Given the increasing abundance of water quality monitoring data, this study develops a variational data assimilation model for one-dimensional unsteady river water quality simulation based on the optimal control theory of partial differential equations, with the goal of improving the accuracy of river water quality simulation and prediction. Twin numerical experiments are conducted using the water quality decay coefficient, initial conditions, upstream boundary conditions, and river pollutant load processes as control variables. The results indicate that the model can extract useful information from observations, rapidly correct the control variables, and identify spatially distributed water quality decay coefficients, initial conditions, upstream boundary conditions, and river pollutant load processes, thereby enabling model predictions to approach the actual water quality dynamics of the river.