After describing the process of the back-propagation training of the network, let's now concentrate on the Logic Decision Tool based on ANN models that this paper proposes.
4. Logic decision tool and ANN models
Fig. 2. Work-flow logic decision tree about on-conditions predictions.Figure optionsDownload full-size imageDownload as PowerPoint slide
Consequently, two new mathematical ANN models are developed in this document showing the aptitudes of ANN to replicate reality self-adaptively in complex and noised operating conditions: Case A) of direct ANN in absence of failure data, reproducing Salirasib production of the power inverter which is has a physic complex equation; and Case B) of Survival ANN with enough non-formal failure data and complex covariates interaction, trying to fit the Survival Function of solar trackers.
4.1. Case A) Failure mode prediction: lack of insulation
The selected failure mode of the power inverter is the “lack of insulation” failure, which due to the fact that production losses are significant; this failure mode is considered in SCADA with priority. This failure mode emerges due to corrosion and, the environmental conditions could be determinants in different areas and besides the inverter operating time. The most representative variables of operation, external environment and internal conditions have to be selected and tested to show their effects in the failure mode. The available variables in our SCADA in the case of power inverters are: ambient temperature (°C), the internal temperature of the power inverter (°C), the global horizontal radiation (W/m2), the operation time of the power inverter (h), and the active energy accumulated of the inverter (kWh).