Relevance. Currently, the share of centrifugal electric pumps operating in intermittent modes is increasing, allowing for increased oil production from low- and medium-flow wells. This is achieved by switching the well to two-position control modes with separate pumping and fluid accumulation cycles. However, this significantly reduces the time it takes to bring the well up to operating mode, which can reach several minutes. This approach reduces the mean time between failures of submersible electric motors and components of the submersible section of the unit due to increased vibration amplitude and torsional oscillations compared to quasi-stationary processes during continuous well operation. The impact of starting currents and torques, which lead to excessive fatigue wear of the submersible section components, can be reduced by using closed-loop sensorless control systems for submersible electric motor drives. This necessitates the development of methods for identifying unmeasured state variables of submersible electric motors, such as the flux vector and rotor angular velocity, using full-order state observers. Aim. Development and research of a rotor speed and load torque full-order observer for the dynamic system «cable line – permanent magnet synchronous motor» based on explicit mathematical models in discrete time for the implementation of closed-loop sensorless control systems for the electric drive of centrifugal electric pump units. Methods. The work used methods of system analysis and structural-parametric synthesis of dynamic systems, methods of identifying unmeasured state variables of dynamic systems based on state observers, methods of numerical mathematical modeling of dynamic systems described by systems of stiff differential equations, and methods of solving systems of nonlinear algebraic equations. Results. The authors have developed the adaptive mathematical model of a rotor speed and load torque full-order observer for the dynamic system «cable line – permanent magnet synchronous motor». The observer performance was studied under typical operating conditions and when the parameters of its customizable model deviated from the reference values. For all operating conditions, the relative errors in estimating the rotor angular velocity and shaft drag torque were less than 1.5 and 1%, respectively.