Symmetries of PDEs

Question

I wonder who raise the concept that in a PDE system $\mathcal{L}$ with $$ \mathcal{L}(u,x,y)=0$$

and $$\mathcal{L}(\mathcal{T}(u,x,y))=0$$ implies that operator $\mathcal{T}$ is a symmetry of $\mathcal{L}$.

I would appreciate it if the original paper could be mentioned.Thank you.