CN110674578A

CN110674578A - Fuel cell polarization curve segmentation identification algorithm applicable to single chip microcomputer

Info

Publication number: CN110674578A
Application number: CN201910909583.2A
Authority: CN
Inventors: 陈晓高; 熊保鸿; 顾胜升
Original assignee: Wuxi Martin Green Photovoltaic Science And Technology Ltd
Current assignee: Wuxi Martin Green Photovoltaic Science And Technology Ltd
Priority date: 2019-09-19
Filing date: 2019-09-19
Publication date: 2020-01-10

Abstract

A fuel cell polarization curve segmentation identification algorithm for a single chip microcomputer is characterized in that a linear segment of a polarization curve is extracted, namely a low current density interval and a large current density interval introduced above, after extraction, a theta angle and a median filter are calculated through a formula to determine a threshold range, and after the threshold range is determined, a data point of the polarization interval and the current density of the last point of the interval can be extracted to serve as J_fAfter the data are obtained, J is taken according to the interval segmentation idea and is less than or equal to J_fInterval obtained as E_eU of A and R_iData identification parameters, take J > J_fInterval obtains U of m' and n_iAnd identifying parameters by data, and finally fitting the parameters identified by the algorithm to obtain a required polarization curve of the fuel cell. The algorithm converts the polarization curve nonlinear parameter identification of the fuel cell into two linear least square identification subproblems by using the interval segmentation concept.

Description

Fuel cell polarization curve segmentation identification algorithm applicable to single chip microcomputer

Technical Field

The invention relates to an identification algorithm of a fuel cell polarization curve, in particular to a fuel cell polarization curve segmentation identification algorithm which can be used for a single chip microcomputer.

Background

The electric energy conversion efficiency of the fuel cell decreases with the increase of the operation time, mainly due to the fact that the output voltage is lower than that in the initial stage under the same current density. When the efficiency is degraded too much, if the controller of the fuel cell still calculates the required power of the fuel cell according to the initial polarization curve, when the required power is constant, the calculated required current will be smaller than the actual required current of the fuel cell, which will result in that the flow rate of the inflowing air and hydrogen is lower than the actual required flow rate of the fuel cell, resulting in the deterioration of the dynamic and static characteristics of the system.

Although most of the existing optimization algorithms have high identification precision, the existing optimization algorithms need strict initial value setting and the like, the calculation amount is large, and the single chip microcomputer with limited calculation resources cannot be met.

Disclosure of Invention

The invention aims to overcome the defects, so that the fuel cell polarization curve segmentation identification algorithm for the single chip microcomputer is provided, the initial value of the parameter is not required to be set, the occupied computing resource is less, the precision is equivalent to that of a Newton method, and the method is very suitable for the environment of the single chip microcomputer.

The technical scheme of the invention is as follows: the algorithm is obtained through an interval segmentation idea based on the following numerical model (formula (1)).

U_i＝E_e-A ln(J_i)-RJ_i-m′exp(nJ_i) (1)

In the formula (1), E_e＝E₀+A ln(j₀)-RJ_n，m′＝mexp(nJ_n)。

There are 5 parameters to be identified in formula (1), which are: e_eA, R, m', and n. The model is a nonlinear model, parameters of the model are estimated by a nonlinear least square method (such as a Newton method) generally, but high-dimensional matrix inversion and initial value setting are required, the robustness is poor, and the calculated amount is large.

The algorithm of the invention provides the following identification method through an interval segmentation pair formula (1):

will-A ln (J)_i) Corresponding to the activated polarization region;

will-RI_iCorresponding to the ohmic polarization region;

mixing-m' exp (nJ)_i) Corresponding to the concentration polarization zone.

The active polarization characteristic term and the ohmic polarization characteristic term represent a low current density interval, and the concentration polarization characteristic term represents a high current density interval.

Low current density interval:

U_i＝E_e-A ln(J_i)-RJ_i(2)

passing formula (1) through a reactive polarizerData pair parameter E for zone and ohmic polarization zone_eAnd A and R are subjected to linear least square identification to obtain a formula (3)

Y_i＝m′exp(nJ_i) (3)

In the formula (3), Y_i＝-[U_i-E_e+A ln(J_i)+RI_i]Taking natural logarithm from two sides of the formula (3) to obtain the formula (4)

P＝Q+nJ_n(4)

In formula (4), P ═ ln (Y)_i) And Q ═ ln (m'). In the high current density interval, the parameters n and Q of the formula (4) can be identified by a linear least square method, so as to obtain the parameters m 'and n of the concentration polarization characteristic term, wherein m' is exp (Q).

By applying the piecewise linear parameter identification method of the formula (2) and the formula (4), the interval segmentation is carried out on the polarization curve, so that the segmentation points J of the ohmic polarization region and the concentration polarization region are estimated_f。

The relation between the voltage and the current density of the ohmic polarization area presents approximate linear characteristic, and a data point of an ohmic polarization area can be extracted according to the principle that the connecting line between any adjacent points on a straight line is equal to the X-axis included angle theta, and the current density of the last point of the ohmic polarization area, namely J_f. The approximate linearity of the ohmic polarization region can lead the included angle theta to generate certain fluctuation, and the included angle theta can tend to be stable by adopting median filtering, so that the threshold range can be conveniently determined.

3 data points near the current density range 1/2 were selected and the average calculated

And then +/-delta theta is taken as a threshold range for judging the ohmic polarization region theta, and data points positioned in the range are considered as data points of the ohmic polarization region. The specific estimation method is shown in formula (5).

The function of multiplying the coefficient mu before the act (eta) conversion is performed, so that the coordinate of the current density in the polarization curve is compressed, and the data of the ohm polarization interval can be extracted.

The invention has the beneficial effects that: the invention provides a fuel cell polarization curve segmentation identification algorithm for a single chip microcomputer, which does not need to set parameter initial values, occupies less computing resources, has the precision equivalent to that of a Newton method, and is very suitable for the environment of the single chip microcomputer.

Drawings

Fig. 1 is a schematic flow chart of a fuel cell polarization curve segmentation identification algorithm applicable to a single chip microcomputer according to the present invention. .

Detailed Description

The invention will be further described with reference to the accompanying drawings.

As shown in FIG. 1, the invention relates to a fuel cell polarization curve segmentation identification algorithm for a single chip microcomputer, which converts the polarization curve nonlinear parameter identification of a fuel cell into two linear least square identification subproblems by using an interval segmentation idea.

The algorithm flow of the invention is as follows: first, linear segments of the polarization curve are extracted, such as the low current density interval and the high current density interval described above. After extraction, the theta angle is calculated through a formula, and a median value is filtered to determine a threshold value range. After the threshold range is determined, the data point of the polarization interval and the current density of the last point of the interval can be extracted as J_f。

After the data are obtained, J is not more than J and can be taken according to the interval segmentation idea_fInterval obtained as E_eU of A and R_iData identification parameters, take J > J_fInterval obtains U of m' and n_iThe data identifies a parameter.

Finally, the parameters identified by the algorithm are used for fitting to obtain a needed polarization curve of the fuel cell.

Claims

1. A fuel cell polarization curve segmentation identification algorithm for a singlechip is firstly obtained by an interval segmentation idea based on the following digital model (formula (1)),

U_i＝E_e-A ln(J_i)-RJ_i-m′exp(nJ_i) (1)

in the formula (1), the reaction mixture is,E_e＝E₀+A ln(j₀)-RJ_n，m′＝mexp(nJ_n)，

there are 5 parameters to be identified in formula (1), which are: e_eA, R, m' and n, the model is a non-linear model, the parameters of the model are usually estimated by a non-linear least square method (such as Newton method, etc.), but the inversion of a high-dimensional matrix and the setting of initial values are required, the robustness is poor, the calculation amount is large,

will-A ln (J)_i) Corresponding to the activated polarization region;

will-RJ_iCorresponding to the ohmic polarization region;

mixing-m' exp (nJ)_i) Corresponding to the concentration polarization zone,

wherein, the activation polarization characteristic term and the ohm polarization characteristic term represent a low current density interval, the concentration polarization characteristic term represents a high current density interval,

low current density interval:

U_i＝E_e-A ln(J_i)-RJ_i(2)

passing equation (1) through the data pair parameter E for the active polarization region and the ohmic polarization region_eAnd A and R are subjected to linear least square identification to obtain a formula (3)

Y_i＝m′exp(nJ_i) (3)

In the formula (3), Y_i＝-[U_i-E_e+A ln(J_i)+RJ_i]Taking natural logarithm from two sides of the formula (3) to obtain the formula (4)

P＝Q+nJ_n(4)

In formula (4), P ═ ln (Y)_i) Q ═ ln (m '), in the high current density interval, the parameter n, Q of the formula (4) can be identified by the linear least square method, so as to obtain the parameters m' and n of the concentration polarization characteristic term, wherein m ═ exp (Q),

by applying the piecewise linear parameter identification method of the formula (2) and the formula (4), the interval segmentation is carried out on the polarization curve, so that the segmentation points J of the ohmic polarization region and the concentration polarization region are estimated_f，

The relation between the voltage and the current density of the ohmic polarization area presents approximate linear characteristic, and a data point of an ohmic polarization area can be extracted according to the principle that the connecting line between any adjacent points on a straight line is equal to the X-axis included angle theta, and the current density of the last point of the ohmic polarization area, namely J_fThe approximate linearity of the ohmic polarization area can lead the included angle theta to generate certain fluctuation, the included angle theta tends to be stable by adopting median filtering, the threshold range is convenient to determine,

Then +/-delta theta is taken as a threshold range for judging the ohmic polarization region theta, and the data point positioned in the range, namely the data point of the ohmic polarization region, is considered as a concrete estimation method shown in formula (5),

the method is characterized in that the coefficient mu is multiplied before actan conversion is carried out, so that the coordinate of current density in a polarization curve is compressed, the data of an ohm polarization interval are extracted, and the algorithm converts the polarization curve nonlinear parameter identification of the fuel cell into two linear least square identification subproblems by utilizing an interval segmentation concept.