The heat dissipation rate between a CaF2 optical micro-cavity bulk and the environment can be detected by measuring the thermo-optic oscillation period. However, there exists a nonlinear relation between the heat dissipation rate and multiple oscillation periods, and thus the heat dissipation rate cannot be effectively detected at a certain oscillation period. A sensing data measurement model based on the back-propagation artificial neural network was applied, and the heat dissipation rate was effectively measured by measuring the oscillation periods. The neural network parameters were optimized to improve the measurement accuracy of heat dissipation rate. The numerical simulation results demonstrate that the proposed method can effectively detect the heat dissipation rate in a CaF2 optical micro-cavity, which is essential for realizing thermal parameter sensing based on an optical micro-cavity.