[105] | 1 | /*! |
---|
[81] | 2 | \file |
---|
| 3 | \brief Models for synchronous electric drive using IT++ and BDM |
---|
| 4 | \author Vaclav Smidl. |
---|
| 5 | |
---|
| 6 | ----------------------------------- |
---|
| 7 | BDM++ - C++ library for Bayesian Decision Making under Uncertainty |
---|
| 8 | |
---|
| 9 | Using IT++ for numerical operations |
---|
| 10 | ----------------------------------- |
---|
| 11 | */ |
---|
| 12 | |
---|
| 13 | #include <itpp/itbase.h> |
---|
| 14 | #include <stat/libFN.h> |
---|
| 15 | #include <estim/libKF.h> |
---|
[108] | 16 | //#include <estim/libPF.h> |
---|
| 17 | #include <math/chmat.h> |
---|
[81] | 18 | |
---|
| 19 | #include "pmsm.h" |
---|
| 20 | #include "simulator.h" |
---|
[117] | 21 | #include "sim_profiles.h" |
---|
[81] | 22 | |
---|
[94] | 23 | #include <stat/loggers.h> |
---|
[81] | 24 | |
---|
| 25 | using namespace itpp; |
---|
| 26 | //!Extended Kalman filter with unknown \c Q |
---|
| 27 | |
---|
| 28 | int main() { |
---|
| 29 | // Kalman filter |
---|
| 30 | int Ndat = 90000; |
---|
| 31 | double h = 1e-6; |
---|
| 32 | int Nsimstep = 125; |
---|
| 33 | |
---|
[94] | 34 | dirfilelog L("exp/sim_var",1000); |
---|
[105] | 35 | //memlog L(Ndat); |
---|
[94] | 36 | |
---|
[81] | 37 | // SET SIMULATOR |
---|
| 38 | pmsmsim_set_parameters ( 0.28,0.003465,0.1989,0.0,4,1.5,0.04, 200., 3e-6, h ); |
---|
| 39 | double Ww = 0.0; |
---|
| 40 | vec dt ( 2 ); |
---|
| 41 | vec ut ( 2 ); |
---|
[135] | 42 | vec dut ( 2 ); |
---|
[131] | 43 | vec dit (2); |
---|
[81] | 44 | vec xtm=zeros ( 4 ); |
---|
[135] | 45 | vec xte=zeros ( 4 ); |
---|
[81] | 46 | vec xdif=zeros ( 4 ); |
---|
| 47 | vec xt ( 4 ); |
---|
[94] | 48 | vec ddif=zeros(2); |
---|
[81] | 49 | IMpmsm fxu; |
---|
| 50 | // Rs Ls dt Fmag(Ypm) kp p J Bf(Mz) |
---|
| 51 | fxu.set_parameters ( 0.28, 0.003465, Nsimstep*h, 0.1989, 1.5 ,4.0, 0.04, 0.0 ); |
---|
| 52 | OMpmsm hxu; |
---|
| 53 | mat Qt=zeros ( 4,4 ); |
---|
[94] | 54 | mat Rt=zeros ( 2,2 ); |
---|
[81] | 55 | |
---|
| 56 | // ESTIMATORS |
---|
| 57 | vec mu0= "0.0 0.0 0.0 0.0"; |
---|
| 58 | vec Qdiag ( "62 66 454 0.03" ); //zdenek: 0.01 0.01 0.0001 0.0001 |
---|
| 59 | vec Rdiag ( "100 100" ); //var(diff(xth)) = "0.034 0.034" |
---|
| 60 | mat Q =diag( Qdiag ); |
---|
| 61 | mat R =diag ( Rdiag ); |
---|
| 62 | EKFfull Efix ( rx,ry,ru ); |
---|
| 63 | Efix.set_est ( mu0, 1*eye ( 4 ) ); |
---|
| 64 | Efix.set_parameters ( &fxu,&hxu,Q,R); |
---|
| 65 | |
---|
| 66 | EKFfull Eop ( rx,ry,ru ); |
---|
| 67 | Eop.set_est ( mu0, 1*eye ( 4 ) ); |
---|
| 68 | Eop.set_parameters ( &fxu,&hxu,Q,R); |
---|
| 69 | |
---|
[94] | 70 | EKFfull Edi ( rx,ry,ru ); |
---|
| 71 | Edi.set_est ( mu0, 1*eye ( 4 ) ); |
---|
| 72 | Edi.set_parameters ( &fxu,&hxu,Q,R); |
---|
| 73 | |
---|
[81] | 74 | epdf& Efix_ep = Efix._epdf(); |
---|
| 75 | epdf& Eop_ep = Eop._epdf(); |
---|
[94] | 76 | epdf& Edi_ep = Edi._epdf(); |
---|
[105] | 77 | |
---|
[94] | 78 | //LOG |
---|
| 79 | RV rQ("10", "{Q }", "16","0"); |
---|
| 80 | RV rR("11", "{R }", "4","0"); |
---|
[105] | 81 | RV rUD("12 13 14 15", "{u_isa u_isb i_isa i_isb }", ones_i(4),zeros_i(4)); |
---|
[131] | 82 | RV rDu("16 17 18 19","{dux duy duxf duyf }",ones_i(4),zeros_i(4)); |
---|
| 83 | RV rDi("20 21","{disa disb }",ones_i(2),zeros_i(2)); |
---|
[94] | 84 | int X_log = L.add(rx,"X"); |
---|
| 85 | int Efix_log = L.add(rx,"XF"); |
---|
| 86 | int Eop_log = L.add(rx,"XO"); |
---|
| 87 | int Edi_log = L.add(rx,"XD"); |
---|
[105] | 88 | int Q_log = L.add(rQ,"Q"); |
---|
| 89 | int R_log = L.add(rR,"R"); |
---|
| 90 | int D_log = L.add(rUD,"D"); |
---|
[131] | 91 | int Du_log = L.add(rDu,"Du"); |
---|
| 92 | int Di_log = L.add(rDi,"Di"); |
---|
[94] | 93 | L.init(); |
---|
[81] | 94 | |
---|
| 95 | for ( int tK=1;tK<Ndat;tK++ ) { |
---|
| 96 | //Number of steps of a simulator for one step of Kalman |
---|
| 97 | for ( int ii=0; ii<Nsimstep;ii++ ) { |
---|
[135] | 98 | sim_profile_steps1 ( Ww , false); |
---|
[81] | 99 | pmsmsim_step ( Ww ); |
---|
| 100 | }; |
---|
| 101 | // simulation via deterministic model |
---|
| 102 | ut ( 0 ) = KalmanObs[0]; |
---|
| 103 | ut ( 1 ) = KalmanObs[1]; |
---|
| 104 | dt ( 0 ) = KalmanObs[2]; |
---|
| 105 | dt ( 1 ) = KalmanObs[3]; |
---|
[131] | 106 | dut ( 0 ) = KalmanObs[4]; |
---|
| 107 | dut ( 1 ) = KalmanObs[5]; |
---|
| 108 | dit ( 0 ) = KalmanObs[8]; |
---|
| 109 | dit ( 1 ) = KalmanObs[9]; |
---|
[81] | 110 | |
---|
[135] | 111 | xte = fxu.eval ( xtm,ut ); |
---|
[81] | 112 | //Results: |
---|
[135] | 113 | // in xt we have simulation according to the model |
---|
[81] | 114 | // in x we have "reality" |
---|
[135] | 115 | xt ( 0 ) =x[0];xt ( 1 ) =x[1];xt ( 2 ) =x[2];xt ( 3 ) =x[3]; |
---|
| 116 | xdif = xt-xte; //xtm is a copy of x[] |
---|
[131] | 117 | if (xdif(0)>3.0){ |
---|
| 118 | cout << "here" <<endl; |
---|
| 119 | } |
---|
[81] | 120 | if ( xdif ( 3 ) >pi ) xdif ( 3 )-=2*pi; |
---|
| 121 | if ( xdif ( 3 ) <-pi ) xdif ( 3 ) +=2*pi; |
---|
[94] | 122 | |
---|
[135] | 123 | ddif = hxu.eval(xt,ut) - dit; |
---|
[81] | 124 | |
---|
| 125 | //Rt = 0.9*Rt + xdif^2 |
---|
[135] | 126 | Qt*=0.1; |
---|
| 127 | Qt += 10*outer_product ( xdif,xdif ); //(x-xt)^2 |
---|
[81] | 128 | |
---|
[135] | 129 | if (Qt(0,0)>3.0){ |
---|
| 130 | cout << "here" <<endl; |
---|
| 131 | } |
---|
| 132 | // For future ref. |
---|
| 133 | xtm = xt; |
---|
| 134 | |
---|
| 135 | Rt*=0.1; |
---|
| 136 | // Rt += 10*outer_product ( ddif,ddif ); //(x-xt)^2 |
---|
| 137 | |
---|
[81] | 138 | //ESTIMATE |
---|
| 139 | Efix.bayes(concat(dt,ut)); |
---|
| 140 | // |
---|
[135] | 141 | Eop.set_parameters ( &fxu,&hxu,(Qt+1e-8*eye(4)),(Rt+1e-6*eye(2))); |
---|
[81] | 142 | Eop.bayes(concat(dt,ut)); |
---|
[94] | 143 | // |
---|
[105] | 144 | Edi.set_parameters ( &fxu,&hxu,(diag(diag(Qt))+1e-16*eye(4)), (diag(diag(Rt))+1e-3*eye(2))); |
---|
[94] | 145 | Edi.bayes(concat(dt,ut)); |
---|
[81] | 146 | |
---|
[94] | 147 | //LOG |
---|
[105] | 148 | L.logit(X_log, vec(x,4)); //vec from C-array |
---|
[94] | 149 | L.logit(Efix_log, Efix_ep.mean() ); |
---|
[105] | 150 | L.logit(Eop_log, Eop_ep.mean() ); |
---|
| 151 | L.logit(Edi_log, Edi_ep.mean() ); |
---|
| 152 | L.logit(Q_log, vec(Qt._data(),16) ); |
---|
| 153 | L.logit(R_log, vec(Rt._data(),4) ); |
---|
| 154 | L.logit(D_log, vec(KalmanObs,4) ); |
---|
[131] | 155 | L.logit(Du_log, dut ); |
---|
| 156 | L.logit(Di_log, dit ); |
---|
[94] | 157 | |
---|
| 158 | L.step(false); |
---|
[81] | 159 | } |
---|
| 160 | |
---|
[94] | 161 | L.step(true); |
---|
[135] | 162 | //L.itsave("sim_var.it"); |
---|
[105] | 163 | |
---|
[81] | 164 | |
---|
| 165 | return 0; |
---|
| 166 | } |
---|