<p>Probabilistic hesitant fuzzy sets (PHFSs) extend hesitant fuzzy sets by attaching probabilities to alternative membership degrees, enabling a more faithful representation of experts’ hesitant judgments in multi-criteria decision making (MCDM). However, many existing probabilistic hesitant fuzzy (PHF) distance-based approaches either require probabilistic hesitant fuzzy elements (PHFEs) to have equal lengths or fail to adequately capture intrinsic hesitancy, which may compromise uncertainty quantification and objective weight estimation. To address these issues, this paper studies PHF-MCDM problems with completely unknown attribute weights and proposes an integrated distance-entropy-TOPSIS framework. A counting unit splitting standardization method is developed to reconcile unequal-length PHFEs without artificial padding, thereby preserving the original probabilistic structure and reducing subjective distortion. Then, a three-dimensional hesitancy degree is defined in terms of membership deviation, element dispersion, and probability proximity. On this basis, several new PHF distance measures are constructed. Subsequently, a distance-based PHF entropy is formulated and incorporated into an entropy-weighting scheme, which is then combined with a modified TOPSIS procedure to obtain the final ranking. The effectiveness and robustness of the proposed method are validated through a safety performance assessment of four lithium battery energy storage power stations, complemented by parameter-sensitivity analysis and comparative studies with representative MCDM methods.</p>