CN108197392A - The assist characteristic curve design method of automobile electric booster steering system based on SOC - Google Patents

Abstract

The invention discloses a kind of assist characteristic curve design methods of the automobile electric booster steering system based on SOC, consider the situation of battery SOC, SOC is divided into three classes, respectively SOC<40%th, 40%<SOC<70% and SOC>70%；To SOC<40% situation carries out segment processing, to 40% corresponding to shaped form assist characteristic curve<SOC<70% situation carries out double smoothing processing；To SOC>70% situation carries out single exponential smoothing processing.The present invention rationally designs assist characteristic curve by considering on the basis of electric boosted automobile batteries energy management, power-assisted steering performance can be obviously improved, it avoids since assist motor power supply is discontinuous caused by the fast or big hand torque operation of the variation of hand-power square, and then cause and turn to difficulty etc., ensure that battery can generate stable electric energy output in the various electricity stages and make to stablize by the power steering that assist characteristic curve calculates, improve power-assisted ability to the greatest extent while ensureing that power-assisted stablizes safe, it is energy saving.

Description

The assist characteristic curve design method of automobile electric booster steering system based on SOC

Technical field

The invention belongs to automobile technical fields, are related to electric boosting steering system and Energy Management System, especially one Kind considers the assist characteristic Curve Design in the case of battery SOC.

Background technology

The assist characteristic curve of EPS refers to the power-assisted electric current of motor and steering wheel input torque, Vehicle Speed, preceding Take turns the relationship of air pressure, front shaft again etc. between parameters, consider under normal circumstances steering wheel input torque, Vehicle Speed this two A major parameter can meet control requirement.The assist characteristic curve of EPS is the key technology of EPS software systems.Assist characteristic Curve determines that controller (ECU) goes the size of control power-assisted electric current according to which type of target, goes to meet different driving cycles Under requirement to booster torquemoment.Basic assist characteristic curve is roughly divided into：Linear type, broken line type, shaped form.It can expire substantially Sufficient design requirement, but the actual effect of power-assisted is poor, then, the Curve guide impeller to assist characteristic curve occurs, wherein, there is ginseng The adjustable Curve-type Assistance Characteristics Curve Design of number carries out shape by the parameter of setting to Curve-type Assistance Characteristics curve Variation, can make curve variable at any time.It, can be according to small power-assisted area and big power-assisted area not with double-gradient smooth assist curves Independent calibration is carried out to parameter of curve with feel requirement, has the advantages that power-assisted smooth-going, parameter function is independent, is easy to debugging, energy Preferably meet expectation of the driver in terms of feel is turned to.Above design does not account for battery only from curve The influence that electricity generates vehicle booster, when hand-power square changes greatly and hand-power square is larger, battery probably can not The EPS electric currents met the requirements are provided, EPS system can power off suddenly, and steering boost system can not work moment, seriously affects and drives Sail safety.

Application No. is 200810202069.7 Chinese patents, and disclosing " has the individually adjusted shaped form power-assisted of parameter The electric boosting steering system of characteristic "；There are parameter adjustments to make the Curve Design period long for it, meanwhile, the reality of vehicle can not be contacted Border situation, so as to carry out curve adjustment according to the actual conditions of vehicle.

Application No. is 201010600546.2 Chinese patents, and disclosing " has the electronic of double-gradient smooth assist curves Servo steering system, " it is segmented the reality that can not ensure power-assisted curve during carrying out diclinic counting smooth processing there are broken line Whether application effect can guarantee sustainable supply of battery capacity etc..

Invention content

In order to overcome the above-mentioned deficiencies of the prior art, the present invention provides a kind of automobile electric power-assisted steerings based on SOC The assist characteristic curve design method of system.

A kind of assist characteristic curve design method of the automobile electric booster steering system based on SOC, the assist characteristic Curve design consider battery SOC situation, during Curve Design on the basis of virgin curve type assist characteristic curve into Row smoothing processing, while constrained to electric current needed for EPS system to ensure battery even if can in the case of low battery There is steady and sustained electric current to export, ensure the realization of the entire power-assisted processes of EPS, Curve Design includes the following steps：

Step 1) estimates the SOC of battery with kalman filtering theory；SOC is divided into three classes, respectively SOC< 40%th, 40%<SOC<70% and SOC>70%；

Step 2) is to SOC<40% situation carries out segment processing corresponding to shaped form assist characteristic curve, is divided into 5 sections 5 areas, respectively：Power-assisted area 1 is power-assisted dead zone, and power-assisted area 2 is small power-assisted area, and power-assisted area 3 is middle power-assisted area, and power-assisted area 4 is Big power-assisted area, power-assisted area 5 is saturation region；Curve smoothing processing is carried out respectively to 5 sections of assist characteristic curve, and power-assisted area 2 is hand Torque T₁<T_d<T₂, handled using single exponential smoothing；Middle Resistance Level 3 is T₂<T_d<T₃, handled using double smoothing；It helps greatly Power area 4 is T₃<T_d<T₄, handled using Three-exponential Smoothing, improve the change rate of shaped form power-assisted curve, so as to stablize battery Electric current exports；

Step 3) is to 40%<SOC<70% situation carries out double smoothing processing to Curve-type Assistance Characteristics curve；

Step 4) is to SOC>70% situation carries out single exponential smoothing processing to Curve-type Assistance Characteristics curve；

During step 5) is to utilizing hand Calculating Torque during Rotary target current, i.e. I_a=ζ T_d, proportionate relationship ζ therein is changed Processing,

Wherein, λ is transformation coefficient, determines the corresponding power-assisted curve of speed；Wherein, I_aFor target current, T_dFor hand-power square, ζ For hand-power square and the conversion relation of target current；

Step 6) is under different speeds, by obtaining the complete assist characteristic under different speeds about the continuous item of speed Curve, i.e.,

I=e^-av·ζT_d (2)

Wherein I be each vehicle speed condition under target current, e^-avFor the compensation coefficient comprising speed, a is constant.

This patent considers the remaining capacity of battery in Curve Design, meanwhile, the design of curve is helped in virgin curve type On the basis of force characteristic curve, in the case where meeting practical application, the calculating of target current can be preferably realized, ensure real Border follow current stablizes output；During power-assisted curve is designed on the one hand be on the basis of shaped form power-assisted curve into Capable smoothing processing, effect are more good；On the other hand, this patent makes the design carried out on the basis of battery capacity is considered, It can ensure the sustainable supply of battery current.

The assist characteristic curve design method of automobile electric booster steering system based on SOC uses double card in step 1) Kalman Filtering theory makes the state of dynamical system the optimal estimation in minimum variance meaning, estimates applied to battery SOC, electricity Pond is seen as dynamical system, and SOC is an internal state of system；The general mathematical form of battery model is：

State equation：x_k+1=A_kx_k+B_ku_k+w_k=f (x_k,u_k)+w_k (3)

Observational equation：y_k=c_kx_k+v_k=g (x_k,u_k)+v_k (4)

Wherein, the input vector u of system_k, to include battery current, temperature, residual capacity and internal resistance variable；

The output y of system_k, the operating voltage for battery；

Battery SOC is included in the quantity of state x of system_kIn, f (x_k,u_k) and g (x_k,u_k) be all determined by battery model it is non- Linear equation is linearized in calculating process；SOC algorithms are estimated, including SOC estimation and reaction evaluated error, association side The recursion equation of poor matrix, covariance matrix are used for providing evaluated error range, this equation is in battery model state equation In, SOC is described as to the foundation of state vector：

Specifically double card Kalman Filtering algorithm steps are：

Step (1) gives initial value：

To enable iteration Fast Convergent, by the relatively actual value of the initial value setting of SOC, R₀Initial value then lead to It crosses current SOC and tables look-up and provide, other two state can be set as 0, thus just obtain X (0) and R (0)；

Step (2) SOC estimates：

The system mode of kth step is estimated using the systematic parameter of -1 step of kth, the system mode then walked again with kth is estimated Count the systematic parameter of kth step：

Step (2.1) is estimated using the system mode that current integration walks kth：

X (kk-1)=A_s(k)X(k-1)+B_s(k)I(k)+ω_s (6)

Wherein,

Wherein, ω_sFor the process noise of system, substantially determined by the noise of electric current；

The optimal estimation of step (2.2) system mode：

X (k)=X (k | k-1)+K_s(k)(V(k)-V(k)') (9)

Wherein V (k) is the battery both end voltage measured, and V (k) ' is the terminal voltage estimated using battery model；

V (k) '=F_v(S(k))-R₀(k)I(k)-U_RC1(k)-U_RC2(k)+v (10)

Step (2.3) seeks Ks (k) (kalman gain), calculates variance matrix：

Wherein Q_sFor the covariance matrix of systematic procedure noise, r_sThe variance of measurement noise for voltage；F_v(S) it is open circuit Function of the voltage about SOC；

Step (2.4) updates variance matrix P, terminates the Kalman filtering of the SOC of this step；

P_s(k)=[I-K_s(k)C_s(k)]P_s(k|k-1) (14)

It is special to choose shaped form power-assisted for the assist characteristic curve design method of automobile electric booster steering system based on SOC Linearity curve function is：

Wherein,F (v) is the corresponding power-assisted curvilinear system of friction speed；F (v) is vehicle The function of fast v, v is faster, and F (v) is smaller, takes the exponential decrease function of v, G (T_d) for different directions disk torque input when power-assisted Curve is increasing function, i.e. the input torque of steering wheel is bigger, light in order to turn to, and booster torquemoment also must be bigger.

The assist characteristic curve design method of automobile electric booster steering system based on SOC, used smoothing processing Or exponential smoothing, it is a kind of Time Series Analysis Forecasting method to grow up on the basis of the method for moving average, it is to pass through meter Exponential smoothing value is calculated, cooperation regular hour sequential forecasting models predict the future of phenomenon；Its principle is any phase Exponential smoothing value is all the weighted average of current period actual observation value and previous phase index smooth value, the fundamental formular of exponential smoothing It is：

S_t=α y_t+(1-α)S_t-1 (16)

In formula, S_t-- the smooth value of time t；y_t-- the actual value of time t；S_t-1-- the smooth value of time t-1；α -- it is smooth Constant, value range are [0,1]；

According to smooth number difference, exponential smoothing is divided into：Single Exponential Smoothing, Secondary Exponential Smoothing Method and index three times Exponential smoothing；

For small power-assisted area 2, T₁<T_d<T₂, handled using single exponential smoothing, predictor formula is：

y_t+1'=ay_t+(1-a)y'_t (17)

Wherein, y_t+1' -- the smooth value S of the predicted value of t+1 phases, i.e. current period (t phases)_t ⁽¹⁾；y_t-- the actual value of t phases；y_t'-- The predicted value of t phases, i.e. last smooth value S_t-1 ⁽¹⁾；The formula can be write again：y_t+1'=y'_t+a(y_t-y'_t), it is seen then that the next period Predicted value be current period predicted value again with using a as the sum of current period actual value of discount and predicted value error；

For middle power-assisted area 3, T₂<T_d<T₃, handled using double smoothing, double smoothing is that an index is put down Sliding is smooth again, it is suitable for the time series of tool linear trend, and predictor formula is：

y_t+m=(2+a_m/(1-a))y'_t-(1+a_m/(1-a))y_t=(2y'_t-y_t)+m(y'_t-y_t)a/(1-a) (18)

In formula, y_t=ay'_t-1+(1-a)y_t-1

Obviously, double smoothing is a linear equation, and intercept is：(2y_t'-y_t), slope is：(y_t'-y_t)a/(1- A), independent variable is predicts number of days, double smoothing fundamental formular：

S_t ⁽²⁾=aS_t ⁽¹⁾+(1-a)S_t-1 ⁽²⁾ (19)

Y_t+ Y=a_t+b_tT (20)

a_t=2S_t ⁽¹⁾-S_t ⁽²⁾ (21)

b_t=(a/1-a) (S_t ⁽¹⁾-S_t ⁽²⁾) (22)

In formula, S_t ⁽¹⁾-- the single exponential smoothing value S of t phases_t ⁽²⁾-- the double smoothing value α of t phases -- smoothing factor Y_t+T-- t+T phase predicted value T-- elapse issue backward by the t phases；

For big power-assisted area 4, T₃<T_d<T₄, handled using Three-exponential Smoothing, Three-exponential Smoothing prediction is secondary smooth On the basis of it is smooth again, predictor formula is：

y_t+m=(3y'_t-3y_t)+[(6-5a)y'_t-(10-8a)y_t+(4-3a)y_t]*a_m/2(1-a)²+(y'_t-2y_t+y'_t)* a²m²/2(1-a)² (23)

In formula, y_t=ay_t-1+(1-a)y_t-1

Predicted value is the weighted sum of former observation, and different data are given with different power, and new data is to larger Power, legacy data give smaller power, by trend adjustment, add trend correction value, can improve exponential smoothing to a certain extent Prediction result, the formula of the exponential smoothing after adjustment are：

Include trend prediction (Y_tT_t)=new prediction (Y_t)+trend corrects (T_t)

There are three steps for the Smoothing Prediction of progress trend adjustment：

(1) the simple index number smoothing prediction (Y of t phases is calculated using previously described method_t)；

(2) trend is calculated, formula is：

T_t=(1-b) T_t-1+b(Y_t-Y_t-1) (24)

Wherein, T_t=the t phases smoothed trend；T_t-1=the t phases last smoothed trend；B=selections become Gesture smoothing factor；Y_t=to t phase simple index number smoothing predictions；Y_t-1=to t phase last issue simple index number smoothing predictions,

(3) the Smoothing Prediction value (Y after trend adjustment is calculated_tT_t), calculation formula is：

Y_tT_t=Y_t+T_t。 (25)

The technical solution adopted by the present invention is when being designed to assist characteristic curve, it is contemplated that the situation of SOC, especially It is to be changed greatly in hand-power square and greatly in the case of hand-power square, change the change rate of power-assisted curve, control the electric current of EPS in electricity Within the supply district of pond.

SOC numerical value is observed in real time by Kalman Filter observer, according to SOC numerical values reciteds to virgin curve type assist characteristic Curve carries out a point situation and handles, to SOC<40 situation is segmented virgin curve type assist characteristic curve smoothly to be located respectively Reason, to SOC>40% situation with regard to carrying out level-one respectively, two level exponential smoothing handles, while the demand current of EPS is carried out about Beam ensures lasting supply of the battery to EPS electric currents, so as to be that servo steering system persistently plays a role.

Compared with prior art, the beneficial effects of the invention are as follows battery electricity is considered when assist characteristic curve is designed The situation of SOC is measured, steering boost system can be made to be run in the case of any battery capacity SOC, ensure the steering of driver Safety.

Description of the drawings

Fig. 1 is the assist characteristic curve of the prior art.

Fig. 2 is SOC of the present invention<Assist characteristic curve in the case of 40%.

Fig. 3 is the present invention 40%<SOC<Assist characteristic curve in the case of 70%.

Fig. 4 is SOC of the present invention>Assist characteristic curve in the case of 70%.

Fig. 5 is the assist characteristic Curve Design flow chart the present invention is based on SOC.

Specific embodiment

Below in conjunction with attached drawing to further instruction of the present invention.

Fig. 1 is engaged it is found that the design of the prior art does not account for battery capacity for vehicle only from curve The influence that power-assisted generates, when hand-power square changes greatly and hand-power square is larger, battery can not probably be provided and be met the requirements EPS electric currents, EPS system can power off suddenly, and steering boost system can not work moment, seriously affect driving safety.

Fig. 2-Fig. 4 is engaged to understand；A kind of assist characteristic of automobile electric booster steering system based on SOC of the present invention is bent Line design method, the design of the assist characteristic curve considers the situation of battery SOC, in virgin curve type during Curve Design It is smoothed on the basis of assist characteristic curve, even if while being constrained to ensure battery to electric current needed for EPS system Can also there be steady and sustained electric current to export in the case of low battery, ensure the realization of the entire power-assisted processes of EPS, Curve Design Include the following steps：

Step 1) estimates the SOC of battery with kalman filtering theory；SOC is divided into three classes, respectively SOC< 40%th, 40%<SOC<70% and SOC>70%；

Step 2) is to SOC<40% situation carries out segment processing corresponding to shaped form assist characteristic curve, is divided into 5 sections 5 areas, respectively：Power-assisted area 1 is power-assisted dead zone, and power-assisted area 2 is small power-assisted area, and power-assisted area 3 is middle power-assisted area, and power-assisted area 4 is Big power-assisted area, power-assisted area 5 is saturation region；Curve smoothing processing is carried out respectively to 5 sections of assist characteristic curve, and power-assisted area 2 is hand Torque T₁<T_d<T₂, handled using single exponential smoothing；Middle Resistance Level 3 is T₂<T_d<T₃, handled using double smoothing；It helps greatly Power area 4 is T₃<T_d<T₄, handled using Three-exponential Smoothing, improve the change rate of shaped form power-assisted curve, so as to stablize battery Electric current exports；For small power-assisted area 2, T₁<T_d<T₂, handled using single exponential smoothing.In small 2 sections of power-assisted area, n point is chosen a₁, a₂, a₃..., obtain a series of point (x₁, y₁), (x₂, y₂)(x₃, y₃) ... it is T to take initial value₁, by taken each point successively It substitutes into：y_t+1'=y'_t+a(y_t-y'_t).Meanwhile it is α to take smoothing factor, damped coefficient μ simultaneously changes damped coefficient, changes successively For operation until obtaining final result.

For middle power-assisted area 3, T₂<T_d<T₃, handled using double smoothing.In middle 3 sections of power-assisted area, n point a are chosen₁, a₂, a₃..., obtain a series of point (x₁, y₁), (x₂, y₂)(x₃, y₃) ... it is T to take initial value₂, by taken each point successively generation Enter：y_t+m=(2+a_m/(1-a))y'_t-(1+a_m/(1-a))y_t=(2y'_t-y_t)+m(y'_t-y_t) a/ (1-a) simultaneously, sliding system of making even Number is α, and damped coefficient for μ and changes damped coefficient with this, is iterated operation until obtaining final result.

For big power-assisted area 4, T₃<T_d<T₄, handled using Three-exponential Smoothing.In middle 3 sections of power-assisted area, n point a are chosen₁, a₂, a₃..., obtain a series of point (x₁, y₁), (x₂, y₂)(x₃, y₃) ... it is T to take initial value₂, by taken each point successively generation Enter：y_t+m=(3y'_t-3y_t)+[(6-5a)y'_t-(10-8a)y_t+(4-3a)y_t]*a_m/2(1-a)²+(y'_t-2y_t+y'_t)*a²m²/2 (1-a)²Meanwhile it is α to choose smoothing factor, damped coefficient is μ and changes damped coefficient successively, is iterated operation until obtaining Final result.

Step 3) is to 40%<SOC<70% situation carries out double smoothing processing to Curve-type Assistance Characteristics curve；

To 40%<SOC<70% situation carries out single exponential smoothing processing to virgin curve type power-assisted curve, takes n point a₁, a₂, a₃..., obtain a series of point (x₁, y₁), (x₂, y₂)(x₃, y₃) ... it is T to take initial value₁, by taken each point successively It substitutes into：y_t+m=(2+a_m/(1-a))y'_t-(1+a_m/(1-a))y_t=(2y'_t-y_t)+m(y'_t-y_t) a/ (1-a) simultaneously, cunning of making even Coefficient is α, and damped coefficient for μ and changes damped coefficient with this, is iterated operation, until obtaining final result.

Step 4) is to SOC>70% situation carries out single exponential smoothing processing to Curve-type Assistance Characteristics curve；

To SOC>70% situation carries out single exponential smoothing processing to virgin curve type power-assisted curve, takes n point a₁, a₂, a₃..., obtain a series of point (x₁, y₁), (x₂, y₂)(x₃, y₃) ... it is T to take initial value₁, taken each point is substituted into successively： y_t+1'=y'_t+a(y_t-y'_t) simultaneously, taking smoothing factor, damped coefficient for μ and changes damped coefficient with this, is iterated fortune for α It calculates, until obtaining final result.

During step 5) is to utilizing hand Calculating Torque during Rotary target current, i.e. I_a=ζ T_d, proportionate relationship ζ therein is changed Processing,

Wherein, λ is transformation coefficient, determines the corresponding power-assisted curve of speed；Wherein, I_aFor target current, T_dFor hand-power square, ζ For hand-power square and the conversion relation of target current；When the hand-power square size for measuring torque sensor is converted into target current, By change conversion coefficient can gentle assist characteristic curve, target current value is made to remain at the discharge current range of battery It is interior, i.e.,Curve can be made to generate the effect of horizontal elongation.

Step 6) is under different speeds, by obtaining the complete assist characteristic under different speeds about the continuous item of speed Curve, i.e.,

I=e^-av·ζT_d (27)

Wherein I be each vehicle speed condition under target current, e^-avFor the compensation coefficient comprising speed, a is constant.

This patent considers the remaining capacity of battery in Curve Design, meanwhile, the design of curve is helped in virgin curve type On the basis of force characteristic curve, in the case where meeting practical application, the calculating of target current can be preferably realized, ensure real Border follow current stablizes output；During power-assisted curve is designed on the one hand be on the basis of shaped form power-assisted curve into Capable smoothing processing, effect are more good；On the other hand, this patent makes the design carried out on the basis of battery capacity is considered, It can ensure the sustainable supply of battery current.

The assist characteristic curve design method of automobile electric booster steering system based on SOC uses double card in step 1) Kalman Filtering theory makes the state of dynamical system the optimal estimation in minimum variance meaning, estimates applied to battery SOC, electricity Pond is seen as dynamical system, and SOC is an internal state of system；The general mathematical form of battery model is：

State equation：x_k+1=A_kx_k+B_ku_k+w_k=f (x_k,u_k)+w_k (28)

Observational equation：y_k=c_kx_k+v_k=g (x_k,u_k)+v_k (29)

Wherein, the input vector u of system_k, to include battery current, temperature, residual capacity and internal resistance variable；

The output y of system_k, the operating voltage for battery；

Battery SOC is included in the quantity of state x of system_kIn, f (x_k,u_k) and g (x_k,u_k) be all determined by battery model it is non- Linear equation is linearized in calculating process；SOC algorithms are estimated, including SOC estimation and reaction evaluated error, association side The recursion equation of poor matrix, covariance matrix are used for providing evaluated error range, this equation is in battery model state equation In, SOC is described as to the foundation of state vector：

Specifically double card Kalman Filtering algorithm steps are：

Step (1) gives initial value：

To enable iteration Fast Convergent, by the relatively actual value of the initial value setting of SOC, R₀Initial value then lead to It crosses current SOC and tables look-up and provide, other two state can be set as 0, thus just obtain X (0) and R (0)；

Step (2) SOC estimates：

The system mode of kth step is estimated using the systematic parameter of -1 step of kth, the system mode then walked again with kth is estimated Count the systematic parameter of kth step：

Step (2.1) is estimated using the system mode that current integration walks kth：

X (k | k-1)=A_s(k)X(k-1)+B_s(k)I(k)+ω_s (31)

Wherein,

Wherein, ω_sFor the process noise of system, substantially determined by the noise of electric current；

The optimal estimation of step (2.2) system mode：

X (k)=X (k | k-1)+K_s(k)(V(k)-V(k)') (34)

Wherein V (k) is the battery both end voltage measured, and V (k) ' is the terminal voltage estimated using battery model；

V (k) '=F_v(S(k))-R₀(k)I(k)-U_RC1(k)-U_RC2(k)+v (35)

Step (2.3) seeks Ks (k) (kalman gain), calculates variance matrix：

Wherein Q_sFor the covariance matrix of systematic procedure noise, r_sThe variance of measurement noise for voltage；F_v(S) it is open circuit Function of the voltage about SOC；

Step (2.4) updates variance matrix P, terminates the Kalman filtering of the SOC of this step；

P_s(k)=[I-K_s(k)C_s(k)]P_s(k|k-1) (39)

It is special to choose shaped form power-assisted for the assist characteristic curve design method of automobile electric booster steering system based on SOC Linearity curve function is：

Wherein,F (v) is the corresponding power-assisted curvilinear system of friction speed；F(v) For the function of speed v, v is faster, and F (v) is smaller, takes the exponential decrease function of v, G (T_d) it is when different directions disk torque inputs Power-assisted curve is increasing function, i.e. the input torque of steering wheel is bigger, light in order to turn to, and booster torquemoment also must be bigger.

The assist characteristic curve design method of automobile electric booster steering system based on SOC, used smoothing processing Or exponential smoothing, it is a kind of Time Series Analysis Forecasting method to grow up on the basis of the method for moving average, it is to pass through meter Exponential smoothing value is calculated, cooperation regular hour sequential forecasting models predict the future of phenomenon；Its principle is any phase Exponential smoothing value is all the weighted average of current period actual observation value and previous phase index smooth value, the fundamental formular of exponential smoothing It is：

S_t=α y_t+(1-α)S_t-1 (41)

In formula, S_t-- the smooth value of time t；y_t-- the actual value of time t；S_t-1-- the smooth value of time t-1；α -- it is smooth Constant, value range are [0,1]；

According to smooth number difference, exponential smoothing is divided into：Single Exponential Smoothing, Secondary Exponential Smoothing Method and index three times Exponential smoothing；

For SOC>70% situation and SOC<Small power-assisted area 2 in the case of 40%, i.e. T₁<T_d<T₂When.Using one Secondary exponential smoothing processing, predictor formula are：

y_t+1'=ay_t+(1-a)y'_t (42)

Wherein, y_t+1' -- the smooth value S of the predicted value of t+1 phases, i.e. current period (t phases)_t ⁽¹⁾；y_t-- the actual value of t phases；y_t'-- The predicted value of t phases, i.e. last smooth value S_t-1 ⁽¹⁾；The formula can be write again：y_t+1'=y_t'+a(y_t-y_t'), it is seen then that the next period Predicted value be current period predicted value again with using a as the sum of current period actual value of discount and predicted value error；

For 40%<SOC<70% situation and SOC<Middle power-assisted area 3 in the case of 40%, i.e. T₂<T_d<T₃When.It should It is handled with double smoothing, double smoothing is to the smooth again of single exponential smoothing, it is suitable for tool linear trend Time series, predictor formula are：

y_t+m=(2+a_m/(1-a))y'_t-(1+a_m/(1-a))y_t=(2y'_t-y_t)+m(y'_t-y_t)a/(1-a) (43)

In formula, y_t=ay'_t-1+(1-a)y_t-1

Obviously, double smoothing is a linear equation, and intercept is：(2y'_t-y_t), slope is：(y_t'-y_t)a/(1- A), independent variable is predicts number of days, double smoothing fundamental formular：

S_t ⁽²⁾=aS_t ⁽¹⁾+(1-a)S_t-1 ⁽²⁾ (44)

Y_t+ Y=a_t+b_tT (45)

a_t=2S_t ⁽¹⁾-S_t ⁽²⁾ (46)

b_t=(a/1-a) (S_t ⁽¹⁾-S_t ⁽²⁾) (47)

In formula, S_t ⁽¹⁾-- the single exponential smoothing value S of t phases_t ⁽²⁾-- the double smoothing value α of t phases -- smoothing factor Y_t+T-- t+T phase predicted value T-- elapse issue backward by the t phases；

For SOC<Big power-assisted area 4 in 40% situation, i.e. T₃<T_d<T₄When, it handles using Three-exponential Smoothing, refers to three times Number smoothing predictions be it is secondary it is smooth on the basis of it is smooth again, predictor formula is：

y_t+m=(3y'_t-3y_t)+[(6-5a)y'_t-(10-8a)y_t+(4-3a)y_t]*a_m/2(1-a)²+(y'_t-2y_t+y'_t)* a²m²/2(1-a)² (48)

In formula, y_t=ay_t-1+(1-a)y_t-1

Predicted value is the weighted sum of former observation, and different data are given with different power, and new data is to larger Power, legacy data give smaller power, by trend adjustment, add trend correction value, can improve exponential smoothing to a certain extent Prediction result, the formula of the exponential smoothing after adjustment are：

Include trend prediction (Y_tT_t)=new prediction (Y_t)+trend corrects (T_t)

There are three steps for the Smoothing Prediction of progress trend adjustment：

(1) the simple index number smoothing prediction (Y of t phases is calculated using previously described method_t)；

(2) trend is calculated, formula is：

T_t=(1-b) T_t-1+b(Y_t-Y_t-1) (49)

Wherein, T_t=the t phases smoothed trend；T_t-1=the t phases last smoothed trend；B=selections become Gesture smoothing factor；Y_t=to t phase simple index number smoothing predictions；Y_t-1=to t phase last issue simple index number smoothing predictions,

(3) the Smoothing Prediction value (Y after trend adjustment is calculated_tT_t), calculation formula is：

Y_tT_t=Y_t+T_t。 (50)

Engage Fig. 5 it is found that the assist characteristic curve design method of the automobile electric booster steering system based on SOC just It is the situation that battery SOC is considered in 1 design process of the assist characteristic curve, to the virgin curve type in the case of different SOC Assist characteristic curve carries out exponential smoothing processing, even if while being constrained to ensure battery low to electric current needed for EPS system Also there can be steady and sustained electric current to export in the case of electricity, ensure the realization of the entire power-assisted processes of EPS.

The design method process mainly includes：The prediction of battery SOC, carried out according to the value of battery SOC classification smoothing processing, Processing is changed using hand Calculating Torque during Rotary target current process comparative example relationship ζ, the curve under full speed is handled.Its In, the prediction of battery SOC ensures the precision of prediction using double Kalman filtering algorithms；It is flat that classification is carried out according to the value of battery SOC Sliding processing, considers SOC respectively<40%, 40%<SOC<70%, SOC>70% 3 kind of situation, to virgin curve type assist characteristic curve It is smoothed respectively.

The technical solution adopted by the present invention is when being designed to assist characteristic curve, it is contemplated that the situation of SOC, especially It is to be changed greatly in hand-power square and greatly in the case of hand-power square, change the change rate of power-assisted curve, control the electric current of EPS in electricity Within the supply district of pond.

SOC numerical value is observed in real time by Kalman Filter observer, according to SOC numerical values reciteds to virgin curve type assist characteristic Curve carries out a point situation and handles, to SOC<40 situation is segmented virgin curve type assist characteristic curve smoothly to be located respectively Reason, to SOC>40%

Situation with regard to carrying out level-one respectively, two level exponential smoothing is handled, while the demand current of EPS is constrained, protect Lasting supply of the battery to EPS electric currents is demonstrate,proved, so as to be that servo steering system persistently plays a role.

Compared with prior art, the beneficial effects of the invention are as follows battery electricity is considered when assist characteristic curve is designed The situation of SOC is measured, steering boost system can be made to be run in the case of any battery capacity SOC, ensure the steering of driver Safety.

Claims (4)

1. a kind of assist characteristic curve design method of the automobile electric booster steering system based on SOC, it is characterised in that：It is described The design of assist characteristic curve considers the situation of battery SOC, in virgin curve type assist characteristic curve during Curve Design On the basis of be smoothed, even if while being constrained situation to ensure battery in low battery to electric current needed for EPS system Under can also there is steady and sustained electric current to export, ensure the realization of the entire power-assisted processes of EPS, Curve Design includes the following steps：

Step 1) estimates the SOC of battery with kalman filtering theory；SOC is divided into three classes, respectively SOC<40%th, 40%<SOC<70% and SOC>70%；

Step 2) is to SOC<40% situation carries out segment processing corresponding to shaped form assist characteristic curve, is divided into 5 sections 5 Area, respectively：Power-assisted area 1 is power-assisted dead zone, and power-assisted area 2 is small power-assisted area, and power-assisted area 3 is middle power-assisted area, and power-assisted area 4 is to help greatly Power area, power-assisted area 5 are saturation regions；Curve smoothing processing is carried out respectively to 5 sections of assist characteristic curve, and power-assisted area 2 is hand-power square T₁<T_d<T₂, handled using single exponential smoothing；Middle Resistance Level 3 is T₂<T_d<T₃, handled using double smoothing；Big power-assisted area 4 i.e. T₃<T_d<T₄, handled using Three-exponential Smoothing, improve the change rate of shaped form power-assisted curve, so as to stablize the electric current of battery Output；

Step 3) is to 40%<SOC<70% situation carries out double smoothing processing to Curve-type Assistance Characteristics curve；

Step 4) is to SOC>70% situation carries out single exponential smoothing processing to Curve-type Assistance Characteristics curve；

During step 5) is to utilizing hand Calculating Torque during Rotary target current, i.e. I_a=ζ T_d, proportionate relationship ζ therein is changed place Reason,

Wherein, λ is transformation coefficient, determines the corresponding power-assisted curve of speed；Wherein, I_aFor target current, T_dFor hand-power square, ζ is hand The conversion relation of torque and target current；

Step 6) is under different speeds, by obtaining the song of the complete assist characteristic under different speeds about the continuous item of speed Line, i.e.,

I=e^-av·ζT_d (2)

Wherein I be each vehicle speed condition under target current, e^-avFor the compensation coefficient comprising speed, a is constant.

2. the assist characteristic curve design method of the automobile electric booster steering system based on SOC according to claim 1, It is characterized in that：With pair kalman filtering theories the state of dynamical system is made in step 1) optimal in minimum variance meaning Estimation estimates that battery is seen as dynamical system, and SOC is an internal state of system applied to battery SOC；Battery model General mathematical form is：

State equation：x_k+1=A_kx_k+B_ku_k+w_k=f (x_k,u_k)+w_k (3)

Observational equation：y_k=c_kx_k+v_k=g (x_k,u_k)+v_k (4)

Wherein, the input vector u of system_k, to include battery current, temperature, residual capacity and internal resistance variable；

The output y of system_k, the operating voltage for battery；

Battery SOC is included in the quantity of state x of system_kIn, f (x_k,u_k) and g (x_k,u_k) be all determined by battery model it is non-linear Equation is linearized in calculating process；SOC algorithms are estimated, including SOC estimation and reaction evaluated error, covariance square The recursion equation of battle array, covariance matrix are used for providing evaluated error range, this equation is in battery model state equation, is incited somebody to action SOC is described as the foundation of state vector：

Specifically double card Kalman Filtering algorithm steps are：

Step (1) gives initial value：

To enable iteration Fast Convergent, by the relatively actual value of the initial value setting of SOC, R₀Initial value then by work as Preceding SOC, which tables look-up, to be provided, other two state can be set as 0, has thus just obtained X (0) and R (0)；

Step (2) SOC estimates：

The system mode of kth step is estimated using the systematic parameter of -1 step of kth, the system state estimation kth then walked again with kth The systematic parameter of step：

Step (2.1) is estimated using the system mode that current integration walks kth：

X (k | k-1)=A_s(k)X(k-1)+B_s(k)I(k)+ω_s (6)

Wherein,

Wherein, ω_sFor the process noise of system, substantially determined by the noise of electric current；

The optimal estimation of step (2.2) system mode：

X (k)=X (k | k-1)+K_s(k)(V(k)-V(k)') (9)

Wherein V (k) is the battery both end voltage measured, and V (k) ' is the terminal voltage estimated using battery model；

V (k) '=F_v(S(k))-R₀(k)I(k)-U_RC1(k)-U_RC2(k)+v (10)

Step (2.3) seeks Ks (k) (kalman gain), calculates variance matrix：

Wherein Q_sFor the covariance matrix of systematic procedure noise, r_sThe variance of measurement noise for voltage；F_v(S) it is open-circuit voltage Function about SOC；

Step (2.4) updates variance matrix P, terminates the Kalman filtering of the SOC of this step；

P_s(k)=[I-K_s(k)C_s(k)]P_s(k|k-1) (14) 。

3. the assist characteristic curve design method of the automobile electric booster steering system according to claim 1 based on SOC, It is characterized in that：Choosing shaped form assist characteristic curvilinear function is：

Wherein,G(T_d)=b₁T_d ²+b₂T_d+B₃；F (v) is the corresponding power-assisted curvilinear system of friction speed；F (v) is speed v Function, v is faster, and F (v) is smaller, takes the exponential decrease function of v, G (T_d) for the input of different directions disk torque when power-assisted it is bent Line is increasing function, i.e. the input torque of steering wheel is bigger, light in order to turn to, and booster torquemoment also must be bigger.

4. the assist characteristic curve design method of the automobile electric booster steering system based on SOC according to claim 1, It is characterized in that：

Smoothing processing or exponential smoothing are a kind of Time Series Analysis Forecastings to grow up on the basis of the method for moving average Method, it is by gauge index smooth value, and cooperation regular hour sequential forecasting models predict the future of phenomenon；It is former It is all the weighted average of current period actual observation value and previous phase index smooth value that reason, which is the exponential smoothing value of any phase, exponential smoothing The fundamental formular of method is：

S_t=α y_t+(1-α)S_t-1 (16)

In formula, S_t-- the smooth value of time t；y_t-- the actual value of time t；S_t-1-- the smooth value of time t-1；α -- smoothing constant, Its value range is [0,1]；

According to smooth number difference, exponential smoothing is divided into：Single Exponential Smoothing, Secondary Exponential Smoothing Method and Three-exponential Smoothing Method；

For small power-assisted area 2, T₁<T_d<T₂, handled using single exponential smoothing, predictor formula is：

y_t+1'=ay_t+(1-a)y_t' (17)

Wherein, y_t+1' -- the smooth value S of the predicted value of t+1 phases, i.e. current period (t phases)_t ⁽¹⁾；y_t-- the actual value of t phases；y_t' -- the t phases Predicted value, i.e., last smooth value S_t-1 ⁽¹⁾；The formula can be write again：y_t+1'=y_t'+a(y_t-y_t'), it is seen then that the next period is pre- Measured value be current period predicted value again with using a as the sum of current period actual value of discount and predicted value error；

For middle power-assisted area 3, T₂<T_d<T₃, handled using double smoothing, double smoothing is to single exponential smoothing Smooth again, it is suitable for the time series of tool linear trend, and predictor formula is：

y_t+m=(2+a_m/(1-a))y_t'-(1+a_m/(1-a))y_t=(2y_t'-y_t)+m(y_t'-y_t)a/(1-a) (18)

In formula, y_t=ay_t'_-1+(1-a)y_t-1

Obviously, double smoothing is a linear equation, and intercept is：(2y_t'-y_t), slope is：(y_t'-y_t) a/ (1-a), from Variable is predicts number of days, double smoothing fundamental formular：

S_t ⁽²⁾=aS_t ⁽¹⁾+(1-a)S_t-1 ⁽²⁾ (19)

Y_t+ Y=a_t+b_tT (20)

a_t=2S_t ⁽¹⁾-S_t ⁽²⁾ (21)

b_t=(a/1-a) (S_t ⁽¹⁾-S_t ⁽²⁾) (22)

In formula, S_t ⁽¹⁾-- the single exponential smoothing value S of t phases_t ⁽²⁾-- the double smoothing value α of t phases -- smoothing factor Y_t+T-- t+T phase predicted value T-- elapse issue backward by the t phases；

For big power-assisted area 4, T₃<T_d<T₄, handled using Three-exponential Smoothing, Three-exponential Smoothing prediction is secondary smooth basis On it is smooth again, predictor formula is：

y_t+m=(3y_t'-3y_t)+[(6-5a)y_t'-(10-8a)y_t+(4-3a)y_t]*a_m/2(1-a)²+(y_t'-2y_t+y_t')*a²m²/ 2(1-a)² (23)

In formula, y_t=ay_t-1+(1-a)y_t-1

Predicted value is the weighted sum of former observation, and different data are given with different power, and new data gives larger power, old Data give smaller power, by trend adjustment, add trend correction value, can improve Smoothing Prediction knot to a certain extent Fruit, the formula of the exponential smoothing after adjustment are：

Include trend prediction (Y_tT_t)=new prediction (Y_t)+trend corrects (T_t)

There are three steps for the Smoothing Prediction of progress trend adjustment：

(1) the simple index number smoothing prediction (Y of t phases is calculated using previously described method_t)；

(2) trend is calculated, formula is：

T_t=(1-b) T_t-1+b(Y_t-Y_t-1) (24)

Wherein, T_t=the t phases smoothed trend；T_t-1=the t phases last smoothed trend；The trend of b=selections is put down Sliding coefficient；Y_t=to t phase simple index number smoothing predictions；Y_t-1=to t phase last issue simple index number smoothing predictions,

(3) the Smoothing Prediction value (Y after trend adjustment is calculated_tT_t), calculation formula is：

Y_tT_t=Y_t+T_t。 (25)。