Facing the climate crisis, more and more renewable inverter-based resources (IBRs), such as wind energy conversion systems (WSs), replace grid-connected synchronous machines (SMs). SMs inherently provide balanced and sinusoidal grid voltages but also power system inertia, whereas standard grid-following (GFL) control of IBRs leads to decreasing power system inertia. This paper proposes grid-forming (GFM) control for doubly-fed induction machine (DFIM) based wind energy conversion systems (WSs) based on the virtual synchronous machine (VSM) concept with the following extensions: (i) Maximum power power tracking (MPPT) compensation for accurate inertia emulation, (ii) feedforward torque control (FTC) for fast power reference tracking, (iii) de-rated operation or reference power point tracking (RPPT) to provide additional power reserves, (iv) dynamic droop power saturation control to avoid excessive power overshoots (as available power reserves are taken into account), and (v) grid voltage control utilizing reactive power from both, DFIM stator and rotor-side back-to-back inverter. The modeling and control of the DFIM-based WSs is shown in detail and the grid model is based on the IEEE 9-bus test system. Comprehensive simulation results show that the GFM control enables unlimited power penetration of IBRs and that the grid frequency dynamics of VSM-based power systems can be interpreted and handled analogously to SM-based power systems. During normal operation, compared to existing VSM control without FTC, the proposed VSM control with FTC increases the wind energy yield, i. e. typical MPPT or RPPT performance is achieved, similar to the performance of GFL control. For high power penetration of IBRs, the proposed VSM control enables stable operation due to its inertia emulation or GFM capability, whereas GFL control tends to instability. During faults and system splits, the VSM provides higher power system damping than a real SM due to (i) virtual/internal damping that can be adapted by software (hence almost independently from the hardware, whereas the response of SM damper windings depends on the physical machine design and cannot be altered during operation), and due to (ii) faster droop control which adapts the virtual turbine power without the mechanical delays that dominate real turbine dynamics.