The quality of field development optimization is largely determined by the quality of the oil reservoir model used. The oil reservoir model is built by solving the inverse problem (automatic history matching problem). For this purpose, the model is represented as a set of 3D sub-areas for which rock and fluid properties are determined by minimizing a functional that includes weighted squares of deviations between observed and simulated data (oil flow rates, cumulative volumes of produced oil, water cut, pressure) from wells during field development. In the proposed approach, well data for the reservoir model and sensitivity to reservoir model parameters are obtained by finite element 3D modeling of multiphase flow, which provides high quality model matching and forecasts, but requires quite high computational costs. We propose a way to increase the computational efficiency of solving the inverse problem through automatic partitioning of the computational domain. The method makes it possible to calculate the sensitivity of production data to the parameters of the reservoir model on small meshes constructed in a special way practically without compromising the accuracy of the solution. The studies of the accuracy and computational efficiency of the developed method, which are conducted by using practical data from high-viscosity oil fields, showed that the proposed approach does not degrade the quality of the resulting reservoir models and, at the same time, allows us to speed up their construction by 3 times even for relatively small fields. Note that the construction time for reservoir models takes several days, so this acceleration is significant. As the size of the deposits increases, the effectiveness of the proposed approach becomes even more significant. The resulting reservoir models can be used in optimization procedures to construct a plan for further development. We propose a special objective function including weighted squares of target development indicators and regularizing additives with adaptive search of coefficients. The target indicators and their sensitivity to the parameters to be optimized are calculated by finite element 3D multiphase flow simulation for the history-matched reservoir model. Using the example of one of the fields, we show that optimized plans built on the basis of the history-matched reservoir model can provide more efficient oil field development in comparison with a plan used in practice, where decisions on changing modes were applied during development. In addition, the optimization approach allows the analysis of different strategies and options, including in changing economic conditions. The proposed history-matching and optimization approaches reduce the development design effort and improve the quality of decision-making.