Automation control of desulfurization system: intelligent algorithm innovation for real-time monitoring of pH of magnesium hydroxide slurry

1. Pain points of traditional pH monitoring and breakthrough direction of intelligent algorithm

In the wet desulfurization process of magnesium hydroxide, the pH value of slurry is both the "baton" of desulfurization efficiency and the "barometer" of system stability. However, traditional monitoring methods face three challenges:

Short electrode life: high concentration of magnesium sulfate crystallization and dust wear in slurry result in an average life of pH electrode of only 3-6 months, and the annual maintenance cost exceeds 200,000 yuan/set;

Measurement hysteresis: offline sampling and analysis takes 15-30 minutes, which is difficult to match the instantaneous fluctuation of flue gas SO₂ concentration (the fluctuation range can reach ±200mg/m³);

Rough control: the response delay of manually adjusting the opening of the slurry supply valve is as long as 10-20 minutes, which is easy to cause pH value overshoot (fluctuation range of more than ±0.5), exacerbating the risk of scaling in the tower.

The introduction of intelligent algorithms is reconstructing the automation control paradigm of desulfurization system through the dual-wheel drive of "soft measurement + predictive control". Based on the measured data of a 660MW unit, the intelligent algorithm can improve the pH control accuracy to ±0.1, improve the stability of desulfurization efficiency by 12%, and save 8%-15% of magnesium hydroxide consumption annually.

2. Intelligent algorithm technology architecture and core modules

1. Soft measurement model of multi-source data fusion

Break through the dependence on a single electrode and build a **"physical measurement + data-driven" hybrid sensor network**:

Hardware layer: Use slurry density/pH joint measurement device, reduce electrode wear through DN20 micro-flow sampling (flow <8L/min), differential pressure density compensation (accuracy ±10kg/m³), and extend the calibration cycle to 90 days1;

Data layer: Integrate DCS real-time data (slurry circulation volume, oxidation air volume, inlet SO₂ concentration) and LSTM time series prediction model to establish a dynamic digital twin of pH value2. When the electrode fails, the model can still output the predicted value through 14 auxiliary variables such as flue gas flow (Qgas) and calcium-sulfur ratio (Ca/S), with an error rate of <2%.

2. LSTM-ARIMA hybrid prediction algorithm

A hierarchical prediction architecture is designed to target the **“large inertia + nonlinear” characteristics** of pH value changes:

Long-term trend capture: LSTM network (hidden layer 128 nodes) is used to learn historical 72-hour data to extract deep features such as slurry buffer capacity and magnesium ion dissolution rate;

Short-term fluctuation correction: ARIMA model (p=3, d=1, q=2) is superimposed to correct the prediction deviation caused by sudden changes in flue gas load in real time, reducing the 15-minute prediction root mean square error (RMSE) to 0.082.

3. Multi-objective optimization control engine

Develop MMPC (multi-model predictive control) algorithm to achieve coordinated optimization of slurry supply and oxidation air volume:

Objective function: min{J=α·(pH-pH_set)² + β·ΔQ_slurry + γ·C_gypsum}, where α, β, and γ are weight coefficients, corresponding to desulfurization efficiency, magnesium hydroxide consumption, and gypsum quality, respectively;

Constraints: pH∈[5.8,6.5], slurry density <1250kg/m³, oxidation rate >95%36. Through rolling time domain optimization, the frequency of slurry valve opening adjustment is reduced from 5 times/hour to 1.2 times/hour, and the valve life is extended by 3 times.

III. Engineering practice and benefit verification

The application case of a 660MW ultra-supercritical unit in Ningbo shows:

Indicator Manual control Intelligent algorithm control Improvement

pH fluctuation range ±0.5 ±0.15 70%

SO₂ emission compliance rate 92% 99.7% +7.7%

Magnesium hydroxide unit consumption 0.85t/h 0.73t/h -14%

Gypsum crystallinity 80% 93% +16%

Annual maintenance cost 1.85 million yuan 1.02 million yuan -45%

The core innovation of this system lies in **"three-order closed-loop control"**:

Perception closed loop: The joint measurement device transmits density/pH value to DCS every 60 seconds, and synchronously triggers the electrode self-cleaning program (high-pressure water jet + ultrasonic oscillation);

Prediction closed loop: The LSTM model updates the pH trend of the next 30 minutes every 5 minutes to guide the slurry feedforward adjustment;

Execution closed loop: The fuzzy PID controller dynamically adjusts the P, I, and D parameters according to the prediction deviation, and the response time is shortened to 45 seconds.

IV. Technology evolution and industrial value

1. Edge computing and cloud-edge collaboration

Deploy embedded AI chips (such as NVIDIA Jetson series) to compress the prediction model inference time from 2.1 seconds to 0.3 seconds, meeting the millisecond response requirements of the ship desulfurization system.

2. Deep application of digital twins

Build a digital twin of slurry reaction dynamics, optimize the spray layer layout through CFD simulation, reduce the liquid-gas ratio from 8L/m³ to 5.5L/m³, and reduce power consumption by 18%.

3. Collaborative upgrading of the industrial chain

Upstream: Localization of high-precision pH sensors (accuracy ±0.01), breaking the monopoly of European and American companies;

Downstream: Output of "desulfurization efficiency guarantee + by-product value-added" combined services, gypsum purity increased to 98.5%, can be directly used in medical gypsum production.

The intelligent upgrade of magnesium hydroxide desulfurization system is essentially a deep integration of "process mechanism + data science". When LSTM network meets magnesium ion mass transfer equation, when predictive control encounters buffer capacity dynamics, an algorithm-driven desulfurization efficiency revolution is accelerating. It is estimated that the global intelligent desulfurization control system market size will exceed US$5 billion in 2028, of which pH real-time monitoring modules account for more than 60%. In this transformation, companies that master core algorithms will dominate the industrial landscape of the next generation of environmental protection equipment.