This study builds a green job-shop scheduling model considering speed-adjustable operation and charging of automated guided vehicle (AGV), aiming to minimize the makespan and total energy consumption. A multi-strategy proximal policy optimization (MSPPO) algorithm is proposed to meet the collaborative optimization scheduling requirement between operation sequencing and AGV transportation task assignment, with the complexity and dynamic correlation of decision space considered. A composite state representation is designed by integrating disjunctive graphs with artificial states, providing the agent with more comprehensive environmental information. A composite reward function combining the makespan and energy consumption is established by combining multi-objective optimization requirements to guide the solution toward the Pareto front. Furthermore, a dual-policy collaborative network architecture is designed. Specifically, the operation policy network is responsible for process sequencing, while the AGV policy network is responsible for AGV transportation task assignment. The sequential execution of the two networks constitutes a complete scheduling action, realizing collaborative optimization of processing and transportation. The two networks share advantage estimation, employ policy feedback mechanisms, and perform joint state updates to enable dynamic mutual adaptation and improve both algorithm robustness and convergence, thus enhancing inter-policy collaboration and adaptability. The effectiveness and superiority of the proposed algorithm are verified by experimental analysis and comparison on extended test cases.