Scale Inhibitor Squeeze Treatments Lifetime Modelling: Data-Driven Versus Conventional Approaches

A scale-inhibitor squeeze-treatment is a preventive well-intervention technique where scale-inhibitor chemicals are injected into the formation to provide long-term protection against inorganic-scale deposition in the production tubing and near-wellbore region. Chemicals retain on rock surfaces by adsorption/precipitation and slowly release into produced brine when the well is back-in-production. If the produced chemical-concentration is above a certain threshold, normally few ppm, the well will be protected. Conventional squeeze-treatment modelling is based on first-principles physics and chemistry that describe transport and retention mechanisms of scale-inhibitors in porous-media, normally described by adsorption-isotherms derived from laboratory coreflood experiments, or from historical field treatment data. This approach remains the industry standard for squeeze design due to its strong theoretical foundation and proven reliability, despite requiring significant laboratory data and making simplifying assumptions about complex reservoir behaviors. In contrast, data-driven approaches leverage machine learning algorithms, historical treatment performance databases, and real-time monitoring data to identify complex, non-linear relationships between operational parameters and squeeze-lifespan that traditional models might overlook. While conventional methods provide a well-established framework that works well, they struggle when historical data from wells under treatment are not available. Data-driven methods excel at optimizing treatments in mature fields with extensive historical data, even if data for the well in question are not available. This study compares a conventional approach with a data-driven method, in terms of prediction capability, based on the same historical data, i.e. historical squeeze-treatments with the same inhibitor type for a given field. After applying both methodologies, it is apparent that in terms of predictability they are very similar. The conventional approach provides a deeper understanding of the problem but relies on the availability of historical data of the well under study. Data-driven models cannot provide deep insights but provide good predictions where historical data are available. Two field-datasets with adequate historical data are used to compare both approaches. For the conventional approach a history matched isotherm was derived, whereas the data-driven model was trained with the historical data. The matches to historic treatments were close, but differences were evident in the optimisation of subsequent treatments, where data-driven models are observed to be less conservative. Conventional methods rely on simplified mathematical models that extrapolate limited field data, often resulting in conservative designs with significant safety margins that may sacrifice treatment longevity or cost-efficiency. The emerging-industry trend favours hybrid approaches that combine the mechanistic understanding of conventional methods with the predictive power of data-driven methods, creating adaptive models that continuously improve with each treatment while maintaining interpretability for engineering decision-making. Therefore, our recommendation is not to limit practice of using one or the other alone: they should be used to complement each other.