In agricultural production, the scheduling of multi-task and multi-agricultural machinery faces challenges like uneven task distribution and operational inefficiencies, resulting in prolonged makespan and low resource utilization. Therefore, we propose a hybrid algorithm combining the Non-dominated Sorting Genetic Algorithm II (NSGA-II) with improved Variable Neighborhood Search (IVNS), named as INSGA-II. IVNS integrates dynamic trigger probability, search depth control and elite retention mechanism. These strategies strengthen local exploitation and global exploration, effectively balancing optimization, leading to faster convergence and improved solution quality. Simulation results indicate that the proposed INSGA-II exhibits superior efficacy compared to the other four baseline algorithms in minimizing the maximum completion time for all tasks across fields. For instance, with a field size of 15, the maximum completion time is reduced by 11.29% (PSO), 5.66% (ACO), 8.29% (GA), and 6.34% (NSGA-II), respectively, which demonstrates the effectiveness and reliability of the improved algorithm. The proposed method demonstrates strong potential to mitigate these challenges, providing an efficient and scalable approach for optimizing agricultural machinery scheduling problems.