Abstract: The heavy-load or overload of distribution transformers is one of the primary causes leading to transformer failures. Therefore, accurately predicting the operational status of distribution transformers is crucial for the safe and reliable operation of power systems. Due to the influence of numerous complex variables on the load of distribution substations, it is often challenging to reliably model and estimate the effects of these complex variables, resulting in a certain level of uncertainty in the final prediction results. Traditional single-point predictions aim to forecast a single optimal value, failing to adequately quantify the uncertainty of the predictions. This paper combines multi-dimensional features with historical operational data of transformers and employs quantile regression methods to model the load situation of substations. Conditional quantiles are integrated with general linear or nonlinear models to construct a probabilistic prediction model. Traditional point predictions, due to their inability to estimate the uncertainty of the prediction results, entail a certain level of risk for business decision-making. In contrast, probabilistic predictions not only provide the optimal forecast point in the future but also offer information about the distribution of future forecast values. Probabilistic predictions can better estimate the occurrence of overloading in the form of prediction intervals or quantiles, providing a basis for decision-making in ensuring the safe and stable operation of the power grid.