A Digital-Twin-Based Intelligent Operation and Maintenance System for Photovoltaic Power Stations

Abstract

Operation and maintenance of photovoltaic (PV) power stations is conventionally reactive: faults are found by periodic manual inspection, so the interval between a performance loss and its diagnosis can be long, and the energy lost during that interval is unrecoverable. This paper presents a low-cost intelligent operation-and-maintenance (O&M) system in which a digital twin — a cloud-resident virtual counterpart of the physical installation — is continuously updated from field sensors and used to interpret plant behaviour. The bench-scale implementation combines a solar panel with a rectifier diode, a 3.7 V battery bank, a 7805 regulator, a voltage sensor, a light-dependent resistor (LDR), a rain-drop sensor and a NodeMCU ESP8266, with remote visualisation through the Blynk IoT platform. Three environmental conditions were instrumented and recorded. Under normal light with no rain the panel delivered 10.44 V; under low light with rain the output collapsed to 0.43 V; and under low light without rain it recovered only to 1.08 V. These three measurements are the complete recorded dataset and are reported as such: they establish that the system correctly distinguishes a productive state from two distinct unproductive states, and that low irradiance rather than precipitation is the dominant factor, since the no-rain low-light case remained more than an order of magnitude below the normal-light case. Because irradiance was not logged in physical units and no measurement was repeated, no efficiency figure, no power curve and no quantified predictive-maintenance accuracy can be derived from this dataset, and none is claimed. The contribution is a demonstrated, fully traceable sensing-to-dashboard path together with an explicit statement of the additional instrumentation required to make the assessment quantitative.