Kyatasan
Building on more than 20 years of experience in the energy industry, I explore and explain technologies out of pure curiosity, together with AI agents. Because everything here follows Kyatasan's own interests, it is far removed from the latest news. And since I create as the mood takes me, the topics have little connection to one another.
Research and commentary at Kyatasan's whim
On data-driven models as a substitute for reservoir simulators
A lecture that—framed by the computational cost of conventional reservoir simulation—organizes the need for proxy/surrogate models and their classification (statistical methods / machine learning / advanced methods).
A lecture on building surrogate models and validating their accuracy—from response surface methods and regression, to machine learning such as random forests and neural networks, to advanced methods like ROM and Fourier Neural Operator.
A hands-on exercise guide using Equinor Volve field data, covering Python/Jupyter setup, data loading, preprocessing, visualization (EDA), and feature preparation.
A hands-on exercise guide that builds surrogate models with regression, random forests, and neural networks from Volve production and pressure data, through to accuracy validation.
A record of actually working through Exercise 1, summarizing insights unique to real data—such as the reality of missing data on a per-well basis.
A record of Exercise 2: building surrogate models with RSM, random forest, and NN, and confirming that accuracy differs greatly between random and time-series splits.
Checking a 2003 paper's outlook on methane hydrate against reality
A verification document that checks the four claims of a 2003 PhD thesis—MH technology development, energy markets, climate policy, and Japan's energy security—against the facts as of August 2026.
A detailed technical report that compares the thesis's evaluation of production methods (depressurization / hot-water injection) and its technical targets against 23 years of technology trends, including production-test results.
On the potential of optimization through integrated simulation
A review of integrated analysis methods (IPM/IAM) that solve reservoir simulation, wellbore flow, and surface facilities as one—from definition and overview to coupling mechanisms (nodal analysis, VFP tables, explicit/implicit coupling), the current status of application to digital oilfields and CCS, and the value and challenges of adoption.