The invention discloses an intelligent breathing training method for a critical patient, which comprises the following steps: S1, a medical worker inputs patient conditions including age, gender, occupational condition, illness condition and treatment condition information of the patient into a central control computer, and sets a breathing training evaluation standard; s2, acquiring the heart rate of the patient in real time through an electrocardio sensor worn on the body of the patient, acquiring the oxyhemoglobin saturation and respiratory rate of the patient in real time through a pulse oxygen sensor, and acquiring high pressure, low pressure and pulse through a blood pressure sensor; s3, through an artificial intelligence algorithm, according to the condition and time of the patient, a treatment oxygen uptake condition is formulated, an oxygen solenoid valve is autonomously controlled to adjust the oxygen concentration, a heater is controlled to adjust the oxygen temperature, an atomizer is controlled to adjust the oxygen humidity, and the respiratory function of the patient is further exercised. According to the method, the breathing exercise of the critical patient is mainly carried out autonomously according to the disease recovery condition of the patient, the human body biological clock and the surrounding environment factors, and the speed of rapidly improving the health recovery of the patient is achieved.