How to update stackingestimator parameters when the model pipeline was created by the older version of sckit-learn?

Question

I have a model pipeline stored in pkl file which was created by the sckit-learn 0.19. I'm trying to update the missing parameters that need for the new version of sckit-learn 0.22.

The pipeline looks like the following:

Pipeline(memory=None,
     steps=[('stackingestimator-1', StackingEstimator(estimator=DecisionTreeRegressor(criterion='mse', max_depth=2, max_features=None,
           max_leaf_nodes=None, min_impurity_decrease=0.0,
           min_impurity_split=None, min_samples_leaf=8,
           min_samples_split=7, min_weight_fraction_lea...positive=False, precompute='auto', random_state=42,
       selection='cyclic', tol=0.1, verbose=0))])

In sckit-learn 0.22, they have added a new parameter 'ccp_alpha'in the DecisionTreeRegressor. So, I need to add this new parameter into every DecisionTreeRegressor used in the pipeline. I have tried using the set_param_recursive in tpot, but it didn't add the missing attribute into the estimator:

set_param_recursive(pipeline_steps=model.steps, parameter='ccp_alpha', value=0.0)

It will return the error:

AttributeError: 'DecisionTreeRegressor' object has no attribute 'ccp_alpha'

Is there a solution to update missing parameters added in the new version?