The increasing demand for real-time applications of Wireless Sensor Networks (WSNs) makes Quality of Service (QoS)-based charging scheduling models an interesting and hot research topic. Satisfying QoS requirements (e.g. data collection integrity, charging respond delay, etc.) for the different applications of WSNs raises significant challenges. More precisely, an effective scheduling strategy not only needs to improve the charging efficiency of charging vehicles but also needs to reduce the charging respond delay of the requests to be charged, all of which must be based on the integrity of data collection. For such applications, existing studies on charging issue often deployed one or more mobile vehicles, which have deficiencies in practical applications. On one hand, it usually is insufficient to employ just one vehicle to charge many sensors in a large-scale application scenario due to the limited battery capacity of the charging vehicle or energy depletion of some sensors before the arrival of the charging vehicle. On the other hand, while the collaboration between multiple vehicles for large-scale WSNs can significantly increase charging capacity, the cost is too high in terms of the initial investment and maintenance costs of these vehicles. To overcome these deficits, in this work, we propose a novel QoS-based on-demand charge scheduling (abbreviated shortly as QOCS) model that one charging vehicle carries multiple removable battery powered chargers. In the novel QoS-based charging model, we study the charging scheduling problem of requesting nodes to guarantee the integrity of network data collection and maximize the satisfaction of charging services. In the QOCS model, We jointly consider the coverage contribution and energy urgency to sort the charging requests of sensors, and introduce a hybrid power supply mechanism based on supply and demand to improve energy utilization. We evaluate the performance of the proposed model through extensive simulation and experimental results show that our model achieves better performance than existing methods.