data_fv_copy=eolico_df[:"2020-01-01"]
    data_fv_copy[~data_fv_copy.index.duplicated()].resample("H").mean()
    data_fv_copy =data_fv_copy[["{}".format(zone.upper()),"tt_con_it-{}".format(zone),"pro_wnd_it-{}".format(zone),"con_it-{}".format(zone)]].copy()

    data_fe = feature_engineering(data_fv_copy,zone.upper())
    tgt = zone.upper()
    print(tgt)
    ens = TimeSeriesEnsemble()

    ens.set_target(tgt)
    ens.add_model({"mod_name":"m1","mod_type":"lasso","mod_params":{"alpha":0.001,"max_iter":1000},"scaler_type":"robust"})

    #ens.add_model({"mod_name":"m2","mod_type":"svmr","mod_params":{"C":1},"scaler_type":"min_max","detrend_type":"mean"})

    ens.add_model({"mod_name":"m2","mod_type":"xgbr","mod_params":{},"scaler_type":None})

    ens.add_model({"mod_name":"m3","mod_type":"rfrr","mod_params":{},"scaler_type":None})

    ens.add_model( {"mod_name":"m4","mod_type":"lasso","mod_params":{"alpha": 0.001,'max_iter':1000},"scaler_type":"min_max"})
    ens.update_common_params({"train_share":0.8})
    fv= ForwardValidator(ens)


    fv.set_fv_data_fe(data_fe)
    #val size = quante righe è grande la validation
    #val steps : quanti loop deve fare
    #train size =-1 finestra allenamento non è rolling se metti 1 parte sempre dalla stessa riga
    fv.set_forward_validation(val_steps=12, val_size=15*24, train_size=-1 )
    a,b,c=fv.perform_forward_validation()
    plots_eolico[tgt]=c
    results_eolico[tgt]=b