Abstract: Bifacial photovoltaic (BPV) modules convert solar irradiance received by both surfaces into electrical energy and provide a higher energy yield than conventional monofacial modules. Their front-side and back-side photocurrents, however, converge through a shared p-n junction and jointly determine the terminal voltage and current. The corresponding power contributions therefore remain intrinsically coupled and cannot be directly measured, identified, or regulated as outputs from two physically independent photovoltaic sources. This electrical coupling limits flexible BPV power management in stand-alone agrivoltaic systems, where irrigation pumps, ventilation fans, spraying devices, circulation pumps, and temperature-control equipment operate intermittently and frequently generate abrupt variations in power demand. Conventional photovoltaic reserve-control methods usually establish power margins through active power curtailment, empirical maximum-power estimation, or additional electrochemical storage. These approaches cannot distinguish the electrical contributions from the two illuminated surfaces and, consequently, cannot utilize back-side generation as a separately quantified short-term reserve. Therefore, this study proposes an analytical BPV power-decoupling method based on the Lambert W function and develops a coordinated time-division power-reserve control strategy for dynamic agricultural loads.An improved equivalent BPV topology is first constructed from the classical single-diode model. Two illumination-dependent photocurrent sources describe carrier generation produced by front-side and back-side irradiance, while two parallel series-resistance branches represent side-related electrode resistance, contact resistance, and conduction loss. A common diode and shunt resistance preserve the shared junction behaviour and intrinsic leakage characteristics of the physical BPV device. Kirchhoff’s current law and the Shockley diode equation are subsequently combined to formulate a nonlinear voltage-current equation containing terminal current in linear, resistive, and exponential terms. Conventional Taylor-series approximation leads to accuracy degradation outside the selected expansion region, whereas Newton-Raphson iteration requires suitable initial values, repeated calculations, and convergence verification. The proposed method transforms the implicit BPV equation into the standard Lambert W form and obtains an explicit analytical current-voltage relationship without numerical iteration, initial-value selection, or convergence checking. Front-side and back-side currents are subsequently calculated from their relative photocurrent contributions, and the corresponding power components are determined using the common terminal voltage. This analytical allocation retains the physical coupling mechanism while providing separately identifiable power references for subsequent converter control. Comparative model evaluations under sunny and cloudy operating conditions produced power mean absolute errors of 171.62 and 9.12 W, bifacial gains of 13.81% and 48.21%, and an average calculation time of 0.95 s. These results demonstrate a satisfactory balance among prediction accuracy, bifacial-gain representation, and computational efficiency.A coordinated control architecture is then established from the calculated power distribution. The dominant front-side power component operates as the main supply and follows a proportional-integral-assisted perturb-and-observe maximum power point tracking algorithm. The perturb-and-observe component updates the maximum-power voltage reference, while proportional-integral feedback corrects photovoltaic-voltage deviations caused by irradiance variations, temperature changes, and load disturbances. The calculated back-side component is organized as a time-division power reserve. During normal operation, the controller continuously estimates the maximum available BPV power, the front-side maximum-power contribution, the back-side available margin, the load demand, and the instantaneous power deficit. A front-side power ratio maps the analytically decoupled distribution onto the measured BPV output. When load demand exceeds the front-side contribution, the reserve-power command is set to the smaller of the detected power deficit and the available back-side margin. This constraint prevents infeasible reserve commands when instantaneous solar generation cannot satisfy the complete load increment. A power-voltage dual-loop controller regulates the reserve path: the power outer loop converts reserve-power error into a voltage reference, and the voltage inner loop adjusts the converter duty ratio. The reserve path releases the requested power during load connection and withdraws its reserve contribution after the disturbance is removed. The main path simultaneously maintains maximum-power tracking and restores stable operation. This coordinated sequence converts the analytically identified back-side contribution into a dispatchable short-term reserve, improves transient supply-demand balance, and reduces dependence on electrochemical storage for brief compensation of agricultural loads.A 700 W dual-Boost direct-current converter prototype was designed and constructed to implement the proposed strategy. Boost 1 was assigned to the front-side main power path, whereas Boost 2 was assigned to the back-side reserve path. Independent duty-ratio adjustment translated the calculated front-side and back-side references into controllable hardware outputs. Field experiments were conducted using an LR5-72HGD-580M bifacial module under varying outdoor irradiance and temperature conditions. Under front-side irradiance of 957 W/m², back-side irradiance of 55 W/m², and ambient temperature of 19.5 °C, the calculated front-side and back-side powers reached 548 and 15 W, respectively. The measured conversion efficiencies of Boost 1 and Boost 2 reached 97.36% and 97.37%, producing an overall conversion efficiency of 97.36%. Full-day field measurements showed consistent trends in variation between calculated and measured power components. The front-side and back-side mean absolute errors were 17.46 and 1.63 W, respectively, confirming effective power allocation under variable outdoor conditions. When irradiance decreased abruptly from 691 to 226 W/m², the reserve and main channels recovered stable operation within approximately 0.05 and 0.06 s. Following disturbance-load connection and disconnection, the system returned to steady operation within approximately 0.04 and 0.05 s, respectively. The experimental results verify that the proposed method achieves explicit BPV power-component estimation, high-efficiency dual-channel conversion, rapid reserve release, and stable transient regulation. The developed approach provides a practical storage-free solution for short-term power balancing in agrivoltaic supply systems containing intermittent agricultural loads.