Abstract Gaslift operation faces several challenges - optimizing the amount of gas pumped, ensuring that the gas is pumped to the maximum possible depth, combating salt and hydrate formation and, finally, the instability of the flow. The timely detection of these problems and the optimisation of the system requires information on the dynamics of the main parameters (flow rates, pressure measurements at each unit and the amount of gas pumped). In the conditions of high cost of multiphase flowmeters at the Orenburg field well flow rates are not measured often enough, and the task of identification of the most unstable wells in real time by auxiliary parameters measured with greater discreteness becomes urgent. Often important production decisions based on monthly data, and less often on daily measurements. In the paper the necessity to go down to the level of analysis of highly discrete information (dozens of measurements per day) is proved by the example of the Orenburg field, because in conditions of unstable operation the averaging inside the day can lead to distortion of the general view of the operating mode. Within the framework of the analysis of dynamic parameters of gas-lift wells operation in unstable conditions, the notion of well operation quality index was introduced, i.e. a normalized metrics reflecting the general state of stability of a gas-lift well operation. This metric considers uncoordinated changes in the main parameters of the system. The wells with the lowest index are the first candidates to search for hidden production losses (not achieving the regime flow rate). Modelling of well performance, adaptation of models to historical data - a typical approach to production management in the field. But in the conditions of unstable gas-lift wells modeling of their work is much more complicated. In this regard, the authors of the article developed an approach to modeling wells through the approximation relationship between the flow rate of wells and the flow rate of gas-lift gas and linear pressure with further integration of well models with the collection system. To solve well operation modes optimization problem, an integrated Proxy-model of the asset was created, which consists of a simplified reservoir model, developed well models and infrastructure model. The Proxy model was adapted and verified to the actual operating data. The task of finding the optimal distribution of gas injection between the wells from the point of view of oil production was solved.