Steam export
Methane slip
1.0
STEAM_EXPORT
NG_FEED
1
0.32
0.73
-0.049
0.17
0.29
1
-0.11
-0.033
0.34
0.37
0.48
0.8
0.6
FG_ CONSUMPTION
SC
0.32
1
-0.069
-0.26
-0.49
0.42
-0.11
1
0.13
0.093
-0.17
-0.72
0.4
COT_TEMP_ DELTA
NG_FEED
0.73
-0.069
1
0.54
0.3
-0.096
-0.033
0.13
1
-0.071
-0.24
0.21
0.2
REFORMER_DRAFT_ PRESSURE
PREREFORMER_ INLET_TEMP
0.0
-0.049
-0.26
0.54
1
0.22
-0.29
0.34
0.093
-0.071
1
-0.093
0.13
-0.2
PREREFORMER_ OUT_TEMP
0.17
-0.49
0.3
0.22
1
-0.21
0.37
-0.17
-0.24
-0.093
1
0.33
FG_LHV
-0.4
REFORMER_ INLET_TEMP
METHANE_SLIP
0.029
0.42
-0.096
-0.29
-0.21
1
0.48
-0.72
0.21
0.13
0.33
1
-0.6
Figure 4 Steam export and methane slip heatmap
the HGU, variables such as the lower heating value (LHV) of natural gas (NG_LHV), airflow to the furnace (AIR_TO_ FURNACE), S/C ratio, high-pressure steam temperature (HP_STEAM_TEMP), LHV of fuel gas (FG_LHV), and the natural gas feed (NG_FEED) have been identified as having a strong correlation with fuel gas consumption. Notably, the addition of the methane slip (METHANE_SLIP) parameter has been observed to enhance the model’s performance significantly. It was discovered during the model development process that an increase in methane slip leads to a higher propor- tion of methane escaping to the purge gas, consequently elevating the purge gas’s LHV. This observation has led to the finding that as the purge gas LHV increases, there tends to be a reduction in fuel gas consumption. This rela- tionship showcases the intricate dynamics between purge gas characteristics and fuel gas usage, highlighting the model’s capability to capture these nuanced interactions for more precise fuel gas consumption predictions. Analysis has confirmed direct correlations between fuel gas consumption in the HGU and variables such as the volume of air entering the furnace, the S/C ratio, and the temperature of high-pressure steam. Additionally, the LHV of consumed fuel gas, the LHV of consumed natural gas, and the volume of natural gas consumed have been piv- otal in enhancing the model’s accuracy. These relationships and their impacts on fuel gas consumption can be further explored in detail on the heatmap provided in Figure 3. In the development process of the model for predicting fuel gas consumption within the HGU, both regression and tree-based algorithms were explored. A decision was made in favour of a regression-based algorithm, due to its demonstrated superior efficacy. The performance of the model, as validated through time-series validation, is suc- cinctly summarised in Table 1.
The results of this model are fed into the steam export model, thereby directly impacting the performance of the steam export model’s outcomes. Consequently, the accu- racy of the fuel gas model’s results is pivotal for the pro- ject’s effectiveness. Steam production model One valuable output of the hydrogen production process is high-pressure steam, which is fed into the refinery’s utility system, representing an energy gain. The quantity of pro- duced high-pressure steam varies with different process parameters. An analytical model has been developed to In the development process of the model for predicting fuel gas consumption within the HGU, both regression and tree-based algorithms were explored understand how the quantity changes with adjustments in these parameters, enabling a deeper insight into the opera- tional efficiency of the HGU. The model for predicting the amount of high-pressure steam exported from the HGU incorporates inputs such as fuel gas consumption, natural gas consumption, and various temperature parameters from the reformer and pre-reformer. Values derived from the fuel and natural gas models serve as inputs to this model. As illustrated in the model’s heatmap, while natural gas consumption is a pri- mary input, the relationship between the LHV and fuel gas consumption and the exported high-pressure steam is also notably strong (see Figure 4 ).
67
PTQ Q3 2024
www.digitalrefining.com
Powered by FlippingBook