This work focuses on using a dynamics-based approach to understand the effects of modeling choices on simulation of high-performance multi-rotor drones carrying heavy suspended payloads. Three aspects of the modeling and simulation of multi-rotor drones are examined: dynamics and system modeling choices, energy costs of controllers, and gust disturbance modeling. First, an 18-rotor drone is modeled as a rigid body using a quaternion formulation with the aerodynamics of the rotors blades simulated using HARP_Opt. To control the system, both a PID and a PD controller are developed using the same proportional and derivative gains. The gust disturbance is modeled deterministically as a simple boxcar function with a duration of one second. First, the dynamics of the system are explored. Since these large multi-rotor drones are expected to carry heavy payloads, the effect of a heavy suspended payload is found through two means. Firstly, a linear stability analysis reveals that the model only has negative and zero eigenvalues which imply the the system is at least marginally stable. However, the stability of the nonlinear system cannot be guaranteed because of the zero-eigenvalues. To determine the stability of the nonlinear system, the system with a heavy suspended payload's response to a simple wind disturbance is simulated. The system's response reveals that the presence of the heavy suspended payload makes the system's response diverge significantly from the case when there is no heavy suspended payload with as large as several orders of magnitude difference in the responses. These results demonstrate that the nonlinear stability of a large multi-rotor drone with a suspended payload must be considered when analyzing future drone designs for stability. Next, the effect of aerodynamic model choice for the rotors is explored by comparing a lumped blade (LB) model with a blade element theory (BET) model. The two aerodynamic models are simulated with identical inputs to follow four simple trajectories that involve vertical flight with no rotation as well as vertical flight with roll, pitch, and yaw rotation. The results show a maximum trajectory error of up to 91% in the case of vertical flight with yaw rotation. This demonstrates that aerodynamic model choice has a significant effect on simulation results and the BET model needs to be considered despite it being more computationally intensive than the LB model. The last portion of the system dynamics explored is the potential for using rotor groupings to reduce the total energy required to fly the drone. With 18 rotors, there are many ways in which the drone's trajectory can be flown. Two different rotor groupings are considered. The two groupings are used to fly the drone on almost identical trajectories. The resulting energy calculations reveals that the two groupings consistently maintain a 4-5% difference in energy cost for all the different trajectories simulated. Second, the energy costs of controllers are explored. To do this, a method is proposed that uses the difference between open-loop system and closed-loop system energy costs to determine the energy cost of using a control system. As a case study, this method is then used to objectively compare a PID controller to a PD controller and characterize the energy cost of integral control. The resulting simulations reveal that both controllers perform similarly in tracking the desired trajectory with about a 5% average tracking error for low-amplitude trajectories and about a 20% tracking error for large-amplitude trajectories. However, the PID controller consistently uses more energy than the open-loop system (7-12.5% more) while the PD controller uses similar amounts of energy to the open-loop system. Additional simulations were completed looking at variations on the integral control. The resulting plots of energy and tracking error versus integral control gain magnitude demonstrate that there is a trade-off between tracking error and energy that needs to be balanced for each application to ensure all tracking and energy specifications are met. Finally, the necessity of modeling gust disturbances using a stochastic model is evaluated. This is done by using a simple, deterministic gust disturbance with variable windspeed and angle and varying which rotors are affected by the gust. It is found that when all of the rotors are hit by the gust, then the drone responds mainly in the vertical direction. When half of the rotors are hit by the gust, then the drone exhibits about an order of magnitude more motion in the horizontal direction leading to very different overall responses. Changes in windspeed and angle also result in inconsistent variations in the final trajectory and energy usage as well. This unpredictability of the resulting trajectory and energy consumption shows that the drone's gust response is highly sensitive to changes in wind parameters. In addition, the drone's sensitivity to system parameter changes is tested by varying the control gains of the system. It is found that when the proportional, integral, and derivative control gains are all varied by the same amount, then the drone's trajectory and energy usage vary significantly. As the gains are reduced, the energy usage compared to the no-wind case varies by between 0.69-82.73%. In addition, when only the integral control is varied, the energy varies by 8.88-18.07% relative to the no-wind case. These results demonstrate that both trajectory and energy usage are sensitive to changes in the gust parameters which means that for more complicated gust disturbances, the drone's response will be even harder to characterize with deterministic models. Therefore, it is necessary to use stochastic gust modeling and simulations to fully characterize how a drone will respond to a random gust disturbance.