With the rapid development of renewable energy and the increasing intelligence of load management, flexible load regulation has become a critical component in the advancement of smart grids. Conventional intelligent optimization algorithms search for the optimal regulation curve through random variations. However, due to the strong uncertainty inherent in random variations, the search for the optimal solution often requires significant time. Additionally, the high dimensionality of input variables and the multitude of possible random variations further increase the time consumption in finding the optimal solution. To expedite the search for the optimal solution, reduce the time consumption during the optimization process, and improve the accuracy of flexible load regulation, this paper proposes a fast optimization algorithm for intelligent flexible load regulation, tailored to the characteristics of flexible load control. The new method divides the entire optimization process into two steps: fast approach and fine control. The algorithm firstly employs a coarse optimization to quickly approach the optimal solution, followed by an improved stochastic intelligent optimization algorithm to search for the optimal solution, thereby reducing optimization errors. Simulation results demonstrate that, compared to conventional stochastic optimization algorithms, the residual cost optimization method can enhance the regulation effectiveness of flexible loads to a certain extent. Moreover, heuristic search further accelerates the convergence speed of the simulated annealing optimization algorithm, significantly improving the regulation performance of flexible loads. It has been verified that adopting heuristic residual cost simulated annealing can reduce load fluctuations by 3.4%.