Limited data-driven energy economical optimization for quay crane scheduling

Energy economical optimization for quay crane scheduling (QCS) involves optimal crane allocation to reduce electricity costs and maintain logistics efficiency. However, it is difficult to obtain entire perception of port power grid, posing a huge challenge to energy optimization of QCS. This paper proposes a limited data-driven approach to assign quay cranes without requiring measurements from all nodes. Specifically, the limited information driven power flow variables mapping modules are proposed to learn the mapping relationship between active power, reactive power, and voltage of measurable nodes using a deep graph convolutional online learning algorithm. Subject to the mapping modules, a measurable-node data based quay crane load assignment optimization model is established. Then an adaptive argument count projection gradient descent algorithm is proposed to solve QCS strategy with the minimum electricity cost. Numerical experiments demonstrate the proposed method can achieve effective optimization with less than half of measurable nodes.