Determination of the ash content, volatile matter and calorific value in coal with laser-induced breakdown spectroscopy (LIBS) is significantly affected by the matrix effect, due to the physical and chemical properties of different types of coals. This work aimed to achieve more accurate determination of the ash content, volatile matter and calorific value via the classification of coal samples by their ash contents. The support vector machine (SVM) classification method, which was optimized by using a genetic algorithm (GA), was used to classify the coal samples into three categories. And then, the partial least squares regression (PLSR) was employed to establish different models for each type of coal. With this detailed classification scheme, the coefficient of determination (R²) of the training set and test set of volatile matter was improved from 0.9269 and 0.9310 to 0.9959 and 0.9888, respectively. The root-mean-square error of cross-validation (RMSECV) and the root-mean-square error of prediction (RMSEP) of volatile matter were also reduced from 1.9940% and 1.8320% to 0.4989% and 0.7719%, respectively. The results of the ash content and calorific value are also improved. This work demonstrated that this LIBS-based quantitative analysis method can be widely used in the coal industry to improve current coal property analysis.

The prediction of coal fusion characteristics and determination of slagging behaviour are of great significance to the operation of thermal power plants. The traditional method needs a long time to measure coal ash fusion temperatures (AFTs) by heating the coal ash cone. The purpose of this work is to predict the coal AFTs directly from coal rather than coal ash using Laser-Induced Breakdown Spectroscopy (LIBS). On the premise that ash content, volatile matter, fixed carbon and moisture are known on the air-dry base state, these four parameters are added into the model as generalized spectral (GS) variables to improve the accuracy of coal AFT prediction, in which the R² of softening temperature (ST) and hemisphere temperature (HT) reached 0.9958 and 0.9856, the root-mean-square of cross-validation (RMSECV) reached 4.88 °C and 9.11 °C, and the root-mean-square of prediction (RMSEP) reached 8.15 °C and 11.3 °C, respectively. The trend relationships between the AFTs and the intensities (or ratio) of the spectral lines corresponding to individual elements of the LIBS spectrum were analysed. Then the variation trend of coal AFTs with specific element content (or ratio) is deduced qualitatively. We demonstrated that the model and scheme developed in this work can be applied in a real-time guiding role in the actual production process of the power plant.

The influences of the focusing geometry and axial focusing position of the second laser pulse on orthogonal double-pulse LIBS were investigated.

Coal quality analysis is important to promote the rational and sustainable utilization of coal resources. The traditional method of selecting characteristic lines from the coal sample spectrum is time-consuming. This work aims to use laser-induced breakdown spectroscopy (LIBS) to reconstruct principal components (PCs) by utilizing the correlation between full spectrum wave point interpretation variance and calibration values. More components related to the features of interest are obtained and the regression metrics are maximized. Subsequently, quantitative models were established using ordinary least squares (OLS) for ash content, volatile matter, and calorific value. The coefficient of determination (R²) of the training set and test set of ash content and volatile matter was improved from 0.8125 and 0.8222, 0.6502 and 0.6483 to 0.9701 and 0.9818, 0.9458 and 0.9429, respectively. The root-mean-square error of cross-validation (RMSECV) and the root-mean-square error of prediction (RMSEP) were reduced from 1.7843% and 1.7696%, 1.1797% and 1.4301% to 0.7153% and 0.7037%, 0.5678% and 0.6628%, respectively. The training set and test set of calorific value R² were also improved from 0.7086 and 0.7157 to 0.9857 and 0.9811. RMSECV and RMSEP were reduced from 0.9525 MJ kg⁻¹ and 1.0178 MJ kg⁻¹ to 0.1518 MJ kg⁻¹ and 0.1613 MJ kg⁻¹, respectively. To evaluate the interpretability of PC recombination, the cumulative correlations between ash content, volatile matter, calorific value and the PCs of each wave point in the LIBS spectrum were analyzed, and then the characteristic spectral line locations of the elements that contributed the most to the calibration model were extracted. The results of the analysis of metal and non-metal elements in coal samples show that PC recombination chooses more significant characteristics rather than noise as independent variables. The work demonstrates the feasibility and effectiveness of employing LIBS-based quantitative analysis for rapid coal quality detection in the coal industry.

It is of great significance to analyze the calorific value and carbon dioxide emission level of coal with different degrees of coalification for energy conservation and emission reduction in coal-fired power plants, and elemental carbon and fixed carbon are closely related to the carbon dioxide emission level and calorific value. Due to the complexity of matrix effects and differences in regression tasks, modeling only all coal samples will limit the quantitative performance of laser-induced breakdown spectroscopy (LIBS). In this work, we combine a dual-cycle variable selection mechanism with competitive adaptive reweighted sampling to optimize partial least squares regression (PLSR). According to the difference between coal samples in coal quality characteristics and prediction tasks, the matrix correction of elemental carbon and fixed carbon is realized by qualitative classification and by using generalized spectra (GS), respectively. The coefficient of determination (R²) of the training set and test set of elemental carbon was improved from 0.7178 and 0.7095 to 0.9534 and 0.9515, respectively. The root-mean-square error of cross-validation (RMSECV) and the root-mean-square error of prediction (RMSEP) decreased from 1.1723 wt% and 1.2488 wt% to 0.4504 wt% and 0.4871 wt%, respectively. The R² of the training set and test set of fixed carbon was also improved from 0.8252 and 0.8236 to 0.9814 and 0.9749. The RMSECV and RMSEP decreased from 1.2274 wt% and 1.3548 wt% to 0.4421 wt% and 0.4711 wt%, respectively. According to Chinese national standards and IPCC guidelines, the maximum error in CO₂ emission factors for all coal samples during the accounting period is 1.7581%. Taking into account the level of CO₂ emissions and the net calorific value, the selection of coal with different degrees of coalification can guide the combustion of mixed coal in power plants to achieve high efficiency and clean power generation. This work demonstrates the great potential of in situ LIBS for online applications in coal-fired power plants.

At present, buildings are the major energy consumers in most developed and developing countries. Energy efficiency is one of the primary objectives of today’s building projects, especially in the context of carbon-peak and carbon-neutral. In buildings, air-conditioning systems account for nearly half of total building energy consumption. Optimal control and operation can significantly improve air-conditioning’s energy efficiency but the key challenge is the complex dynamics of air-conditioning systems. This study focuses the energy efficient operation of air-conditioners in office buildings, and proposes a simulation-assisted reinforcement learning (RL) method to develop energy efficient operation strategies. This study uses a whole building energy simulation model to build the environmental simulator for the air-conditioning system, and uses asynchronous RL algorithm to learn energy efficient operation strategies. Energy saving and Setpoint Not Met is used as the optimization objective as well as the criteria for evaluating the performance of the RL operation strategies. A validation simulator with varied weather conditions is also built to validate the robustness of reinforcement learning. The results show that, compared to common rule-based control strategy, 16.1% energy saving with better thermal comfort can be achieved by the RL operation strategy. In addition, the results also show the RL operation strategy has a certain level of robustness.