Abstract: The original Tetragonula Carbonaria Optimization Algorithm (TGCOA) tends to converge slowly and may not explore the search space sufficiently when applied to complex optimization problems. To improve its search performance, the mechanisms used at different temperature stages are modified, resulting in a multi-strategy improved algorithm termed ITGCOA. At the low-temperature stage, the Golden Sine strategy is incorporated into the position-update process, gradually directing the search toward currently promising regions and improving local search accuracy. As the algorithm enters the intermediate-temperature stage, adaptive differential evolution is introduced, with its parameters adjusted according to the population state to slow premature aggregation. At high temperatures, Lévy flight generates search steps of different lengths, extending the movement range of individuals and increasing their ability to escape from local optima. ITGCOA is evaluated on the CEC2017 and CEC2022 benchmark functions, as well as on constrained engineering optimization problems. Compared with the original TGCOA and six other algorithms, ITGCOA produces better optimization results and shows a more stable convergence process.