Algorithm 1 Model Training and Forecasting Evaluation

Require: $fsc$, $frc$ Ensure: Forecasting unseen stator FFT current $seg\_fsc$,        Forecasting unseen rotor FFT current $seg\_frc$ 1:Data preprocessing using $MinMaxScale$ method: $r,s\leftarrow MinMaxScale(fsc,frc)$ 2:Split preprocessed data into training and testing data with a ratio of 80:20, respectively // For rotor data: $r\_tr,s\_tr\leftarrow r[:len(r)\times 0.8],r\left[len\right(r)\times 0.8:]$ // For stator data: $r\_te,s\_te\leftarrow s[:len(s)\times 0.8],s\left[len\right(s)\times 0.8:]$ 3:Create function $split\_IO\left(\right)$ to split input and output from training and testing data $(Xr\_tr,yr\_tr),(Xr\_te,yr\_te)\leftarrow split\_IO(r\_tr,r\_te)$ $(Xs\_tr,ys\_tr),(Xs\_te,ys\_te)\leftarrow split\_IO(s\_tr,s\_te)$ 4:Setup several hyperparameters of Bi-LSTM model: $n_i,n_h,n_o\leftarrow numberofinput,hidden,outputnodes$ $op\leftarrow trainingoptimizer$ $loss\leftarrow traininglossfunction$ $val\leftarrow validatingsplitvalue$ $call\leftarrow EarlyStoppingfunction$ 5:while $i\le epochs\left(ep\right)$do 6:    Create BiLSTM layer:     $BiLSTM\leftarrow LSTM(n_i,n_h,n_o,n_f,op,loss)$ 7:    Create $fit\left(\right)$ function to learn:      $f\_r\leftarrow BiLSTM.fit(Xr\_tr,yr\_tr,ep,val,bat,call)$      $f\_s\leftarrow BiLSTM.fit(Xs\_tr,ys\_tr,ep,val,bat,call)$ 8:end while 9:Training evaluation $mse\_r,mse\_s\leftarrow f\_r\left[loss\right],f\_s\left[loss\right]$ 10:Testing evaluation $seg\_frc\leftarrow BiLSTM.prediction(Xr\_te)$ $seg\_fsc\leftarrow BiLSTM.prediction(Xs\_te)$ 11:Calculate RMSE and MAE losses $error\_r\leftarrow seg\_frc-yr\_te$ $error\_s\leftarrow seg\_fsc-ys\_te$ $rmse\_r\leftarrow sqrt\left(mean\right(\left(square\right(error\_r\left)\right)\left)\right)$ $rmse\_s\leftarrow sqrt\left(mean\right(\left(square\right(error\_s\left)\right)\left)\right)$ $mae\_r\leftarrow mean\left(abs\right(error\_r\left)\right)$ $mae\_s\leftarrow mean\left(abs\right(error\_s\left)\right)$ 12:return $seg\_frc,seg\_fsc$