from sklearn.metrics import mean_squared_error, mean_absolute_errortrue = actual_values  # 实测数据
pred = predicted_values  # 模拟数据rmse = mean_squared_error(true, pred, squared=False)
mae = mean_absolute_error(true, pred)
nse = 1 - sum((true - pred)**2) / sum((true - np.mean(true))**2)print(f'RMSE: {rmse:.3f}, MAE: {mae:.3f}, NSE: {nse:.3f}')

NSE越接近1越好，MAE越接近0效果越好，RMSE越小越好。