Hint: A little experimentation leads to the conjecture we can do it using $n-3$ diagonals.
For the induction step, suppose this is true for a convex polygon with $k$ sides. To deal with the case $k+1$, let $P,Q,R$ be consecutive vertices, and join $PR$. That divides our polygon into a triangle and a convex polygon with $k$ sides. Apply the induction hypothesis to this polygon.
Added: The phrasing "are needed" suggests that we also should prove that we cannot do it with fewer than $n-3$ lines. For that, strong induction is the right tool.
In the case $n=3$, we certainly need at least $n-3$ diagonals. Suppose that at least $n-3$ diagonals are needed for all convex $k$-gons with $k\lt n$. We show at least $n-3$ diagonals are needed for a convex $n$-gon.
So take a a division of the polygon into triangles by non-intersecting diagonals. Let $\ell$ be one of the diagonals of this division. This divides our $n$-gon into a $p$-gon and a $q$-gon, where $p+q=n+2$. By the induction hypothesis, at least $p-3$ diagonals are needed for the $p$-gon, and at least $q-3$ for the $q$-gon. This gives (including $\ell$) a total of at least $(p-3)+(q-3)+1=(p+q)-5=(n+2)-5=n-3$ diagonals.