Steam Volume Prediction for Industrial Boilers Group 88

Steam Volume Prediction for Industrial Boilers Group 88 Bolin He Xiaoyuan Jiang Yidong Li 2020/06

Background Boilers are essential devices in diverse aspects of industrial procedures such as electric power networks and transportation engineering. Figure 1 Typical diagram of a coal-fired thermal power station

Background The efficiency of boilers, which is also called total thermal energy, is complicated and hard to be determined, since boiler loss is caused by various parameters including combustion air feed rate, radiation loss, and ambient temperature. Figure 2 A Rankine cycle with a two-stage steam turbine and a single feed water heater.

Literature survey

Literature survey

Dataset Table 1 First 5 rows of the training data Table 2 First 5 rows of the test data

Feature Extraction Figure 3 Data distribution of V 0, V 1, … , V 11

Feature Extraction features with poor correlation should be dropped because it means they are not very relevant. Figure 4 Correlation coefficient matrix

Feature Extraction Box-Cox transformation: Figure 5 The influence of Box-Cox transformation on the data distribution

Feature Extraction Figure 6 Target before Logarithm Transformation Figure 7 Target after Logarithm Transformation

Feature Extraction Figure 8 Outliers visualization

Model Ø Linear Regression

Model Ø Gradient Boosting Regression

Model Ø Ridge Regression

Model Ø MLP Regression Figure 9 An MLP Example

Results Ø Linear Regression Score: 0. 8926894125981535 MSE: 0. 11053158993152493 Ø Gradient Boosting Regression Score: 0. 8917557871111966 MSE: 0. 11149323884215377 Ø Ridge Regression Score: 0. 8911242195088429 MSE: 0. 11214376338896286 Ø MLP Regression Score: 0. 8930182569721768 MSE: 0. 11019287506302151

Future Plan Ø Investigation on model parameters Ø Explore new model Ø Explore new data preprocessing method Ø Organize all these plots, figures and results

References [1]F. Steingress, Low pressure boilers. 4 th ed. (includes cd-rom). American Technical Publishers, 2001. [2]"Prediction of Boilers Emission using Polynomial Networks - IEEE Conference Publication", Ieeexplore. ieee. org, 2020. [Online]. Available: https: //ieeexplore. ieee. org/abstract/document/4054835/. [Accessed: 14 - Apr- 2020]. [3]"Real-Time Boiler Control Optimization with Machine Learning", Ground. AI, 2020. [Online]. Available: https: //www. groundai. com/project/real-time-boiler-controloptimization-with-machine-learning/1. [Accessed: 08 - May- 2020]. [4]"MLP Principle Introduction", Weizn. net, 2020. [Online]. Available: http: //www. weizn. net/? post=253. [Accessed: 01 - Jun- 2020].

Thank You!