This tool overcomes the limitations of the equivalent tool used in Plexos software. The idea is to make a projection – which is different from a forecast – of a half-hourly annual profile for future years, taking into account assumptions about annual peak and consumption. By playing with these three assumptions (shape, peak and energy), the tool generates different demand scenarios. For example, this is useful for creating price paths used by generators in investment studies.
The tool has many advantages over what coworkers were using previously. Firstly, the projection is significantly improved with better mapping of the base profile and public holidays, as well as better preservation of the integrity of the shape when large distortion is required (more intelligent distortion function). Secondly, the tool provides a better understanding of assumptions thanks to a well-designed app with carefully crafted visualisation elements, automatic testing of assumption consistency, and the ability to project in different modes if the assumptions contain inconsistencies. Finally, this app offers greater control and transparency over what is happening, with code that is fully available to company members.
The application was developed using Python, Github, and Snowflake.

One of the country’s largest energy companies has launched the construction of a huge BESS. In anticipation of its use, they asked us to develop a tool to simulate the use of BESS in different power markets. This allows traders to practise optimising their bids. The tool took the form of a comprehensive app integrated into the client’s information system. There were three of us working on the project, plus two people on the client side. The application was developed using Python, SQL, Docker and Azure.

An australian gentailer was co-developing an EaR (Earnings at Risk) calculator with us. This metric is designed to evaluate the market risk associated with the portfolio exposure to spot price, including retail load, generation and financial products in the middle to long term.
My contribution was mainly to put together a study of the codebase in a pre-release version. This work required me to understand the code in detail in order to assess compliance with best practices (implementation, readability, error handling, etc.), make recommendations and validate the functionality of the code. The checklist developed for this purpose fuelled internal discussions on improving code quality in projects.

Development of various applications for internal use in energy portfolio management using Python and VBA.

Complete development of an application for load forecasting for businesses with a contracted power between 36kVA and 250kVA. This tool takes load curves as input, each corresponding to a POD (Point Of Delivery), and makes forecasts in record time (it processes 2000 curves in less than 25 minutes). The tool has been designed for efficiency and is based on a simple model. For each POD, the model performs a linear projection of a base profile built from the POD’s load curve, de-correlated from temperature. It also handles edge cases where data is missing or inconsistent. This tool was quickly operational and helped win several tenders.
Excel, Python, AWS.

Complete development of a tool for forecasting dynamic coefficients for profiles with a contracted power of 36kVA or less. The tool has provided a substantial gain in accuracy compared to the previous tool, which used static coefficients with a statistical method for forecasting, especially in a context of declining consumption due to the 2022 energy crisis. A variation in forecasting impacts the pricing of offers via a variation in the ARENH rate and therefore in the cost of sourcing, the cost of flex and the cost of capacity. The study of this impact demonstrated that the new tool’s forecasting provided a competitive advantage. The tool operates autonomously, launching on the local service machine at regular intervals to update data and forecasts.
Python, Bash.