One way of understanding it is to introduce a one-variable function and then differentiate that function in the way you already know. For instance, if you are asked to compute $\frac{\partial z}{\partial x}$ at $(x,y)=(1,2)$, you can introduce $r(x)=z(x,2)=x^2+2x+4$. Then $\frac{\partial z}{\partial x}(1,2)=r'(1)$.
If you want to get $\frac{\partial z}{\partial x}$ for general $(x,y)$, you can imagine having a different function $r$ for each value of $y$, and differentiating that function. This is not a very convenient way to do calculations, but it is a simple way of thinking about what's going on.
As a practical matter, when you compute $\frac{\partial z}{\partial x}$, you treat $y$ as if it were a number. Thus for instance $\frac{\partial}{\partial x}(xy)=y$, for the same reason that $\frac{d}{dx}(2x)=2$. Similarly $\frac{\partial}{\partial x}(y^2)=0$, for the same reason that $\frac{d}{dx}(4)=0$.
$\frac{∂z}{∂x}(x^2+xy+y^2)=\frac{d}{dx}(x^2)+\frac{d}{dx}(xy)+\frac{d}{dx}(y^2)$
And, if we work those out individually:
$\frac{d}{dx}(x^2)=2x$
(x term, but no y term, therefore: differentiate with respect to x, $x^2$ = 2x)
$\frac{d}{dx}(xy)=y$
(x term and a y term, therefore: differentiate with respect to x, x = 1 and y remains untouched, as it is held constant)
$\frac{d}{dx}(y^2)=0$
(y term but no x term, therefore: differentiate with respect to x, $y^2$ = 0)
Yielding:
$\frac{∂z}{∂x}(x^2+xy+y^2)=2x+y$
Is this correct?
– user913304 Oct 20 '15 at 23:04