US Patent 11933678 Temperature sensor calibration using recurrent network

A method for determining ambient temperature of a device in a building control system. The method includes obtaining a measurement of ambient temperature from a first temperature sensor of the device and obtaining a measurement of temperature of a processing circuit of the device. These temperature measurements along with a brightness level of a display of the device and a state of a relay of the device are provided as input to a neural network. The neural network is trained to recognize these inputs in order to more accurately determine the ambient temperature of the device. The reading from the first temperature sensor may otherwise be inaccurate due to heat generated by electronic components of the device.