 Hi, my name is Meder Givanovich and I'm the VP Technology at TDE. In this video, I will present a case study of how automated well operations planning, supported by OSTU, can help you to introduce consistency and scalability in your well planning process, so to save time and cost. But before diving into the technology, let's have a short introduction of our company. TDE is a data engineering company founded in 2000. Since then, we have focused on turning real-time data to value for customers in the upstream oil and gas sector. TDE have processed real-time drilling data of more than 50,000 wells globally and we are looking to transfer this experience also to other fields of the energy industry. In our sector, we have pioneered with software as a service and cloud offerings and we are working with AWS for many years and recently we also became a partner in the AWS Energy Competency Program. During these 20 years of business, we have not only focused on real-time monitoring of drilling operations, but also on the planning aspect of it, especially the operations sequence planning. And we have observed that our customers share some of the same issues. Although there is software to support the operations planning, the process is still very manual, which makes it lengthy, costly and inconsistent. The reason why the process is still so manual is the lack of reference data. Either it is not there at all, or it is sold in applications or databases and inaccessible. Eventually, these issues lead to over or unallocated budgets, which incur high cost of capital. Now at TDE, we developed a software to meet these challenges and enable an automated, fast, consistent and reliable operations planning process, where the data is liberated by the OSDU platform. The software predicts technical limits by using artificial intelligence and physics-based modeling, and it defines KPI targets for the actual operation. With this, it improves risk management and budget allocation. Now here you can see the main interface of the software. The data input requirement is minimal, merely a well name, a simple well schematic, and then the user needs to select a rig. From there, the system takes over automatically and computes the operation sequence with P10, P50 and P90 time and cost estimates. The operation sequence is modeled in multiple levels of detail. The activities track at the bottom shows phases, runs like drilling or casing runs, and even in more detailed main operations which split the runs into activities, like making a PHA, tripping in stand, filling a pipe or drilling a specific formation. On the sidebar, you can see some of the many parameters which the software uses for the time estimation. Now most of these parameters are automatically derived from the rig sensor data. However, one specific parameter required is the netROP, which is the actual speed of penetrating the rock. This single KPI defines an average 30% of the total drilling time, and it is heavily dependent on the formation, the tool selection and operating parameters. Thus, we do not want to rely on the plain offset data only. We want to introduce also a more sophisticated method to predict this KPI. In the project of AWS, we focused on our machine learning based approach to model the ROP. The machine learning model requires not only raw and processed real-time sensor data, but also other input data like trajectory service, geological information and P10 information. This is where we saw potentially the OSTU data platform as a central place to access all of this data in a unified and structured way. In addition, we want to write back the results in form of planned activities to OSTU, so it can be used in other applications, for example in a drilling advisory system in the field. In the future, we will also implement physics-based modeling based on UCS to estimate the theoretical maximum ROP achievable, which is limited by the geology. For this, the system needs even more input data like ultrasonic logs, which will be also retrieved from OSTU. Now here is a quick look on the data flow of the solution. The OSTU data platform enabled us to split the data from the application. We ingest the input data as files into the OSTU, making it indexable and searchable on the platform. On the application side, the data can be searched and downloaded with an OSTU connector tool, and then it is further ingested into the ROP app microservice, where it is used for training the machine learning model and for modeling the ROP. For this six-week proof-of-concept project, we chose to create a simple Windows UI to support the process. In production, the process would be controlled through the main planning interface, which you have seen before. The actual ROP modeling application is a Python microservice, which communicates via APIs with the UI. In step one, we are searching the offset data on OSTU and downloading it. In step two, we ingest this data into the ROP app, and we run some pre-processing algorithms, and we also configure the planned well. In step three, we launch the actual machine learning prediction for the planned well and inspect the results. The app automatically selects the best-fitting machine learning algorithm, and in this plot you can see how the modeled ROP fits the actual. From there, the system would take the ROP, averaged by formation depth interval, and to the planned operation sequence to compute a more precise time-cost depth estimate. So to sum up, OSTU provides a reliable and standardized way of getting the various input data from multiple disciplines into the ROP modeling app. It eliminates the need to build several connectors to individual data sellers to collect reports, logs, and so on, and it offers a standardized way to provide the planned activities back to the other applications. Now this is a schematic of the system architecture. TDE planned with the ROP microservers is making use of the most state-of-the-art cloud technology, making it super scalable and flexible at the same time. It uses APIs to connect to the cloud agnostic architectures of the ProNova platform and the OSTU data platform. So finally, we took the following conclusions from the POC project. Automated well-operations planning significantly mitigates planning risks and improves budgeting and there's an industry need for this. Machine learning-based automated ROP modeling adds value to this process. The OSTU data platform proved good as data integration and exchange platform, and the well-known SKYMUS proved okay for this use case. Drilling states and performance KPIs from ProNova can be made available and searchable on the OSTU data platform. So that was the final slide. If you experienced the same challenges in well-operations planning, or you just want to learn more about TDE, contact me directly, or check out our website where you can find contact details of our regional offices. My name is Miodra Givanovich. Thank you for watching this video.