Edge-gateway acquisition and cloud monitoring of a multi-inverter rooftop photovoltaic system in a zero-carbon park: vendor-independent design and field deployment

Abstract

Rooftop photovoltaic (PV) generation is a core contributor to zero-carbon parks, but the inverter fleets that produce it are often locked inside a manufacturer’s closed cloud, so the owner cannot freely access, aggregate or audit the data. This paper reports the design and ongoing field deployment of a vendor-independent edge-gateway and cloud-monitoring system for a 15-inverter rooftop PV installation in a building-integrated PV (BIPV) project in Wuhan, China. The fleet comprises three Growatt models (10000UE, 20000UE and MID 20KTL3-X). The originally installed OEM datalogger was bound to a closed platform whose backend was inaccessible and whose reporting address could not be changed, blocking redirection of the data to the owner’s self-hosted platform. To recover data ownership, an independent industrial IoT gateway is deployed as an RS-485 Modbus-RTU master that polls the inverters over two buses and publishes to a self-hosted server over MQTT/HTTP via a cellular link. The cloud platform stores the telemetry, computes generation, specific yield and inverter efficiency, and applies threshold, peer-comparison and communication-status checks on a zero-carbon dashboard. At the time of writing, 3 of the 15 inverters are integrated and uploading data, while the remaining 12 are offline or pending verification of wiring, Modbus address, model and rated-capacity information. The integrated subset has recorded about 399.9 MWh cumulatively on the platform. Full-fleet clear-day, overcast-day and carbon-reduction figures are therefore presented as preliminary estimates derived from the integrated units under an explicitly stated comparable-output assumption, not as measured 15-inverter totals. We also describe a field case of an inverter that displays a fault while still generating. Irradiance and module-temperature sensors are not yet installed, so the performance ratio is left to future work. The system demonstrates a practical, reproducible path to vendor-independent PV monitoring.