Automated, Real-Time Analysis of PV Manufacturing Data Using a “loss component” Approach

As photovoltaic (PV) manufacturing volume grows, so too will the opportunities and potential benefits of sophisticated real time analysis of high volumes of manufacturing data. There are several challenges for any sophisticated analysis gaining widespread use. Firstly, the analysis needs to be run in a continuous and unsupervised way and so the algorithms employed to do this need to be robust. The analysis needs to be done on commonly available end of line data in order to be relevant to the majority of the industry. The information provided by the analysis needs to be useful and interpretable to both specialist and generalist personnel and the overall philosophy should be to provide the right information rather than more information. Both of these last points are essential to the challenges of better utilizing high value specialized staff while also presiding over ever growing volumes of data. This paper describes a new analysis approach that has been developed to address all of these needs. This new approach uses the relationships inferred in standard end of line data to express the performance of each cell as a series of "loss components". Changes in the performance of the line can be directly interpreted through changes in the loss components, leading to an improved understanding of any given change, or problem, immediately when it arises. This paper will introduction the concept of analysis through loss components. It will broadly describe the approach used to calculate the loss components and demonstrate a validation of the approach using simulated data with known loss mechanisms. The paper will conclude with a discussion of the limitations of the approach and a discussion of the future development strategies.