Abstract: This study developed and validated an energy-adaptive smart hydraulic drive architecture (EA-SHDA) for 4UZL-1600 self-propelled potato collecting harvester operating under widely varying traction and speed demands. The design aims to reduce redundant hydraulic supply during high-speed light-load transit without sacrificing traction reserve for low-speed harvesting, climbing or obstacle crossing. It also examines whether an asymmetric front–rear motor arrangement with on-demand rear drive engagement can deliver stable pressure, flow, power and trajectory responses during two-wheel/four-wheel drive switching, and distinguishes energy-saving evidence from controlled simulations and stability evidence from field operation. A whole-machine resistance model and hydraulic component matching procedure were established for three typical working conditions: low-speed full-load harvesting, high-speed light-load transit, and climbing/obstacle crossing. The front axle was equipped with 32 mL/r small-displacement motors, while the rear axle adopted 63 mL/r large-displacement motors. A 100-second weighted duty cycle was defined, consisting of 80 s low-speed harvesting, 15 s high-speed transit and 5 s climbing. The proposed architecture was compared with conventional fixed two-wheel and four-wheel drive schemes under identical external load boundaries. A dynamic simulation model was built on the AMESim platform, incorporating fluid compressibility, pipeline resistance, valve throttling and motor efficiency characteristics. Field prototype tests were conducted in sandy loam soil with calibrated pressure and flow sensors, and travel trajectories were recorded. Component matching results show that four engaged motors require a total flow of 34.20 L/min at 2.5 km/h for low-speed harvesting and climbing, while two front motors demand 138.24 L/min at 35 km/h for transit. Under the same theoretical boundary, conventional fixed-drive systems need 432.00 L/min at high speed as their motors are sized for the maximum traction condition. Disconnecting the large-displacement rear branch reduces the flow requirement to 138.24 L/min, cutting the calculated pump input power from 62.71 kW to 48.02 kW—a 23.42% reduction during high-speed transit. Under low-speed harvesting and climbing conditions, the proposed system consumes 12.71 kW and 19.95 kW respectively, close to the conventional values of 12.80 kW and 21.00 kW, confirming that energy saving is not achieved at the cost of traction performance. After time weighting, the cycle-equivalent pump power decreases from 20.70 kW to 18.37 kW, corresponding to an 11.25% reduction. Dynamic simulation reveals distinct and stable response regions. Pump outlet pressure stays at 20–22 MPa during low-speed heavy-load operation, settles at 18–20 MPa after rear branch disconnection for high-speed transit, and rises to 33–35 MPa after rear drive re-engagement for climbing. Total flow maintains 34–48 L/min in low-speed stages and stabilizes around 130 L/min in the high-speed stage, consistent with theoretical calculations. Drive switching only causes short transient pressure disturbances from fluid compressibility and valve opening, with no sustained overpressure, pressure collapse or supply interruption. Field measurements further confirm the staged operation of the prototype. Flow remains near 34 L/min during low-speed harvesting, rises to approximately 138 L/min during high-speed transit, and returns to 33–35 L/min during climbing. The three power platforms are clearly separated, and switching causes no obvious energy supply gap or prolonged response lag. Straight-line wheel tracks remain parallel, with a measured minimum turning radius of about 2.7 m and no evident lateral slip or soil damage. In conclusion, the EA-SHDA reduces redundant high-speed hydraulic supply through asymmetric displacement allocation, selective drive unit engagement and operating-mode switching, while preserving traction reserve under heavy-load and sloped conditions. The consistency among resistance calculation, dynamic simulation and field measurement verifies the engineering feasibility of the design for multi-condition agricultural machinery. Quantified energy savings are derived from theoretical calculation and simulation, while field tests validate the continuity, stability, maneuverability and practical operability of the prototype. This study provides a technically grounded design route for comparable adaptive hydraulic systems.