Aside from the problem of creating and maintaining otherwise identical and stable systems (nicely described by Joel above), creating a true scientific experiment is not terribly difficult.
In this scenario, three students plus the teacher are needed. These people will be called P1, P2, P3, and T1. T1 is considered a trusted neutral third party (the teacher).
P1 - Generally the person that forms and states the hypothesis. This person would also develop the experimental protocol. This person would be responsible for designing the experimental setup, selecting the equipment, livestock, and so forth. All of this would be documented. This person is absolutely forbidden from ever touching the tanks, and likewise also forbidden from knowing which of the tanks is the experiment and which is the control. (The hypothesis statement is biased by nature, for example, "a rise in temperature of 3-4 degrees C will cause harmful affects to a variety of Acropora corals," which is why this person can't know which tank is which.) As the experiment progresses this person is responsible for making and recording detailed observations of what's happening with the animals inside the tanks.
P2 - "Lab technician" assigns and makes a written note of which tank is the experiment and which is the control. This recorded record is given, before the experiment starts, to T1 for safe keeping. P2 is responsible for preparing the maintenance materials, foods, etc. used in maintaining the two tanks. These materials would be required to be indistinguishable for the two tanks (for instance, if testing CO2 absorption, both tanks would have a CO2 canister filled with a gas that is bubbled through the water - for one of them this might actually be nitrogen while the other is CO2). P1 and P3 are not permitted to be present while P2 is working as that might expose P1 and P3 to knowledge of which tank is the control and which is the experiment. Because this person knows which tank is which, this person is not permitted to make observations, perform tank maintenance, or otherwise touch the tanks.
P3 - "Lab technician" performs tank maintenance using the materials specifically prepared for the individual tanks by P2. This person is not permitted to know which tank is which in order to prevent bias in the maintenance routine and must perform the maintenance while P1 and P2 are not present.
T1 - "Trusted third party" keeps all written records and insures that the experimental protocol is followed by P1, P2, and P3. This person also knows which tank is which, and so is not permitted to make observations, do maintenance, or otherwise touch the tanks.
This is a double-blind setup, as the person making observations doesn't know which tank is which and isn't permitted to affect the maintenance, and likewise the person doing tank maintenance (and thus administering whatever is being experimented with) also doesn't know which tank is which.
There's still a lot that can go wrong - for instance, a janitor might accidentally spill something into one of the tanks and not mention it to anyone. Things like that can be really hard to account for, but should come out in the next step.
This would be quite newsworthy for a high school student (top ranking undergrads sometimes get this far), but the next step would be to let the larger community know about the experiment and the results. To do so the hypothesis, description of the protocol, observations and other supporting data, and any Conclusions would be written up in a paper and submitted for publication. Here's where the test of the quality of the experimental protocol itself comes in - in order to be rigorously scientific, the paper would have to be published in a peer-reviewed journal whose focus is the field being studied. The reviewers are responsible for looking for flaws in the protocol, the implementation of the protocol (as expressed in the data gathered), or leaps to judgment without supporting data within the Conclusion. The field of the journal has to be related to that of the paper, otherwise the peer reviewers are unlikely to be qualified to properly review the paper (reputable journals don't allow such issues to arise in the first place). Let's assume it passes through review with only minor revisions (getting published without minor revisions is rare) and accepted for publication (because of print cycles, layout, formatting of images and data, actual publication in a physical journal can take months.) Despite being accepted for publication, it's still not taken as a valid hypothesis. It's only an "interesting" hypothesis. The next step would be for another set of experimenters to duplicate the experiment using the same experimental setup and protocol described in the first paper and to publish their results in a similar peer-reviewed journal. Assuming the second experiment's results match, only at that point is the experiment considered to be valid-until-proven-otherwise. If the second experiment's results don't match the first, then everyone involved examines both sets of documentation in detail and tries to figure out what went wrong, (which might result in another publication or two) followed by more experiments and so forth until the problem is better understood, the hypothesis revised to be more accurate and/or the results start to agree because of a documented change.
The news media routinely reports on experimental results as if they are valid after the first paper has been accepted for publication (or sometimes even before then)... Not a good thing.