the present application is composed of input layer, hidden layer and output layer, a total of three layers. The input of the input layer is P inj And ΔP, the output layer is m cyc ; According to the following empirical formula, the number of hidden layers is determined:

[0067]

[0068] Where: l is the number of hidden layers, n is the number of input layer neurons, m is the number of output layer neurons, and a is a constant between 1 and 10; According to the above formula, the number of layers of the neural network is selected as 6 layers;

[0069] The established BP neural network sets the activation function of the hidden layer and the output layer as the logsig function, and the formula is:

[0070]

[0071] The selected BP backpropagation error function is:

[0072]

[0073] Where, t i is the desired output, and O i is the calculated output of the network;

[0074] Step 2.2: Initialize the weight threshold of the network structure;

[0075] The weight matrix from the input layer to the hidden layer is:

[0076]

[0077] The weight matrix from the hidden layer to the output layer is:

[0078] W2 = (w 11 w 12 w 13 w 14 w 15 w 16 )

[0079] The threshold matrix of the hidden layer activation function is:

[0080] θ = (θ1θ2θ3θ4θ5θ6)

[0081] The threshold of the output layer activation function is θ';

[0082] Step 2.3: Based on the experimental device of step 1, take the feedback signal and the actual jet flow, the feedback signal jet pressure P inj and the maximum drop value of inlet pressure ΔP;

[0083] Step 2.4: The jet pressure P inj and the maximum drop value of inlet pressure ΔP and the actual jet flow are normalized to [0, 1] to make the data fall within the sensitive interval of the activation function. The data normalization formula is:

[0084]

[0085] In the formula: x i is the original data, x' i is the normalized data, x min is the minimum value in the original data, x min is the maximum value in the original data;

[0086] Step 2.5: The jet flow obtained in step 1 and the jet pressure and inlet pressure drop value ΔP normalized in step 2.4 are substituted into the neural network model, and the network training function adopts the gradient descent method with momentum (Traingdx). The neural network is trained.

[0087] The selected BP backpropagation error function is:

[0088]

[0089] Set the appropriate expected error so that E < 0.0001. If the calculated error after training does not meet the accuracy requirement, update the weight threshold and continuously train the network until the desired accuracy requirement is met. The network training can be completed.

[0090] A PID injection quantity closed-loop control method for feedback information of high-pressure natural gas direct injection engine injection quantity based on online sensing, the PID control algorithm is specifically expressed as formula (1):

[0091]

[0092] Because the digital PID incremental control algorithm is a recursive algorithm, the actuator needs the increment of the control quantity of the current output, and the increment is only related to the deviation of the last several samplings, easy to realize disturbance-free switching manually or automatically, and does not affect integral out-of-control, so the increment is taken after the PID algorithm is discretized;

[0093] Then e(t) in formula (1) is e(k)-e(k-1),

[0094] Then the mathematical expression of the output of the incremental PID control algorithm is shown as formula (2):

[0095] Delta u(k)=K P [e(k)-e(k-1)]+K i [e(k)-e(k-1)]+K d [e(k)-2e(k-1)+e(k-2)] (2)

[0096] Wherein K i is the integral coefficient of the control system, K d is the differential coefficient of the control system, K p is the proportional coefficient, u(k) is the current output, Delta u(k) is the difference value between the current output and the last output, e(k) is the kth deviation, and e(k-2) is the k-2th deviation.

[0097] The input and output of the PID control algorithm are identified in MATLAB, the transfer function is established, the parameter value of the PID is adjusted, and the system is controlled. Further, the Kp value is adjusted first, so that the range of the system output value is 10% above and below the expected range, then the Ki and Kd values are adjusted, so that the output of the system can quickly reach the target value, and the system is stably within 2% above and below the target value.

[0098] (1) The application provides a PID closed-loop control method for high-pressure natural gas direct injection engine injection quantity based on online sensing as feedback information. Figure 1 As shown in the figure.

[0099] The natural gas is compressed by a compressor, the compressed high-pressure natural gas is injected through a gas rail to an injector, a high-precision pressure sensor is clamped and installed at the inlet of the injector to collect a pressure feedback signal, the inlet pressure feedback signal is amplified by an amplifier and then collected by a PXI, and finally closed-loop control is realized by a PID controller.

[0100] (2) Figure 2 The jet flow neural network calculation model of the application is provided. The complex injection problem in the injector is processed by the simple jet flow neural network calculation model. Real-time, efficient and accurate calculation of the jet flow is realized.

[0101] (3) Figure 3 The jet flow neural network structure diagram of the application is provided. The injection pressure P inj at the inlet of the injector and the maximum drop value ΔP of the inlet pressure are taken as the input layer, the jet flow m cyc is taken as the output layer, and the optimal fifteen hidden layers are provided.

[0102] (4) Figure 4 The PID control block diagram of the application is provided, Figure 5 The algorithm flowchart of the application is provided. The injector inlet pressure signal is taken as an online sensing signal, the difference between the preset jet flow and the calculated jet flow is substituted into the PID controller, and the injection of the jet flow is accurately controlled.

[0103] The above is the specific implementation process of the single PID closed-loop control method, and the above process can be repeated to realize the high-pressure natural gas cylinder direct injection engine jet flow PID closed-loop control method with online sensing as the feedback information.

Claims

1. A control method for a high-pressure natural gas direct injection engine jet quantity PID jet quantity closed-loop control device based on online sensing feedback information, characterized in that, The control device includes a natural gas supply system (1), a clamp-type high-precision pressure sensor (2), an injector (3), a rail pressure controller (4), a PXI trigger device (5), a charge amplifier for charge-to-voltage conversion (6), and a host computer (7). The natural gas (1-1) of the natural gas supply system (1) is connected to the gas rail (1-3) through the compressor (1-2). The gas rail (1-3) is connected to the clamp-type high-precision pressure sensor (2) and the rail pressure controller (4) respectively. The clamp-type high-precision pressure sensor (2) is installed at the gas inlet of the injector (3) to sense the pressure change of the injector (3) in real time and generate a feedback signal. The injector (3) is connected to the PXI trigger device (5). The PXI trigger device (5) and the clamp-type high-precision pressure sensor (2) are both connected to the charge amplifier (6) of charge-to-voltage conversion to receive the amplified signal and trigger the injector (3) to act. The PXI trigger device (5) and the rail pressure controller (4) are both connected to the host computer. The injector (3) ejects a preset amount of air. The pressure change of the injector (3) is sensed by the clamp-type high-precision pressure sensor (2) at the gas inlet of the injector (3). The feedback signal sensed in real time by the clamp-type high-precision pressure sensor (2) is amplified by the charge amplifier (6) of the charge-to-voltage conversion and input to the PXI trigger device (5). The PXI trigger device (5) transmits the signal to the host computer (7) for recognition of the air volume neural network calculation model. The recognized and calculated air volume is compared with the preset air volume to achieve real-time and accurate control of the next air volume. The control method specifically includes the following steps: Step 1: Assemble the control device and debug it; Step 2: Establish the jet pressure P at the jet initiation point measured in the device in Step 1. inj And the BP neural network fuel injection quantity calculation model for the maximum inlet pressure drop value ΔP, yields the output quantity m. cyc ; Step 3: The jet volume m from Step 2 cyc It is input as a feedback signal into the PID control algorithm; Step 4: Use the iterative PID control algorithm to tune the parameters and obtain the optimal solution; The PID control algorithm is specifically expressed as shown in equation (1): Therefore, the PID algorithm is discretized and the increment is taken; Then in equation (1), e(t) = e(k) - e(k-1), The mathematical expression for the output of the incremental PID control algorithm is shown in equation (2): Δu(k)=K P [e(k)-e(k-1)]+K i e(k)+K d [e(k)-2e(k-1)+e(k-2)] (2) Where K i K is the integral coefficient of the control system. d denoted by ...

2. The control method of the PID closed-loop control device for high-pressure natural gas direct injection engine fuel injection quantity based on online sensing feedback information as described in claim 1, characterized in that, Step 2 specifically includes the following steps. Step 2.1: Determine the structure of the jet volume neural network; the input of the input layer is P. inj With ΔP, the output layer is m cyc ; The number of hidden layers can be determined using the following empirical formula: Where: l is the number of hidden layers, n is the number of neurons in the input layer, m is the number of neurons in the output layer, and a is a constant between [1, 10]; according to the above formula, the number of layers in the neural network is selected as 6. Step 2.2: Initialize the weight thresholds of the network structure; Step 2.3: Based on the experimental setup in Step 1, collect the feedback signal and the actual jet volume. The feedback signal jet pressure P inj and the maximum drop in inlet pressure ΔP; Step 2.4: Adjust the jet pressure P inj The maximum drop in inlet pressure ΔP and the actual jet volume are normalized to [0,1] so that the data falls within the sensitive range of the activation function. The data normalization formula is: In the formula: x i Here, x'i represents the original data, and x'i represents the normalized data. min x is the minimum value in the original data. min The maximum value in the original data; Step 2.5: Substitute the normalized values ​​of the jet volume obtained in Step 1 and the jet pressure and inlet pressure drop value ΔP obtained in Step 2.4 into the neural network model. The neural network training function adopts the gradient descent method with momentum (Traingdx) to train the neural network.

3. The control method of the PID closed-loop control device for high-pressure natural gas direct injection engine fuel injection quantity based on online sensing feedback information as described in claim 2, characterized in that, Step 2.1 specifically involves setting the activation functions of both the hidden and output layers of the established BP neural network to the logsig function, as shown in the formula: The selected BP backpropagation error function is: Among them, t i For the desired output, O i This is the computational output of the network.

4. The control method of the PID closed-loop control device for high-pressure natural gas direct injection engine fuel injection quantity based on online sensing feedback information as described in claim 2, characterized in that, Specifically, step 2.2 involves setting the weight matrix from the input layer to the hidden layer as follows: The weight matrix from the hidden layer to the output layer is: W2=(in 11 'In 12 'In 13 'In 14 'In 15 'In 16 ′) The threshold matrix of the hidden layer activation function is: θ = (θ1θ2θ3θ4θ5θ6) The threshold of the activation function in the output layer is θ′.

5. The control method of the PID closed-loop control device for high-pressure natural gas direct injection engine fuel injection quantity based on online sensing feedback information as described in claim 2, characterized in that, Step 2.5 specifically involves selecting the BP backpropagation error function as follows: Set an appropriate expected error such that E < 0.0001. If the calculation error obtained after training does not meet the accuracy requirements, update the weight threshold and continue to train the network until the expected accuracy requirements are met, and then the network training is completed.

Images