Advanced protection and operation of power systems, especially microgrids, requires rapid detection and estimation of system states. Microgrids, however, are a challenge due to the high variability of distributed generation and modern load profiles. Setting of relays and establishing Remedial Action Schemes on these systems requires rapid detection and state estimation and modeling of complex loads. This work demonstrates the design and operation of a high-speed method for estimating states and load composition across a 286-bus campus test-bed utilizing only a few PMUs. The method is based on training data that is analyzed offline via Singular Value Decomposition to determine clustering vectors that can be utilized for state estimation in real-time. This is combined with a priori bus studies to estimate load composition in terms of residential, industrial, and commercial in real-time. A case-study using the Oregon State University campus grid is presented, and it is shown that the estimation method is accurate (to within 1% in the test shown) while requiring less than 0.2ms computation time.