[105] | 1 | /*! |
---|
[81] | 2 | \file |
---|
[223] | 3 | \brief Simulation of disturbances in PMSM model, EKF runs with simulated covariances |
---|
[81] | 4 | \author Vaclav Smidl. |
---|
| 5 | |
---|
[224] | 6 | \ingroup PMSM |
---|
[81] | 7 | ----------------------------------- |
---|
| 8 | BDM++ - C++ library for Bayesian Decision Making under Uncertainty |
---|
| 9 | |
---|
| 10 | Using IT++ for numerical operations |
---|
| 11 | ----------------------------------- |
---|
| 12 | */ |
---|
| 13 | |
---|
[262] | 14 | |
---|
[81] | 15 | #include <stat/libFN.h> |
---|
| 16 | #include <estim/libKF.h> |
---|
[108] | 17 | //#include <estim/libPF.h> |
---|
| 18 | #include <math/chmat.h> |
---|
[81] | 19 | |
---|
| 20 | #include "pmsm.h" |
---|
| 21 | #include "simulator.h" |
---|
[117] | 22 | #include "sim_profiles.h" |
---|
[81] | 23 | |
---|
[94] | 24 | #include <stat/loggers.h> |
---|
[81] | 25 | |
---|
[254] | 26 | using namespace bdm; |
---|
[81] | 27 | |
---|
| 28 | int main() { |
---|
| 29 | // Kalman filter |
---|
| 30 | int Ndat = 90000; |
---|
| 31 | double h = 1e-6; |
---|
| 32 | int Nsimstep = 125; |
---|
| 33 | |
---|
[232] | 34 | dirfilelog L("exp/sim_var2",1000); |
---|
[105] | 35 | //memlog L(Ndat); |
---|
[94] | 36 | |
---|
[232] | 37 | |
---|
[81] | 38 | // SET SIMULATOR |
---|
| 39 | pmsmsim_set_parameters ( 0.28,0.003465,0.1989,0.0,4,1.5,0.04, 200., 3e-6, h ); |
---|
| 40 | double Ww = 0.0; |
---|
| 41 | vec dt ( 2 ); |
---|
[240] | 42 | vec ut ( 2 ); |
---|
| 43 | vec utm ( 2 ); |
---|
[135] | 44 | vec dut ( 2 ); |
---|
[131] | 45 | vec dit (2); |
---|
[240] | 46 | vec x2o(2); |
---|
[81] | 47 | vec xtm=zeros ( 4 ); |
---|
[135] | 48 | vec xte=zeros ( 4 ); |
---|
[81] | 49 | vec xdif=zeros ( 4 ); |
---|
| 50 | vec xt ( 4 ); |
---|
[94] | 51 | vec ddif=zeros(2); |
---|
[232] | 52 | IMpmsm2o fxu; |
---|
[81] | 53 | // Rs Ls dt Fmag(Ypm) kp p J Bf(Mz) |
---|
| 54 | fxu.set_parameters ( 0.28, 0.003465, Nsimstep*h, 0.1989, 1.5 ,4.0, 0.04, 0.0 ); |
---|
| 55 | OMpmsm hxu; |
---|
| 56 | mat Qt=zeros ( 4,4 ); |
---|
[94] | 57 | mat Rt=zeros ( 2,2 ); |
---|
[81] | 58 | |
---|
| 59 | // ESTIMATORS |
---|
| 60 | vec mu0= "0.0 0.0 0.0 0.0"; |
---|
| 61 | vec Qdiag ( "62 66 454 0.03" ); //zdenek: 0.01 0.01 0.0001 0.0001 |
---|
| 62 | vec Rdiag ( "100 100" ); //var(diff(xth)) = "0.034 0.034" |
---|
| 63 | mat Q =diag( Qdiag ); |
---|
| 64 | mat R =diag ( Rdiag ); |
---|
[279] | 65 | EKFfull Efix ; |
---|
[81] | 66 | Efix.set_est ( mu0, 1*eye ( 4 ) ); |
---|
| 67 | Efix.set_parameters ( &fxu,&hxu,Q,R); |
---|
| 68 | |
---|
[279] | 69 | EKFfull Eop ; |
---|
[81] | 70 | Eop.set_est ( mu0, 1*eye ( 4 ) ); |
---|
| 71 | Eop.set_parameters ( &fxu,&hxu,Q,R); |
---|
| 72 | |
---|
[279] | 73 | EKFfull Edi ; |
---|
[94] | 74 | Edi.set_est ( mu0, 1*eye ( 4 ) ); |
---|
| 75 | Edi.set_parameters ( &fxu,&hxu,Q,R); |
---|
| 76 | |
---|
[271] | 77 | const epdf& Efix_ep = Efix.posterior(); |
---|
| 78 | const epdf& Eop_ep = Eop.posterior(); |
---|
| 79 | const epdf& Edi_ep = Edi.posterior(); |
---|
[105] | 80 | |
---|
[94] | 81 | //LOG |
---|
[162] | 82 | RV rQ( "{Q }", "16"); |
---|
| 83 | RV rR( "{R }", "4"); |
---|
| 84 | RV rUD( "{u_isa u_isb i_isa i_isb }", ones_i(4)); |
---|
[232] | 85 | RV rDu("{dux duy }",ones_i(2)); |
---|
[162] | 86 | RV rDi("{disa disb }",ones_i(2)); |
---|
[94] | 87 | int X_log = L.add(rx,"X"); |
---|
[232] | 88 | int X2o_log = L.add(rx,"X2o"); |
---|
| 89 | int Xdf_log = L.add(rx,"Xdf"); |
---|
[94] | 90 | int Efix_log = L.add(rx,"XF"); |
---|
| 91 | int Eop_log = L.add(rx,"XO"); |
---|
| 92 | int Edi_log = L.add(rx,"XD"); |
---|
[105] | 93 | int Q_log = L.add(rQ,"Q"); |
---|
| 94 | int R_log = L.add(rR,"R"); |
---|
| 95 | int D_log = L.add(rUD,"D"); |
---|
[131] | 96 | int Du_log = L.add(rDu,"Du"); |
---|
| 97 | int Di_log = L.add(rDi,"Di"); |
---|
[94] | 98 | L.init(); |
---|
[81] | 99 | |
---|
| 100 | for ( int tK=1;tK<Ndat;tK++ ) { |
---|
| 101 | //Number of steps of a simulator for one step of Kalman |
---|
| 102 | for ( int ii=0; ii<Nsimstep;ii++ ) { |
---|
[135] | 103 | sim_profile_steps1 ( Ww , false); |
---|
[81] | 104 | pmsmsim_step ( Ww ); |
---|
| 105 | }; |
---|
| 106 | // simulation via deterministic model |
---|
[232] | 107 | ut ( 0 ) = KalmanObs[4]; |
---|
| 108 | ut ( 1 ) = KalmanObs[5]; |
---|
| 109 | |
---|
[240] | 110 | x2o = fxu.eval2o(utm-ut); |
---|
[232] | 111 | utm = ut; |
---|
| 112 | |
---|
[81] | 113 | dt ( 0 ) = KalmanObs[2]; |
---|
| 114 | dt ( 1 ) = KalmanObs[3]; |
---|
[131] | 115 | dut ( 0 ) = KalmanObs[4]; |
---|
| 116 | dut ( 1 ) = KalmanObs[5]; |
---|
| 117 | dit ( 0 ) = KalmanObs[8]; |
---|
| 118 | dit ( 1 ) = KalmanObs[9]; |
---|
[81] | 119 | |
---|
[135] | 120 | xte = fxu.eval ( xtm,ut ); |
---|
[81] | 121 | //Results: |
---|
[135] | 122 | // in xt we have simulation according to the model |
---|
[81] | 123 | // in x we have "reality" |
---|
[135] | 124 | xt ( 0 ) =x[0];xt ( 1 ) =x[1];xt ( 2 ) =x[2];xt ( 3 ) =x[3]; |
---|
| 125 | xdif = xt-xte; //xtm is a copy of x[] |
---|
[131] | 126 | if (xdif(0)>3.0){ |
---|
| 127 | cout << "here" <<endl; |
---|
| 128 | } |
---|
[81] | 129 | if ( xdif ( 3 ) >pi ) xdif ( 3 )-=2*pi; |
---|
| 130 | if ( xdif ( 3 ) <-pi ) xdif ( 3 ) +=2*pi; |
---|
[94] | 131 | |
---|
[135] | 132 | ddif = hxu.eval(xt,ut) - dit; |
---|
[81] | 133 | |
---|
| 134 | //Rt = 0.9*Rt + xdif^2 |
---|
[135] | 135 | Qt*=0.1; |
---|
| 136 | Qt += 10*outer_product ( xdif,xdif ); //(x-xt)^2 |
---|
[81] | 137 | |
---|
[135] | 138 | if (Qt(0,0)>3.0){ |
---|
| 139 | cout << "here" <<endl; |
---|
| 140 | } |
---|
| 141 | // For future ref. |
---|
| 142 | xtm = xt; |
---|
| 143 | |
---|
| 144 | Rt*=0.1; |
---|
| 145 | // Rt += 10*outer_product ( ddif,ddif ); //(x-xt)^2 |
---|
| 146 | |
---|
[81] | 147 | //ESTIMATE |
---|
| 148 | Efix.bayes(concat(dt,ut)); |
---|
| 149 | // |
---|
[135] | 150 | Eop.set_parameters ( &fxu,&hxu,(Qt+1e-8*eye(4)),(Rt+1e-6*eye(2))); |
---|
[81] | 151 | Eop.bayes(concat(dt,ut)); |
---|
[94] | 152 | // |
---|
[105] | 153 | Edi.set_parameters ( &fxu,&hxu,(diag(diag(Qt))+1e-16*eye(4)), (diag(diag(Rt))+1e-3*eye(2))); |
---|
[94] | 154 | Edi.bayes(concat(dt,ut)); |
---|
[81] | 155 | |
---|
[94] | 156 | //LOG |
---|
[105] | 157 | L.logit(X_log, vec(x,4)); //vec from C-array |
---|
[240] | 158 | L.logit(X2o_log, x2o); |
---|
[232] | 159 | L.logit(Xdf_log, xdif); |
---|
[94] | 160 | L.logit(Efix_log, Efix_ep.mean() ); |
---|
[105] | 161 | L.logit(Eop_log, Eop_ep.mean() ); |
---|
| 162 | L.logit(Edi_log, Edi_ep.mean() ); |
---|
| 163 | L.logit(Q_log, vec(Qt._data(),16) ); |
---|
| 164 | L.logit(R_log, vec(Rt._data(),4) ); |
---|
| 165 | L.logit(D_log, vec(KalmanObs,4) ); |
---|
[131] | 166 | L.logit(Du_log, dut ); |
---|
| 167 | L.logit(Di_log, dit ); |
---|
[94] | 168 | |
---|
[162] | 169 | L.step(); |
---|
[81] | 170 | } |
---|
| 171 | |
---|
[162] | 172 | L.finalize(); |
---|
[135] | 173 | //L.itsave("sim_var.it"); |
---|
[105] | 174 | |
---|
[81] | 175 | |
---|
| 176 | return 0; |
---|
| 177 | } |
---|