A less high-powered approach. I am assuming that since this is a model of daily sales that $t$ is the number of days, and it doesn't make sense to push the value of $t$ any further than the nearest integer. So we just want to find the day that sales pass 4500.
You went a step too far already. you want to put it in this form:
$$200(t -15)e^{-0.01t} + 3000 = 4500$$
Then solve it for the $t$ not in the exponent:
$$200(t -15)e^{-0.01t} = 1500$$
$$(t - 15)e^{-0.01t} = 7.5$$
$$t -15 = 7.5e^{0.01t}$$
$$t = 15 + 7.5e^{t/100}$$
Since $t$ is positive, and is evidently much less than $100$, we can expect that $1<e^{t/100}<2$.
Now we go through a process of refining that estimate by use the equation above to produce new limits for $t$. Those new limits provide new limits for $e^{t/100}$, which provides new limits for $t$, etc. We keep this up until the limits on $t$ are within a single day:
$$1<e^{t/100}<2$$
$$22.5 = 15 + 7.5\cdot 1 < t < 15 + 7.5\cdot 2 =30 $$
$$1.25 \approx e^{0.225 }< e^{t/100} < e^{0.3} \approx 1.35$$
$$24.39 < t < 25.124$$
$$1.276 \approx e^{0.2439} < e^{t/100} < e^{0.25124} \approx 1.286$$
$$24.57 < t < 24.64$$
So by the model, sales reach 4500 on day 24.