Intelligent control algorithm for marine engine

Abstract

According to the intelligent control algorithm for the marine engine, by optimizing the fuel injection timing and the rail pressure in the transient process, the purpose of collaborative optimization of NOx emission and the IMEP is achieved, and the contradiction that engine emission and fuel consumption are difficult to reconcile is solved.

Description

Technical Field

[0001] The invention relates to the technical field of marine engines, and in particular to an intelligent control algorithm for marine engines. Background Art

[0002] With the rapid development of technologies such as electronic control and new fuels, diesel engines have increasingly intelligent and high-precision control systems in order to meet increasingly stringent emission regulations and higher performance requirements. As the system becomes increasingly complex, engine control variables continue to increase and couple with each other, which poses new challenges to control algorithms. In order to solve the irreconcilable contradiction between engine emissions and fuel consumption, controllable parameters are optimized to solve complex multi-objective optimization problems. Summary of the invention

[0003] The purpose of the present invention is to optimize the injection timing and rail pressure in the transient process, so as to achieve the purpose of synergistically optimizing NOx emissions and mean effective pressure IMEP.

[0004] The technical solution of the present invention provides an intelligent control algorithm for a marine engine, comprising the following steps:

[0005] According to the MPC system theory, the state quantity, control input quantity and MPC system output quantity of the MPC controller in the form of state space matrix are obtained;

[0006] According to the state variables of the marine engine such as intake pressure, injection amount, cycle speed, NOx emission, mean effective pressure, injection timing and rail pressure, combined with the state quantity of the MPC controller, control input quantity and MPC system output quantity, the following formula is obtained:

[0007] X k =[p intake m fiel N cycle NOx IMEP] T

[0008] Y k =U k =[T fuel p rail ] T

[0009] Among them, X is the state variable of the MPC system, U is the control variable of the MPC system, Y is the output variable of the MPC system, k is the sampling time, the unit is cycle, p intake is the intake pressure, in kPa, m fuel is the fuel injection amount, in mg, N cycle is the cycle speed in r / min, NOx is the NOx emission in ppm, IMEP is the mean effective pressure in bar, Tfuel is the injection timing, unit is °CA ATDC, p rail is the rail pressure, in bar;

[0010] The partial derivatives of the state variables of the marine engine with respect to the control input are obtained to construct the B matrix in the state space matrix, thereby obtaining the A matrix and the C matrix in the state space matrix;

[0011] The injection timing constraint formula and rail pressure constraint formula are used to constrain the injection timing and rail pressure;

[0012] Construct a cost function for optimizing NOx emissions and mean effective pressure IMEP;

[0013] The continuous least square method is used to solve the cost function under the injection timing constraint formula and the rail pressure constraint formula, so that the cost function returns the minimum NOx emission and mean effective pressure value;

[0014] NOx emissions and mean effective pressure are optimized based on the expressions corresponding to the state quantities of the MPC controller in the form of a state space matrix, the control input quantities and the MPC system output quantities, and the cost function formula.

[0015] Preferably, the injection timing constraint formula is as follows:

[0016] -16≤T fuel ≤-6.

[0017] Preferably, the rail pressure constraint formula is as follows:

[0018] 800≤p rail ≤1200.

[0019] Preferably, the cost function formula is as follows:

[0020]

[0021] Where N is the prediction time step, α and β are weighting coefficients, and k+1 is the sampling time.

[0022] Preferably, the intake pressure and the cycle speed are directly acquired by the MPC system.

[0023] Preferably, the fuel injection amount is calculated and obtained according to the circulation speed and the target speed in combination with a PID controller.

[0024] Preferably, the initial values ​​of the injection timing and the rail pressure are obtained by uniformly distributed random sampling.

[0025] Preferably, the prediction time step N and weighting coefficients α and β are automatically adjusted based on the BP neural network.

[0026] Preferably, the input of the BP neural network is intake pressure, injection amount, cycle speed, injection timing and rail pressure, the middle hidden layer is 1 layer of 10 neurons, and the output is the prediction time step N and weighting coefficients α and β.

[0027] Preferably, the state space matrix is ​​as follows:

[0028]

[0029]

[0030] The technical solution of the present invention proposes an intelligent control algorithm for marine engines, which optimizes the injection timing and rail pressure in the transient process, thereby achieving the purpose of synergistically optimizing NOx emissions and mean effective pressure IMEP, and solving the irreconcilable contradiction between engine emissions and fuel consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 A schematic diagram showing the comparison between NOx and IMEP obtained using MPC control and the original values;

[0032] Figure 2 This is a schematic diagram of the optimized rail pressure and injection timing. DETAILED DESCRIPTION

[0033] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the appended claims of the application equally.

[0034] An embodiment of the present invention provides an intelligent control algorithm for a marine engine, and the algorithm is implemented as follows:

[0035] (1) According to the MPC system theory, the state space matrix form of the system is expressed as follows:

[0036] X k+1 =AX k +BU k

[0037] Y k =CX k

[0038] Where: X is the state variable of the MPC system, U is the control variable of the MPC system, Y is the output variable of the MPC system, A, B, C are state space matrices, and subscripts k and k+1 represent sampling times.

[0039] Design the state quantity, control input quantity, and MPC system output quantity of the MPC controller as follows:

[0040] X k =[p intake m fuel N cycle NOx IMEP] T

[0041] Y k =U k =[T fuel p rail ] T

[0042] The variables are defined as follows:

[0043] p intake : Inlet pressure, m fuel : Fuel injection quantity, N cycle : Cycle speed, NOx: NOx emissions, IMEP: Mean effective pressure, T fuel : Injection timing, p rail : Rail pressure.

[0044] The model is locally linearized using the Euler approximation method, and the local area of ​​the nonlinear curve is approximated by a line segment. The partial derivatives of each system state quantity with respect to the control variable (control input quantity) are calculated to construct the B matrix, and finally the state space matrices are obtained as follows:

[0045]

[0046] In order to ensure the accuracy of the model, the injection timing and rail pressure are constrained:

[0047] -16≤T fuel ≤-6

[0048] 800≤p rail ≤1200

[0049] In order to optimize NOx emissions and mean effective pressure IMEP, the cost function is constructed as follows:

[0050]

[0051] Where N is the prediction time step, α and β are weighting coefficients, that is, the system state variables and control variables at the next N moments are obtained through prediction, and the continuous least squares method, namely SLSQP method, is used to solve the control variable value (NOx emission and mean effective pressure value) that makes the cost function return the minimum under the constraints.

[0052] (2) From the above, it can be seen that the system state variables are intake pressure, injection amount, cycle speed, NOx emissions and mean effective pressure. Among them, the intake pressure and cycle speed are directly obtained, and the injection amount is calculated by the cycle speed, target speed and PID controller; in order to obtain IMEP and NOx emission variables, a transient NOx emission prediction model, an IMEP prediction model and a cycle speed prediction model are constructed respectively, and the model input is selected according to the correlation analysis. The following table shows the input parameters determined by each model, among which the speed prediction model is used to obtain the cycle speed at the next moment, so as to input the NOx transient prediction model and obtain the NOx emission prediction value.

[0053]

[0054] The following is an implementation plan for building an MPC controller:

[0055] (1) Obtain transient process data sets under different operating conditions of marine engines and perform normalization processing.

[0056] (2) The injection amount is calculated based on the current cycle speed and target speed through the PID algorithm.

[0057] (3) The initial values ​​of the target parameters, namely the injection timing and rail pressure, are obtained by using uniformly distributed random sampling.

[0058] (4) Since the real-time optimization of MPC is slow, in order to achieve real-time optimization at different speeds, the rolling window length N and multi-parameter weights α and β are automatically adjusted based on the BP neural network. The neural network structure is a 5-10-3 three-layer structure, that is, the input is the 5 parameters of the MPC state variables, the middle hidden layer is a layer of 10 neurons, the output is the prediction time step N, the weighting coefficients α and β are three nodes, the hidden layer activation function is sigmoid, and the learning algorithm uses back propagation to calculate the gradient of the error layer by layer through the chain rule, and uses the gradient descent method to update the parameters in the network.

[0059] (5) During rolling optimization, the model prediction values ​​of the X(k+1), X(k+2), …, X(k+N) states are calculated based on the discrete linear state space expression combined with the transient NOx emission prediction model, IMEP prediction model, cycle speed prediction model and cost function formula, and the optimized outputs of the U(k+1), U(k+2), …, U(k+N) controllers are solved.

[0060] Using MPC control to obtain NOx and IMEP compared with the original values Figure 1 As shown, the optimized rail pressure and injection timing are as follows Figure 2 The IMEP and NOx optimization effects are shown in Table 1.

[0061] Table 1 IMEP and NOx optimization effects

[0062]

[0063] Depend on Figure 1 , Figure 2 It can be seen from Table 1 that the algorithm can achieve 26% NOx optimization and 2% IMEP optimization control, with good optimization performance.

Claims

1. An intelligent control algorithm for marine engines, It is characterized in that The following steps are involved: According to the MPC system theory, the state quantity, control input quantity and MPC system output quantity of the MPC controller in the form of state space matrix are obtained; According to the state variables of the marine engine such as intake pressure, injection amount, cycle speed, NOx emission, mean effective pressure, injection timing and rail pressure, combined with the state quantity of the MPC controller, control input quantity and MPC system output quantity, the following formula is obtained: X k =[p intake m fuel N cycle NOx IMEP] T Y k =U k =[T fuel p rail ] T Among them, X is the state variable of the MPC system, U is the control variable of the MPC system, Y is the output variable of the MPC system, k is the sampling time, p intake is the intake pressure, m fuel is the fuel injection amount, N cycle is the cycle speed, NOx is the NOx emission, IMEP is the mean effective pressure, T fuel is the injection timing, p rail is the rail pressure; The partial derivatives of the state variables of the marine engine with respect to the control input are obtained to construct the B matrix in the state space matrix, thereby obtaining the A matrix and the C matrix in the state space matrix; The injection timing constraint formula and rail pressure constraint formula are used to constrain the injection timing and rail pressure; Construct a cost function for optimizing NOx emissions and mean effective pressure IMEP; The continuous least square method is used to solve the cost function under the injection timing constraint formula and the rail pressure constraint formula, so that the cost function returns the minimum NOx emission and mean effective pressure value; NOx emissions and mean effective pressure are optimized based on the expressions corresponding to the state quantities of the MPC controller in the form of a state space matrix, the control input quantities and the MPC system output quantities, and the cost function formula.

2. An intelligent control algorithm for a marine engine as claimed in claim 1, It is characterized in that The injection timing constraint formula is as follows: -16≤T fuel ≤-6。 3. An intelligent control algorithm for a marine engine as claimed in claim 1, It is characterized in that The rail pressure constraint formula is as follows: 800≤p rail ≤1200。 4. An intelligent control algorithm for a marine engine as claimed in claim 1, It is characterized in that The cost function formula is as follows: Where N is the prediction time step, α and β are weighting coefficients, and k+1 is the sampling time.

5. An intelligent control algorithm for a marine engine as claimed in claim 1, It is characterized in that The intake pressure and the cycle speed are directly obtained by the MPC system.

6. An intelligent control algorithm for a marine engine as claimed in claim 1, It is characterized in that The injection amount is obtained by calculating the cycle speed and the target speed in combination with a PID controller.

7. An intelligent control algorithm for a marine engine as claimed in claim 1, It is characterized in that The initial values ​​of the injection timing and the rail pressure are obtained by uniformly distributed random sampling.

8. An intelligent control algorithm for a marine engine as claimed in claim 4, It is characterized in that The prediction time step N and weighting coefficients α and β are automatically adjusted based on the BP neural network.

9. An intelligent control algorithm for a marine engine as claimed in claim 8, It is characterized in that The input of the BP neural network is intake pressure, injection amount, cycle speed, injection timing and rail pressure, the middle hidden layer is 1 layer with 10 neurons, and the output is the prediction time step N and weighting coefficients α and β.

10. An intelligent control algorithm for a marine engine as claimed in claim 1, It is characterized in that The state space matrix is ​​as follows:

Images