This paper introduces valuation structures associated with preferential models. Based on
KLM valuation structures, we present a canonical approach to obtain injective preferential models for any
preferential relation satisfying the property INJ, and give uniform proofs of representation theorems for
injective preferential relations appeared in the literature. In particular, we show that, in any propositional
language (finite or infinite), a preferential inference relation satisfies INJ if and only if it can be represented
by a standard preferential model. This conclusion generalizes the result obtained by Freund. In addition,
we prove that, when the language is finite, our framework is sufficient to establish a representation theorem
for any injective relation.