BIP Consulting
Contact: Andrea Magatti
Country: Italy
Award Category: Innovative Problem Solver
Tell us about the business problem you were trying to solve.
Our customer needed to define the correct physical monitoring frequency for the high-pressure network transporting natural gas.
The network counts over 34k km and, by current laws, is needed to check if any digging, building, or depositing activity could happen , near the placement of the gas pipeline. The customer needs to know where and when is required to check on specific sites for any interference event.
Since managing events is a lengthy and costly task, we were asked to build a model capable of calculating a threshold about the network segment's risk and optimizing the monitoring frequency to minimize foreseeable management costs.
What SAS products did you use and how did you use them?
To address this problem, we have used various SAS components (9.4, Viya 3.5, and Viya 4), GIS software, and Python code.
We used QGis to manipulate the shape file containing the complete network since it's a common event to have a lot of issues with the practical geometry. The geometry was manipulated using QGis add-ons and some external Python code.
Once the shapefile is fixed, we use the capability of proc GIS to prepare a dataset for the next steps.
The model uses PROC PLS to reduce dimensionality and PROC COUNTREG to calculate the probability of counting events using a zero-inflated spatial regressor. We had to write the scoring code with IML since the SCORE statement of the countreg doesn't support the ZIP distribution with spatial regressors.
Once the model had been validated, we used the OPTMODEL procedure to achieve the best monitoring frequency and minimize the cost of managing interference events.
What were the results or outcomes?
The model has produced a remarkable 85% of recall for the relevant event (at least one event) and similar accuracy metrics.
Using that model, we have defined the best monitoring frequency for different event probabilities (low, middle, high). Such approach is actually field tested, on a specific monitoring district. We expect to have final results before June 2025.
Why is this approach innovative?
The main innovation in this field involves modeling events, considering their risk, and using the probability of such events to implement precautionary actions to minimize future costs arising from the standard modality of defining the monitoring frequency.
