Abstract: Aiming at the problems existing in the SCR De-NOx system of a 300MW coal-fired unit during deep peak shaving—specifically low flue gas temperature at low loads, significant hysteresis in NOx monitoring signals, and insufficient CEMS data representativeness, all under the conditions of no economizer bypass, no flue gas bypass, and without the use of wide-temperature catalysts—this paper conducts research on the optimization of full-process De-NOx monitoring.A multi-dimensional monitoring network covering the SCR inlet full cross-section, reactor interior, outlet full cross-section, and chimney emission end was established to analyze the coupling variation laws of flue gas temperature, flow rate, and ammonia-nitrogen molar ratio across a wide load range, thereby identifying the main sources of monitoring deviation under low-load conditions.A Smith prediction compensation mechanism was introduced to address the CEMS sampling transmission hysteresis. Combined with anti-interference filtering and wide-condition adaptive correction methods, a multi-algorithm fusion optimization model was constructed to enhance the stability and measurement accuracy of the monitoring system under harsh, low-load conditions.Field tests demonstrated that after optimization, within the 30%–100% full load range, the CEMS data validity rate increased from 80.5% to 95.8%, the NOx concentration response time was reduced by over 60%, and the absolute error of ammonia escape monitoring was controlled within ±0.2 ppm.This scheme achieves precise wide-load De-NOx monitoring merely through improvements in monitoring logic and algorithms, avoiding the need for extensive equipment retrofitting. It effectively reduces the risk of environmental non-compliance and provides engineering references for the compliant operation of similar bypass-free coal-fired units, offering theoretical and technical support for De-NOx monitoring optimization in deep peak shaving scenarios.