Diesel engine partial premixing combustion process prediction method and combustion model

Abstract

The invention provides a modeling method of a diesel engine partial premixed combustion model based on data driving. The modeling method comprises the following steps: establishing a heat release rate model about four oil injection parameters of diesel engine pilot injection timing, pilot injection fuel oil mass, main injection timing and main injection fuel oil mass and a crank angle based on a single Weber equation; according to the first law of thermodynamics, the cylinder pressure model is reversely pushed by the heat release rate model and is defined as a main model; a deviation model is established through the Gaussian process and used for describing the deviation between the main model and the real cylinder pressure, a training data set is updated in real time, and hyper-parameter optimization is carried out; the method comprises the following steps: predicting deviation through Gaussian regression, predicting the combustion process of the diesel engine in combination with a main model, establishing a diesel engine part premixed combustion model into two parts, namely a main model and a deviation model, describing the main model based on a single Weber equation, and performing regression fitting on the deviation model based on an online Gaussian process. The workload of determining model parameters in a traditional method is effectively reduced, and more accurate cylinder pressure prediction is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of diesel engines, and particularly relates to a method for predicting the partial premixed combustion process of a diesel engine and a combustion model. Background Art

[0002] Due to the rapid growth of global transportation trade, the importance of maritime cargo transportation has become more prominent. High reliability, low fuel consumption, and wide application range have enabled diesel engines to hold a leading position in the global marine diesel engine market, and more than 80% of the ships worldwide use diesel engines as the power source.

[0003] Among the combustion modes of diesel engines, compared with traditional combustion modes, partial premixed combustion (PPC) optimizes the fuel-air mixing degree through partial premixing of air and fuel, making the combustion more uniform and stable. Uniform combustion reduces heat loss, improves combustion efficiency, and also reduces the emissions of nitrogen oxides (NOx) and particulate matter (PM). In addition, by optimizing the fuel injection timing and the mixture formation process, PPC can avoid the problems of too fast combustion or uneven combustion in traditional compression ignition engines, and reduce the knocking risk during the combustion process.

[0004] The combustion process of PPC is a complex physical and chemical process, and the model established through a detailed chemical reaction kinetic mechanism is not conducive to control. Therefore, on the premise of meeting the real-time requirement, in order to ensure the accurate prediction of combustion parameters, the multi-Weber equation is generally used to predict the cylinder pressure. However, the multi-Weber equation requires a large number of parameters to be identified, consuming a lot of material and human resources. Summary of the Invention

[0005] Aiming at the problems of many identified parameters and low accuracy in establishing the PPC combustion model at present, the present invention proposes a method for predicting the partial premixed combustion process of a diesel engine, which can accurately predict the cylinder pressure with only a small number of Weber parameters to be identified.

[0006] A method for predicting the partial premixed combustion process of a diesel engine, characterized by comprising the following steps:

[0007] Step 1: Establish a heat release rate model described by a single Weber function for the four injection parameters of the pre-injection timing, pre-injection fuel mass, main injection timing, and main injection fuel mass of the diesel engine and the crankshaft angle;

[0008] Step 2: According to the first law of thermodynamics, inversely deduce a cylinder pressure model for estimating the cylinder pressure from the heat release rate model, and define it as the main model;

[0009] Step 3: Establish a deviation model using a Gaussian process to describe the cylinder pressure deviation between the main model and the actual cylinder pressure, update the training data set in real time, and perform hyperparameter optimization;

[0010] Step 4: Predict the cylinder pressure deviation through Gaussian regression, and combine the main model to predict the combustion process of the diesel engine. The sum of the cylinder pressure deviation and the estimated cylinder pressure is the final predicted cylinder pressure.

[0011] As a more preferable technical solution of the present invention, in Step 1, the heat release rate model is described by a single Wiebe function, and the expression form is as follows:

[0012]

[0013] In the formula: θ is the crankshaft angle, and the unit is °CA; is the cumulative representative heat release at θ, Q tot is the total representative heat release, and the units are both J; θ soc and θ eoc are the crankshaft angles at the start and end of combustion respectively; a and m are Wiebe parameters.

[0014] As a more preferable technical solution of the present invention, the formula for inversely deriving the cylinder pressure model from the heat release rate model in Step 2 is as follows:

[0015]

[0016] In the formula: γ is the specific heat ratio, is the in-cylinder pressure at θ, the unit is Pa, and V(θ) is the working volume at θ, the unit is m^3.

[0017] As a more preferable technical solution of the present invention, in Step 3, the deviation model is established using a Gaussian process as follows: m data pairs are used as the training data set of the Gaussian process, that is where the first injection timing θ pinj , the second injection timing θ minj , the first injection quantity m pinj , the second injection quantity m minj and the four injection parameters and the crankshaft angle are set as x i = [θ, θ pinj , θ minj , m pinj , m minj , and the deviation between the actual cylinder pressure and the cylinder pressure of the main model corresponding to the crankshaft angle θ is set as y i = [e p,θ . The online acquisition of the data set is determined by the root mean square error RMSE, and the formula is as follows:

[0018]

[0019] Where: n represents the number of crankshaft angles selected from a set of fuel injection parameters. Whenever the RMSE is less than a set threshold, it is considered that the cylinder pressure of the compensated model is consistent with the actual cylinder pressure, and at this time the data set does not need to be updated; when the RMSE is greater than this threshold, it is considered that the deviation model needs to be updated, and at this time the data set needs to be updated.

[0020] As a more preferable technical solution of the present invention, the expression form of the Gaussian process in step three is as follows:

[0021]

[0022] Where: K(x,x) is the covariance matrix, σ 2 is the noise variance, and the form of the kernel function is as follows:

[0023]

[0024] where σ f and l are both hyperparameters to be optimized, obtained by maximum likelihood estimation, and the log-likelihood function is as follows:

[0025]

[0026] As a more preferable technical solution of the present invention, the cylinder pressure deviation is predicted by Gaussian regression in step four as follows:

[0027] Taking x * as the prediction input, and taking μ * and K * as the mean and variance of the predicted output value respectively, the mean of the predicted output value is used as the actual cylinder pressure deviation, and the formula is as follows:

[0028]

[0029] The sum of the cylinder pressure deviation predicted by Gaussian process regression and the estimated cylinder pressure obtained by the main model is the final predicted cylinder pressure of the present invention.

[0030] Another object of the present invention is to provide a data-driven partially premixed combustion model for a diesel engine, including a main model and a deviation model. The main model is described based on a single Wiebe equation for estimating the cylinder pressure and is obtained by back-calculating from the heat release rate model. The deviation model is used to describe the cylinder pressure deviation between the main model and the actual cylinder pressure and is established by a Gaussian process; the sum of the cylinder pressure deviation and the estimated cylinder pressure is the final predicted cylinder pressure.

[0031] As a more preferable technical solution of the present invention, the main model is expressed as:

[0032]

[0033] As a more preferable technical solution of the present invention, the cylinder pressure deviation is predicted by Gaussian regression as follows:

[0034] Let x * represent the prediction input, and let μ * and K * represent the mean and variance of the prediction output value respectively. The mean of the prediction output value is used as the true cylinder pressure deviation. The formula is as follows:

[0035]

[0036] The beneficial effects are as follows:

[0037] The modeling method of the partially premixed combustion model of the diesel engine provided by the present invention includes two parts: establishing a main model and a deviation model. The main model is described based on the single Wiebe equation, and the deviation model is based on online Gaussian process regression fitting, thereby effectively reducing the workload of determining model parameters in the traditional method and achieving more accurate cylinder pressure prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a flowchart of the prediction method for the partially premixed combustion process of the diesel engine in Embodiment 1 of the present invention;

[0039] Figure 2 is a cylinder pressure prediction effect diagram when the main model has a small deviation in Embodiment 1 of the present invention.

[0040] Figure 3 is a cylinder pressure prediction effect diagram when the main model has a large deviation in Embodiment 1 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0041] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0042] The present invention provides a data-driven partially premixed combustion model for a diesel engine, including a main model and a deviation model. The main model is described based on the single Wiebe equation for estimating the cylinder pressure and is obtained by back-calculating from the heat release rate model. The deviation model is used to describe the cylinder pressure deviation between the main model and the true cylinder pressure and is established by Gaussian process; the sum of the cylinder pressure deviation and the estimated cylinder pressure is the final predicted cylinder pressure.

[0043] Embodiment 1

[0044] As Figure 1The following is a method for predicting the partial premixed combustion process of a diesel engine according to the present invention, including the following steps:

[0045] Step 1: Based on the single Wiebe equation, establish a heat release rate model regarding four injection parameters, namely the pre-injection timing, pre-injection fuel mass, main injection timing, and main injection fuel mass of the diesel engine, and the crankshaft angle;

[0046] The heat release rate model is as follows:

[0047]

[0048] In the formula: θ is the crankshaft angle, with the unit of °CA; is the cumulative representative heat release at θ, Q tot is the total representative heat release, and the units are both J; θ soc and θ eoc are the crankshaft angles at the start and end of combustion respectively; a and m are Wiebe parameters.

[0049] Then, calibrate the Wiebe parameters through experimental data, and finally the Wiebe parameters are selected as a = 6 and m = 2. Q tot , θ soc and θ eoc These three parameters are related to the injection strategy. In the case of two injections, four input quantities are defined as the first injection timing θ pinj , the second injection timing θ minj , the first injection quantity m pinj and the second injection quantity m minj . Therefore, there is a corresponding implicit relationship between the combustion parameters [Q tot , θ soc , θ eoc and the injection strategy [θ pinj , θ minj , m pinj , m minj .

[0050] Step 2: According to the first law of thermodynamics, inversely deduce the cylinder pressure model from the heat release rate model and define it as the main model;

[0051] Among them, the formula for inversely deducing the cylinder pressure model from the heat release rate model is as follows:

[0052]

[0053] In the formula: γ is the specific heat ratio; is the in-cylinder pressure at θ, with the unit of Pa; V(θ) is the working volume at θ, with the unit of m^3. The combustion process represented by the above formula is defined as the crankshaft angle within the range of -12 - 36. This is the focus of the present invention, and other parts are represented by the adiabatic process. The working volume is calculated based on the cylinder parameters and the piston position, and the calculation formula is as follows:

[0054]

[0055] In the formula, V c is the clearance volume, B is the cylinder diameter, l is the connecting rod length, and R is the crankshaft radius. The clearance volume can be calculated from the compression ratio r c as follows:

[0056]

[0057] Step 3: Establish a deviation model using the Gaussian process to describe the deviation between the main model and the real cylinder pressure, update the training data set in real time, and perform hyperparameter optimization;

[0058] In establishing the deviation model using the Gaussian process, m data pairs are used as the data set, that is where the injection parameters and the crankshaft angle are set as x i =[θ, θ pinj , θ minj , m pinj , m minj , and the deviation between the real cylinder pressure and the main model cylinder pressure corresponding to the crankshaft angle θ is set as y i =[e p,θ . The online acquisition of the data set is determined by the root mean square error RMSE, and the formula is as follows:

[0059]

[0060] In the formula: n represents the number of crankshaft angles selected from a set of injection parameters. Whenever RMSE is less than a set threshold, it is considered that the compensated model cylinder pressure is consistent with the real cylinder pressure, and at this time the data set does not need to be updated; when RMSE is greater than the threshold, it is considered that the deviation model needs to be updated, and at this time the data set needs to be updated.

[0061] The update rule is set as follows: 1) When the injection strategy does not change, that is, x i does not change, replace the y i under the current injection strategy with the deviation data of the previous cycle; 2) When the injection strategy changes, add the new (x i , y i ) data pair to the data set.

[0062] In Step 3, the expression form of the Gaussian process is as follows:

[0063]

[0064] Where: K(x, x) is the covariance matrix, and σ 2 is the noise variance. The form of the kernel function is as follows:

[0065]

[0066] where σ f and l are both hyperparameters to be optimized, obtained by maximum likelihood estimation. The log-likelihood function is as follows:

[0067]

[0068] Step 4: Predict the deviation through Gaussian regression, and combine with the main model to predict the combustion process of the diesel engine.

[0069] Among them, let x * represent the prediction input, and let μ * and K * represent the mean and variance of the prediction output value respectively. The mean of the prediction output value is used as the true cylinder pressure deviation. The formula is as follows:

[0070]

[0071] The sum of the cylinder pressure deviation predicted by Gaussian process regression and the estimated cylinder pressure obtained by the main model is the final predicted cylinder pressure of the present invention.

[0072] Taking the simulation experiment of a certain type of diesel engine as an example:

[0073] After applying the modeling method of the present invention, it can be seen from Figure 2 and Figure 3 that regardless of the size of the main model error, after Gaussian process compensation, the RMSE value of the model's predicted cylinder pressure is much smaller than that of the main model established based on the Wiebe equation and the first law of thermodynamics, indicating that applying Gaussian process compensation not only ensures the prediction accuracy but also effectively reduces the workload of determining model parameters under traditional methods.

[0074] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and can be implemented in other forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent conditions of the claims are intended to be included in the present invention. Any reference signs in the claims should not be regarded as limiting the claimed rights.

[0075] In addition, it should be understood that although this specification is described in terms of embodiments, not every embodiment contains only one independent technical solution. This narrative style of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for predicting a diesel engine partially premixed combustion process, characterized in that: The following steps are involved: Step 1, establishing a heat release rate model described by a single Weber function with respect to four injection parameters of the diesel engine, namely, pre-injection timing, pre-injection fuel mass, main injection timing, and main injection fuel mass, and a crankshaft angle; Step 2: According to the first law of thermodynamics, the main model for estimating cylinder pressure is obtained by inverse deduction from the heat release rate model; Step 3: Establish a deviation model for describing the cylinder pressure deviation between the main model and the actual cylinder pressure through Gaussian process, update the training data set in real time, and perform hyperparameter optimization; Step 4: Predict the cylinder pressure deviation through Gaussian regression, and predict the diesel engine combustion process in combination with the main model. The sum of the cylinder pressure deviation and the estimated cylinder pressure is the final predicted cylinder pressure.

2. The method for predicting a diesel engine partially premixed combustion process according to claim 1, characterized in that: In the step 1, the heat release rate model is described by a single Weibull function, which is expressed as follows: Where: θ is the crankshaft angle, unit is °CA; is the cumulative heat release at θ, Q tot It is the total heat release, and the unit is J; θ soc and θ eoc are the crankshaft angles at the start and end of combustion, respectively; a and m are Weber parameters.

3. The method for predicting a diesel engine partially premixed combustion process according to claim 1, characterized in that: The formula for inverting the cylinder pressure model from the heat release rate model in step 2 is as follows: Where: γ is the specific heat ratio, is the cylinder pressure at θ, in Pa, and V(θ) is the working volume at θ, in m^3.

4. The method for predicting a diesel engine partially premixed combustion process according to claim 1, characterized in that: In step 3, the deviation model is established using the Gaussian process as follows: Take m data pairs as the data set, that is The first injection timing θ pinj 、Second injection timingθ minj 、First injection amount m pinj 、The second injection amount m minj The four injection parameters and crankshaft angle are set to x i =[θ,θ pinj ,θ minj ,m pinj ,m minj ], the deviation between the actual cylinder pressure and the main model cylinder pressure at the corresponding crankshaft angle θ is set to y i =[e p,θ ]; The online collection of the data set is determined by the root mean square error RMSE, and the formula is as follows: Where n represents the number of crankshaft angles selected from a set of injection parameters; whenever the RMSE is less than a set threshold, it is considered that the compensated model cylinder pressure is consistent with the actual cylinder pressure, and the data set does not need to be updated; when the RMSE is greater than the threshold, it is considered that the deviation model needs to be updated, and the data set needs to be updated.

5. The method for predicting a diesel engine partially premixed combustion process according to claim 4, characterized in that: The Gaussian process in step 3 is expressed as follows: Where: K(x,x) is the covariance matrix, σ 2 is the noise variance, and the kernel function is as follows: where σ f and l are hyperparameters to be optimized, which are obtained by maximum likelihood estimation. The log-likelihood function is as follows:

6. The method for predicting a diesel engine partially premixed combustion process according to claim 1, characterized in that: In step 4, the cylinder pressure deviation is predicted by Gaussian regression as follows: x * Represents the prediction input, expressed in μ * and K * Represent the mean and variance of the predicted output values ​​respectively. The mean of the predicted output value is taken as the actual cylinder pressure deviation. The formula is as follows:

7. A data-driven diesel engine partial premixed combustion model, characterized in that: It includes a main model and a deviation model. The main model is described based on a single Weber equation and is used to estimate the cylinder pressure, which is obtained by inverse deduction of the heat release rate model. The deviation model is established through a Gaussian process and is used to describe the cylinder pressure deviation between the main model and the actual cylinder pressure. The sum of the cylinder pressure deviation and the estimated cylinder pressure is the final predicted cylinder pressure.

8. The data-driven diesel engine partial premixed combustion model according to claim 7, characterized in that: The main model is expressed as:

9. The data-driven diesel engine partial premixed combustion model according to claim 7, characterized in that: The cylinder pressure deviation is predicted by Gaussian regression as follows: x * Represents the prediction input, expressed in μ * and K * Represent the mean and variance of the predicted output values ​​respectively. The mean of the predicted output value is taken as the actual cylinder pressure deviation. The formula is as follows:

Images