A beautiful partnership story... with Engie Green

The need

The Engie Green company is in the business of design, development, construction and maintenance of all types of RES electricity generation processes, in particular wind farms. Wind farms are required to comply with legislation imposing limits on noise exposure of local residents. The only solutions currently available to achieve this consist very often in reducing the farm’s activity, by slowing or stopping the turbines when necessary. The major disadvantage of the current solutions is that they are relatively rigid. They do not adapt to the real background noise of the site which may change over a period of time. These solutions may therefore sometimes be too restrictive, resulting in unnecessary limitation of energy production not justified by the noise levels emitted. In order to solve this problem, Engie Green asked UMRAE – joint research unit into environmental acoustics, a Cerema and Université Gustave Eiffel (UGE) research laboratory – to develop a system to automatically estimate the noise emergence, in order to allow virtually real-time control of the slowing system controlling noise emissions.

The partnership

A close partnership was set up between Engie Green and the UMRAE research team. UMRAE brings its expertise in developing experimental methodologies for noise characterisation, particularly wind turbine noise, and automatic recognition of environmental noise sources. Engie Green brings its experience as a developer and operator of wind farms. For the duration of the project, the partnership has been able to recruit an acoustics researcher specialised in artificial intelligence. Thanks to the many interactions between the research team and the industrial partner, the knowledge and data required in order to develop the process have been shared very effectively. The partnership contract was initially for one year, but it has been renewed for an additional year, since an encouraging feasibility study of the proposed concept was completed during the first year. The second year is devoted to increasing the system’s reliability, in order to produce an operational industrial prototype.

The innovation

The aim of the innovation is to estimate the contribution of a wind farm’s noise to a measured acoustic signal, and to indicate virtually in real-time whether or not the wind farm is in compliance with the regulations. An artificial intelligence algorithm based on an automatic learning method estimates the acoustic contribution of a wind farm, within a sound signal measured by a sensor placed in proximity to the wind farm. The initial tests showed that the system was capable of evaluating the noise emergence of a wind farm during continuous operation, with a level of uncertainty and accuracy comparable to the level obtained by means of the method used traditionally by acoustic research departments. However, compared to the traditional method, the major advantage of the new process is that no manual analysis by an acoustics expert is required, and it is not necessary to stop the wind farm several times, which inevitably leads to major production losses. The first results of the partnership were communicated at international conferences and will be published in a scientific journal.