I think you have a good pick with R820T2 RTL-SDR. We did use a much inferior RTL-SDR some time ago (RTL2832U Elonics E4000) as a cheap spectrum analyzer and it impressed us.
According to this: http://www.rtl-sdr.com/rtl-sdr-tests-r820t-vs-r820t2-stability-tests-for-radio-astronomy/ , the R802T2 should have pretty good performance for spectrum sensing if you are targeting the 1.0-1.7 GHz frequency range.
However please note that these devices are limited to 8-bit resolution and (for good frequency stability, as stated in the link above) 1.6 MS/s of sampling rate. You need to check if this is sufficient for your sensing algorithm. For example, if you were trying to find the best way to put cognitive radio signals inside empty GSM channels, you would need to see if these specs are enough for your algorithm to monitor sufficient GSM channels.
I believe that the bottleneck would be the RF circuitry, which has a very high thermal noise. You might work around this issue by using a good LNA front-end, otherwise the error probability will be significantly high. Try to use a good antenna as well, maybe an omni antenna with good gain characteristics (3~6 dBi).
i have already seen energy detection sensing experiment examples in RTL SDR site. one application is clear saying it is a best spectrum analyzer. As R820T is a DVB-T tuner, i am more worried about whether DVB-T signals here in India (Hyderabad) available in air or not.
Don't worry about the tuner being originally designed for DVB-T. You can gather raw I/Q samples over the entire frequency range supported by the tuner (which is much wider than DVB-T requires)