Relevance. Plunger units, in particular rod and rodless ones with surface and submerged drive, are widely used in the oil industry for extraction of high-viscosity emulsion from marginal wells. The main advantage of such units is the possibility to operate in conditions of highly viscous emulsions with viscosity values up to 200–300 cPs. However, there are peculiarities of plunger unit operation, which lead to some complications. One of the most widespread complications is high wellhead pressure, which leads to increased dynamic loads on pump and drive units. Qualitative and precise definition of wellhead pressure value depending on well and unit parameters will contribute to more effective design, operation and evaluation of plunger unit reliability. Known studies and methods of wellhead pressure determination have limitations related to single-factor influence of some parameter on wellhead pressure values. This approach is not suitable for taking into account the simultaneous influence of several parameters. Aim. Development of calculation methodology, which allows taking into account complex influence of well operation parameters on wellhead pressure value and dynamic loads on plunger rods. Objects. Fund of wells equipped with borehole rod pump units. Methods. Statistical and regression analysis of the stock of wells equipped with downhole rod pumping units, development of a neural network. Results. A stock of wells equipped with downhole sucker-rod pumps, the technological mode of which is provided by one of the companies of Privolzhsky federal district, was investigated as an example, on the basis of the specified stock the factors influencing the value of wellhead pressure were established and the technique on its forecasting was developed. The authors have proposed the method for discarding the data with a wide range of values when analyzing the factors affecting the wellhead pressure. The method consists in excluding the values of fluid viscosity not lying on the "hump" characteristic of the function "fluid viscosity – watercut". The authors developed the calculation method based on regression dependences and neural network. It allows forecasting more accurately the value of loads on the pump rods.