Exploiting the regulation capability of industrial loads is an important approach to addressing increasingly prominent power and energy balance challenges in new-type power systems. However

the adjustable capability of industrial loads is constrained by complex coupling among multiple internal production processes

making it hard to achieve accurate assessment

while the solution efficiency of existing industrial load control methods still needs improvement. Taking screw production loads as an example

this paper analyzes the regulation mechanisms of individual processes

establishes a production process model and a valid action recognition model

and realizes independent quantification of adjustable capability boundaries before control

thereby improving assessment effectiveness. For electricity shortage scenarios

a Stackelberg game-based control model between a park operator and industrial loads is developed. Based on historical response contribution credits and adjustable capability assessment results

response tasks are fairly allocated and industrial loads are guided for participating autonomously through the Stackelberg game

improving fairness under frequent user participation. Furthermore

this paper proposes a valid action recognition proximal policy optimization (VAR-PPO) algorithm to train industrial load agents and embed them into the game-solving process

thereby improving solution efficiency. Case studies show that screw production loads have considerable regulation potential

and the proposed method can support park-level supply-demand coordination under shortage scenarios.