Ambue scaling retrofit with graph algorithms

点击翻译按钮获取中文摘要

**Improving property portfolio sustainability analysis with graph machine learning**There are a number of key hurdles in upgrading the United Kingdom's property stock, and then delivering sustainable retrofit at pace and scale. One is the securing of an accurate picture of its existing state so as to make timely, cost-effective and targeted retrofits to each property. Another is accessing private finance to execute these projects once they have been identified.Registered Providers ("RP's") have become dependent on unreliable, gamed and increasingly-inaccessible data to make these investment decisions, whilst blanket on-site surveys are time-consuming and disruptive. Often, this data is used to produce deterministic energy assessments via methods such as SAP, which have known limitations in terms of real-world performance gaps, and through which the quality of initial data and associated uncertainty is not materially quantified.Building retrofit strategies on the back of such stock assessment propagates unquantified project risk due to the quality of feedstock data, and increases the likelihood that desktop assessed projects will need to be significantly reworked upon reaching detailed design and procurement. This risk can be mitigated through careful archetyping of the project stock and confirmatory surveys, but this is resource intensive, expensive and can slow down project delivery.The project is to model in enhanced detail, a sub-set of a client's property portfolio, including developing a customised retrofit pattern book, with enhanced property measurements and surveys to significantly improve the accuracy of the baseline and forecast data for these properties. We will then use Graph Algorithms (machine learning) to extrapolate the findings to apply the improved analysis and forecasts, to the whole portfolio. This will enable greater confidence in the data and analysis, to in turn have better confidence in the forecasts and retrofit improvements and outcomes.