Under the background of the digital transformation of water resources asset management, the traditional supervision model faces the problems of multi-source data fragmentation, lagging accounting verification, and difficulty in coordinating economic benefits and ecological constraints. This paper constructs a collaborative optimization model of accounting supervision of water resources assets and water conservancy economy powered by digital twin technology, and forms an overall framework around “physical basin system–digital twin virtual mirror–data-driven decision-making layer”. The model integrates hydrological monitoring, project operation, accounting books, tax payment and water quality ecological data, and realizes the dynamic mapping of water resources assets through standardized processing, spatio-temporal fusion and virtual mirror status update. At the regulatory level, the accounting deviation calculation, dynamic threshold, isolated forest anomaly score and comprehensive risk index were introduced to complete the intelligent identification of water withdrawal, water charges, taxes, water quality and asset amortization. At the optimization level, NSGA-II and reinforcement learning are combined to construct a collaborative decision-making model that takes into account economic net benefits, ecological health, water balance, water quality constraints and accounting compliance. The experimental results show that the accuracy of model supervision reaches 83.3%, which is 12.1 percentage points higher than that of manual auditing, and the false negative rate is reduced to 6.7%. The collaborative optimization scheme achieved 112.7 billion CNY cumulative net economic benefits, the ecological health index remained at 0.83, and the collaborative score remained stable in sensitivity analysis. The research can provide technical support for the fine supervision of water resources assets, the optimization of water conservancy economy and the collaborative management of basin ecology.
