Small index multi-dimensional energy efficiency analysis method based on online detection of coal quality into the furnace

Abstract

This application discloses a method for multidimensional energy efficiency analysis of small indicators based on online detection of coal quality entering the furnace. It relates to the field of data analysis technology and solves the technical problem that existing technologies often ignore real-time fluctuations in coal quality entering the furnace and use fixed benchmark values ​​for evaluation and optimization, leading to low accuracy in small indicator energy efficiency evaluation and optimization. The method generates dynamic benchmark values ​​based on coal quality data; generates multidimensional energy efficiency indicators based on the dynamic benchmark values ​​and operational small indicator data; and constructs a multidimensional energy efficiency correlation graph to generate optimal analysis results. This corrects the evaluation benchmark from a fixed value to a dynamic value that adjusts with real-time coal quality, eliminating the interference of coal quality fluctuations on the evaluation. The method utilizes a data model to reveal the complex mapping relationship between operating parameters and multidimensional energy efficiency, determining the optimal operating strategy for the current coal quality in real time, thus improving the accuracy of small indicator energy efficiency evaluation and the timeliness and accuracy of optimization decisions.

Description

Technical Field

[0001] This application belongs to the field of data analysis technology, specifically a multi-dimensional energy efficiency analysis method based on small indicators of online detection of coal quality entering the furnace. Background Technology

[0002] Against the backdrop of energy structure transformation, thermal power generation, as the cornerstone of electricity supply, is crucial in terms of operational efficiency and environmental protection. Energy efficiency analysis and operational optimization of thermal power plants are key to energy conservation and emission reduction. Traditional energy efficiency analysis often uses fixed benchmark values ​​for the design coal type or agreed-upon coal type to evaluate operational levels, or manages discrete indicators such as flue gas temperature, fly ash carbon content, and main steam pressure. However, the quality of coal fed into the boiler flues in actual operation, directly affecting boiler combustion characteristics, auxiliary equipment power consumption, and pollutant generation. This leads to distorted energy efficiency evaluations under fixed benchmarks, and the management of small indicators lacks sufficient guidance due to the lack of dynamic correlation with coal quality.

[0003] Existing technologies often ignore real-time fluctuations in the quality of coal entering the furnace when conducting energy efficiency assessments and optimizations, and use fixed benchmark values ​​for assessment and optimization, resulting in low accuracy of small-index energy efficiency assessments and optimizations. Therefore, the multi-dimensional energy efficiency analysis method for small-index coal quality still needs further improvement. Summary of the Invention

[0004] This application aims to solve at least one of the technical problems existing in the prior art; to this end, this application proposes a multi-dimensional energy efficiency analysis method for small indicators based on online detection of coal quality entering the furnace, which is used to solve the technical problem that the prior art often ignores the real-time fluctuation of coal quality entering the furnace and uses fixed benchmark values ​​for evaluation and optimization when conducting energy efficiency assessment and optimization, resulting in low accuracy of small indicator energy efficiency assessment and optimization.

[0005] To achieve the above objectives, the first aspect of this application provides a multi-dimensional energy efficiency analysis method based on online detection of coal quality entering the furnace, including: Obtain data on the quality of coal entering the furnace and key operational indicators; Dynamic benchmark values ​​corresponding to operational indicators are generated based on the coal quality data entering the furnace. Multidimensional energy efficiency indicators are generated based on dynamic benchmark values ​​and operational small-scale indicator data. A multidimensional energy efficiency correlation map is constructed based on multidimensional energy efficiency indicators, and the optimal analysis results are obtained through dynamic optimization.

[0006] This application, through the above steps, corrects the evaluation benchmark from a fixed value to a dynamic value that is adjusted in real time according to coal quality, fundamentally eliminating the interference of coal quality fluctuations on the evaluation. At the same time, it uses a data model to reveal the complex mapping relationship between operating parameters and multi-dimensional energy efficiency, and calculates the optimal operating strategy adapted to the current coal quality in real time. This changes the evaluation distortion problem caused by inaccurate benchmarks in traditional methods, significantly improving the accuracy of small-index energy efficiency evaluation, as well as the timeliness and accuracy of optimization decisions.

[0007] Furthermore, the generation of dynamic benchmark values ​​corresponding to operational sub-indicators based on the coal quality data entering the furnace includes: Acquire equipment status characteristics; the equipment status characteristics include unit load, coal mill operating combination mode, and heat exchange surface cleanliness coefficient; Extract coal quality data for furnace feed; the coal quality data for furnace feed includes net calorific value, moisture content, volatile matter on dry ash-free basis, sulfur content and ash content on ash basis; Integrate equipment status characteristics and coal quality data into continuous small index analysis data; Input the continuous small indicator analysis data into the continuous small indicator benchmark prediction model to obtain the dynamic benchmark values ​​corresponding to several continuous small indicators. The continuous small indicators include flue gas temperature, oxygen content, and fly ash carbon content; The dynamic benchmark value corresponding to the theoretical minor index is calculated using an anti-balance calculation formula based on the principle of energy conservation; the theoretical minor index refers to coal consumption for power supply. The dynamic baseline values ​​for the control-related minor indicators are determined using a deterministic computational model based on chemical principles; the control-related minor indicators refer to sulfur dioxide concentration. Operational indicators are determined based on continuous small indicators, theoretical small indicators, and control small indicators.

[0008] Furthermore, the continuous small index benchmark prediction model is a multi-task branch prediction model, which consists of an input and feature preprocessing layer, a shared feature extraction layer, and a multi-task prediction branch. The input and feature preprocessing layer is used to receive continuous small index analysis data and perform feature preprocessing operations on the continuous small index analysis data to obtain a preprocessed feature vector; the feature preprocessing operations include feature standardization of continuous numerical features, encoding of categorical variables, and feature concatenation of all features to obtain a preprocessed feature vector; The shared feature extraction layer consists of several fully connected layers, each of which is followed by a ReLU activation function and a Dropout layer; its input data is a preprocessed feature vector, and its output data is a shared feature representation vector. The multi-task prediction branch consists of several parallel task branches; the task branches include a flue gas temperature branch, an oxygen branch, and a fly ash carbon content branch; the task branches employ several fully connected layers; the input data of the multi-task prediction branch is a shared feature representation vector, and the output data is the dynamic baseline values ​​corresponding to flue gas temperature, oxygen content, and fly ash carbon content.

[0009] Furthermore, the loss function of the continuous small index benchmark prediction model satisfies: Where LOSS represents the total loss; This is represented as multi-task loss. Represented as physical constraint loss; Represented as hyperparameters; The multi-task loss is the uncertainty-weighted sum of the prediction losses corresponding to several task branches; the prediction loss is calculated using the mean squared error loss. The multi-task loss satisfies: Where i represents the number corresponding to the task branch. Let V be the variance corresponding to the i-th task branch. Let be the prediction loss corresponding to the i-th task branch; The physical constraint loss satisfies: ; where max represents the maximum value operation. This is expressed as the current boiler efficiency. and These represent the theoretical minimum and theoretical maximum values ​​of boiler efficiency, respectively.

[0010] This application first integrates real-time data on coal quality and equipment status, and employs a prediction model consisting of a shared feature layer and multi-task branches to synchronously output dynamic benchmark values ​​for continuous small indicators. Simultaneously, it generates theoretical benchmark values ​​for power supply coal consumption and pollutant concentration by combining the inverse equilibrium formula and the stoichiometric model. The prediction model is trained and optimized using a joint loss function that integrates adaptive multi-task weighting and thermodynamic physical constraints. Through a multi-task sharing mechanism and the embedding of physical laws, the synergy and engineering rationality among different benchmark value predictions are significantly improved. This ensures that the generated dynamic benchmark values ​​not only accurately respond to coal quality fluctuations but also conform to the principles of energy conservation and combustion chemistry, thus laying a reliable foundation for subsequent accurate evaluation and optimization.

[0011] Furthermore, the generation of multi-dimensional energy efficiency indicators based on dynamic benchmark values ​​and operational small-scale indicator data includes: Obtain operational small indicator data and dynamic benchmark values ​​within the time window; the operational small indicator data includes continuous small indicators, theoretical small indicators and control small indicators, as well as the actual values ​​corresponding to the continuous small indicators, theoretical small indicators and control small indicators; Calculate the relative deviation rate of each sub-indicator in the continuous sub-indicators; the relative deviation rate is expressed as the ratio between the sub-indicator difference and its corresponding dynamic benchmark value; the sub-indicator difference is expressed as the difference between the actual value and its corresponding dynamic benchmark value; The operational energy efficiency score is determined based on the relative deviation rate corresponding to several consecutive small indicators. The additional coal consumption cost from an economic perspective is calculated using a formula. The formula satisfies: ;in, This is expressed as the actual value of coal consumption for power generation; This represents the dynamic baseline value of coal consumption for power supply; Load represents the equipment load; T represents the size of the time window. This is expressed as the unit price of coal fed into the furnace; Calculate the additional environmental costs for environmental protection dimensions using a formula. The formula satisfies: ;in, This is expressed as additional emissions; Expressed as a pollution emission rate; the additional emissions satisfy: ;in, This is expressed as a dynamic baseline value for sulfur dioxide concentration; Expressed as the actual value of sulfur dioxide concentration; The value is expressed as flue gas flow rate; T represents the size of the time window. Multidimensional energy efficiency indicators are determined based on operational energy efficiency scores, additional coal consumption costs in the economic dimension, and additional environmental protection costs in the environmental dimension.

[0012] Furthermore, the determination of operational energy efficiency score based on the relative deviation rate corresponding to several consecutive small indicators includes: Extract the relative deviation rate corresponding to several consecutive small indicators; Calculate the operating energy efficiency score using a formula. The formula satisfies: Where j represents the number of the sub-indicator in the continuous sub-indicators; n represents the total number of continuous sub-indicators; This is expressed as the relative deviation rate corresponding to the j-th sub-index. Let K represent the weight coefficient corresponding to the j-th sub-indicator; K represents the scaling factor, where K>0.

[0013] This application calculates the deviation rate of continuous small-scale operating parameters relative to their dynamic benchmarks and weights them to form an operational energy efficiency score. Simultaneously, it directly converts the difference between the actual value of coal consumption for power generation and the dynamic benchmark, combined with real-time coal prices and power generation, into additional fuel costs per unit time. Furthermore, it quantifies the deviation of pollutant emission concentrations from the dynamic benchmark, combined with flue gas flow and pollution discharge rates, into additional environmental costs. This transforms the traditional, fragmented, and abstract assessment of technical parameters into a unified and intuitive economic cost and quantitative score, enabling operators and managers to clearly identify the severity, economic impact, and environmental costs of energy efficiency issues. This provides a precise and operable quantitative basis for pursuing optimal operational decisions that balance safety, economy, and environmental protection.

[0014] Furthermore, the step of constructing a multidimensional energy efficiency correlation map based on multidimensional energy efficiency indicators and performing dynamic optimization to obtain the optimal analysis result includes: Obtain multi-dimensional energy efficiency indicators and adjustable operating parameters; Construct a multidimensional energy efficiency correlation map; the multidimensional energy efficiency correlation map includes a key adjustable parameter table and a SHAP dependency graph; Extract at least U key adjustable parameters that coexist in the key adjustable parameter table of the multidimensional energy efficiency index, and use them as adjustment and optimization parameters, where U>0; The values ​​of the adjustment and optimization parameters are determined based on the adjustment and optimization parameters and their corresponding SHAP dependency graph; The optimal analysis result is determined based on the adjusted optimization parameters and their corresponding values.

[0015] Furthermore, the construction of the multidimensional energy efficiency correlation map includes: Obtain several historical multidimensional energy efficiency indicators and historical operating parameters; A sample dataset was obtained by time-aligning several historical multidimensional energy efficiency indicators and historical operating parameters. A LightGBM regression model is trained for each of the multidimensional energy efficiency indicators: operational energy efficiency score, additional coal consumption cost in the economic dimension, and additional environmental protection cost in the environmental dimension. The input data of the LightGBM regression model is the operational parameters, and the output data is the predicted value of the corresponding energy efficiency indicator. The sample dataset is divided into training set, validation set and test set; the LightGBM regression model is trained on the training set, the hyperparameters are adjusted by Bayesian optimization on the validation set, and the model is tested on the test set. When the test accuracy is greater than the test threshold, the prediction model corresponding to the running energy efficiency score, the additional coal consumption cost in the economic dimension and the additional environmental protection cost in the environmental dimension is obtained. In the test set, the SHAP contribution value of each adjustable parameter in each sample to the output of the prediction model is calculated using the TreeSHAP algorithm. A table of key adjustable parameters is determined based on the SHAP contribution value; the table of key adjustable parameters includes several key adjustable parameters. A SHAP dependency graph is plotted for key adjustable parameters and their corresponding SHAP contribution values; the horizontal axis of the SHAP dependency graph represents the key adjustable parameter values, and the vertical axis represents the SHAP contribution values. A multidimensional energy efficiency correlation map was determined based on a table of key adjustable parameters and a SHAP dependency graph.

[0016] Furthermore, the determination of the key adjustable parameter table based on SHAP contribution values ​​includes: Extract the SHAP contribution values ​​for all samples; Statistical analysis was performed on the SHAP contribution values ​​of all samples, the mean absolute value of the SHAP value of each adjustable parameter was calculated, and the mean absolute values ​​of the adjustable parameters were sorted in descending order to obtain the adjustable parameter ranking table corresponding to each multidimensional energy efficiency index. Select the top M adjustable parameters from the adjustable parameter sorting table as the key adjustable parameters corresponding to the current multidimensional energy efficiency index, and form a key adjustable parameter table; where M>0.

[0017] Furthermore, determining the adjustment and optimization parameter values ​​based on the adjustment and optimization parameters and their corresponding SHAP dependency graph includes: Extract and adjust the optimization parameters and their corresponding SHAP dependency graphs; The overall benefit is obtained by weighted summation of the ordinate values ​​corresponding to each adjusted and optimized parameter value in several SHAP dependency graphs. The adjustment and optimization parameter value corresponding to the maximum comprehensive return is selected as the final adjustment and optimization parameter value.

[0018] This application first trains multiple LightGBM models using historical data to predict energy efficiency indicators in various dimensions. It then employs the TreeSHAP algorithm to quantify the impact of each adjustable operating parameter on the prediction results, generating a multi-dimensional energy efficiency correlation graph centered on a key parameter ranking table and a SHAP dependency graph. Furthermore, by identifying common key parameters important across multiple dimensions as optimization variables and weighting and fusing them based on their contribution values ​​in the SHAP dependency graphs of each dimension, it calculates specific parameter settings that maximize overall benefits. This transforms the complex multi-objective optimization problem into an interpretable deterministic search process based on a data-driven quantified graph. This significantly reduces the computational complexity of online optimization and its dependence on the accuracy of the entire model, and allows the optimization results to directly and transparently reflect the synergistic trade-offs between operational, economic, and environmental goals. This provides efficient, reliable, and easily understood and implemented precise operational guidance for on-site operations.

[0019] Compared with the prior art, the beneficial effects of this application are: 1. This application generates dynamic benchmark values ​​corresponding to operational sub-indicators based on coal quality data entering the furnace; generates multi-dimensional energy efficiency indicators based on the dynamic benchmark values ​​and operational sub-indicator data; constructs a multi-dimensional energy efficiency correlation graph based on the multi-dimensional energy efficiency indicators and performs dynamic optimization to obtain the optimal analysis results, thereby correcting the evaluation benchmark from a fixed value to a dynamic value that adjusts in real time with coal quality, fundamentally eliminating the interference of coal quality fluctuations on the evaluation; and uses a data model to reveal the complex mapping relationship between operating parameters and multi-dimensional energy efficiency, and calculates the optimal operating strategy adapted to the current coal quality in real time, changing the evaluation distortion problem caused by inaccurate benchmarks in traditional methods, and significantly improving the accuracy of sub-indicator energy efficiency evaluation and the timeliness and accuracy of optimization decisions.

[0020] 2. This application first integrates real-time data on coal quality and equipment status, and employs a prediction model consisting of a shared feature layer and multi-task branches to synchronously output dynamic benchmark values ​​for continuous small indicators. Simultaneously, it generates theoretical benchmark values ​​for power supply coal consumption and pollutant concentration by combining the inverse equilibrium formula and the stoichiometric model. The prediction model is trained and optimized using a joint loss function that integrates adaptive multi-task weighting and thermodynamic physical constraints. Through a multi-task sharing mechanism and the embedding of physical laws, the synergy and engineering rationality among different benchmark value predictions are significantly improved. This ensures that the generated dynamic benchmark values ​​not only accurately respond to coal quality fluctuations but also conform to the principles of energy conservation and combustion chemistry, thus laying a reliable foundation for subsequent accurate evaluation and optimization.

[0021] 3. This application first uses historical data to train multiple LightGBM models to predict energy efficiency indicators in various dimensions, and then uses the TreeSHAP algorithm to quantify the impact of each adjustable operating parameter on the prediction results, thereby generating a multi-dimensional energy efficiency correlation map with a key parameter ranking table and a SHAP dependency graph as the core. Then, by identifying common key parameters that are important in multiple dimensions as optimization variables, and performing weighted fusion based on their contribution values ​​on the SHAP dependency graphs of each dimension, the specific parameter settings that maximize the comprehensive benefits are calculated. This transforms the complex multi-objective optimization problem into an interpretable deterministic search process based on a data-driven quantified graph. This not only significantly reduces the computational complexity of online optimization and the dependence on the accuracy of the entire model, but also enables the optimization results to directly and transparently reflect the synergistic trade-off between operational, economic, and environmental goals, providing efficient, reliable, and easy-to-understand and implement precise operational guidance for the field. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart of the multi-dimensional energy efficiency analysis method for small indicators based on online detection of coal quality in the furnace, as described in this application. Figure 2 This is a flowchart illustrating the generation process of the optimal analysis results for this application. Detailed Implementation

[0024] The technical solutions of this application will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0025] Please see Figure 1 The first aspect of this application provides a method for multi-dimensional energy efficiency analysis of small indicators based on online detection of coal quality entering the furnace, including: Obtain data on the quality of coal entering the furnace and key operational indicators; Dynamic benchmark values ​​corresponding to operational indicators are generated based on the coal quality data entering the furnace. Multidimensional energy efficiency indicators are generated based on dynamic benchmark values ​​and operational small-scale indicator data. A multidimensional energy efficiency correlation map is constructed based on multidimensional energy efficiency indicators, and the optimal analysis results are obtained through dynamic optimization.

[0026] In this embodiment, the dynamic benchmark value corresponding to the operational sub-indicators is generated based on the coal quality data entering the furnace, including: Acquire equipment status characteristics; these characteristics include unit load, coal mill operating combination mode, and heat exchange surface cleanliness coefficient. Extract coal quality data for furnace feed; coal quality data for furnace feed includes net calorific value, moisture content, volatile matter on dry ash-free basis, sulfur content and ash content on as-received basis; Integrate equipment status characteristics and coal quality data into continuous small index analysis data; Input the continuous small indicator analysis data into the continuous small indicator benchmark prediction model to obtain the dynamic benchmark values ​​corresponding to several continuous small indicators. Continuous minor indicators include flue gas temperature, oxygen content, and fly ash carbon content; The dynamic baseline value corresponding to the theoretical minor index is calculated using an anti-balance calculation formula based on the principle of energy conservation; the theoretical minor index refers to coal consumption for power supply; in this embodiment, the anti-balance calculation formula satisfies: ;in, This is represented as a dynamic baseline value for coal consumption for power generation. This is expressed as the current boiler efficiency. This is expressed as the current turbine efficiency. This is expressed as the current generator efficiency; It is expressed as the received lower heating value; the 3600 in the molecule represents the calorific value corresponding to 1 kilowatt-hour of electrical energy. The dynamic baseline values ​​for the control-related minor indicators are determined using a deterministic calculation model based on chemical principles; the control-related minor indicators refer to sulfur dioxide concentration; in this embodiment, the expression of the deterministic calculation model satisfies: ;in, This is expressed as a dynamic baseline value for sulfur dioxide concentration. This indicates that the sulfur content was received. Represented as ash content; Operational indicators are determined based on continuous small indicators, theoretical small indicators, and control small indicators.

[0027] The continuous small index benchmark prediction model in this embodiment is a multi-task branch prediction model, which consists of an input and feature preprocessing layer, a shared feature extraction layer, and a multi-task prediction branch. The input and feature preprocessing layer is used to receive continuous small index analysis data and perform feature preprocessing operations on the continuous small index analysis data to obtain a preprocessed feature vector. The feature preprocessing operations include feature standardization of continuous numerical features, encoding of categorical variables, and feature concatenation of all features to obtain a preprocessed feature vector. The shared feature extraction layer consists of several fully connected layers, each followed by a ReLU activation function and a Dropout layer; its input data is a preprocessed feature vector, and its output data is a shared feature representation vector; in this embodiment, the shared feature representation vector encodes deep information common to all tasks and related to coal quality and equipment operating status; The multi-task prediction branch consists of several parallel task branches; the task branches include flue gas temperature branch, oxygen content branch and fly ash carbon content branch; the task branches adopt several fully connected layers; the input data of the multi-task prediction branch is a shared feature representation vector, and the output data is the dynamic baseline value corresponding to flue gas temperature, oxygen content and fly ash carbon content. In this embodiment, the number of fully connected layers in the shared feature extraction layer is set to 4, and the number of fully connected layers in each task branch is set to 2.

[0028] The loss function of the continuous small index benchmark prediction model in this embodiment satisfies: Where LOSS represents the total loss; This is represented as multi-task loss. Represented as physical constraint loss; This is represented as a hyperparameter; in this embodiment, the total loss is used for backpropagation to optimize the parameters of the continuous small index benchmark prediction model. The multi-task loss is the uncertainty-weighted sum of the prediction losses corresponding to several task branches; the prediction loss is calculated using the mean squared error loss. Multi-task loss satisfies: Where i represents the number corresponding to the task branch. Let V be the variance corresponding to the i-th task branch. Let be the prediction loss corresponding to the i-th task branch; Physical constraint loss satisfies: ; where max represents the maximum value operation. This is expressed as the current boiler efficiency. and These represent the theoretical minimum and theoretical maximum values ​​of boiler efficiency, respectively.

[0029] This embodiment first integrates real-time data on coal quality and equipment status, and uses a prediction model composed of a shared feature layer and multi-task branches to synchronously output dynamic benchmark values ​​for continuous small indicators. At the same time, it combines the inverse equilibrium formula and the stoichiometric model to generate theoretical benchmark values ​​for power supply coal consumption and pollutant concentration, respectively. The prediction model is trained and optimized by fusing a joint loss function of adaptive multi-task weighting and thermodynamic physical constraints. With the help of the multi-task sharing mechanism and the embedding of physical laws, the synergy and engineering rationality among different benchmark predictions are significantly improved. The resulting dynamic benchmark values ​​can not only accurately respond to coal quality fluctuations, but also ensure that they conform to the principles of energy conservation and combustion chemistry, thus laying a reliable foundation for subsequent accurate evaluation and optimization.

[0030] In this embodiment, the generation of multi-dimensional energy efficiency indicators based on dynamic benchmark values ​​and operational small-scale indicator data includes: Obtain the running sub-indicator data and dynamic benchmark values ​​within the time window; the running sub-indicator data includes continuous sub-indicators, theoretical sub-indicators, and control sub-indicators, as well as the actual values ​​corresponding to the continuous sub-indicators, theoretical sub-indicators, and control sub-indicators; the size of the time window is set based on experience, and in this embodiment it is set to 6 hours; Calculate the relative deviation rate of each sub-indicator in the continuous sub-indicators; the relative deviation rate is expressed as the ratio between the sub-indicator difference and its corresponding dynamic benchmark value; the sub-indicator difference is expressed as the difference between the actual value and its corresponding dynamic benchmark value; The operational energy efficiency score is determined based on the relative deviation rate corresponding to several consecutive small indicators. The additional coal consumption cost from an economic perspective is calculated using a formula. The formula satisfies: ;in, This is expressed as the actual value of coal consumption for power generation; This represents the dynamic baseline value of coal consumption for power supply; Load represents the equipment load; T represents the size of the time window. This is expressed as the unit price of coal fed into the furnace; Calculate the additional environmental costs for environmental protection dimensions using a formula. The formula satisfies: ;in, This is expressed as additional emissions; Expressed as a pollution emission rate; additional emissions must meet the following requirements: ;in, This is expressed as a dynamic baseline value for sulfur dioxide concentration; Expressed as the actual value of sulfur dioxide concentration; The value is expressed as flue gas flow rate; T represents the size of the time window; in this embodiment, the values ​​mentioned above are... In the economic dimension, the formula for calculating additional coal consumption costs is to convert units of grams to tons, while in the environmental dimension, the formula for calculating additional environmental costs is to convert units of milligrams to kilograms. Multidimensional energy efficiency indicators are determined based on operational energy efficiency scores, additional coal consumption costs in the economic dimension, and additional environmental protection costs in the environmental dimension.

[0031] In this embodiment, the determination of operational energy efficiency score based on the relative deviation rate corresponding to several consecutive small indicators includes: Extract the relative deviation rate corresponding to several consecutive small indicators; Calculate the operating energy efficiency score using a formula. The formula satisfies: Where j represents the number corresponding to the sub-indicator in the continuous sub-indicators; n represents the total number of continuous sub-indicators, and in this embodiment, n is 3; This is expressed as the relative deviation rate corresponding to the j-th sub-index. The value is represented by the weight coefficient corresponding to the j-th sub-index. The specific value is set according to experience. In this embodiment, the weight coefficients corresponding to flue gas temperature, oxygen content and fly ash carbon content are set to 0.45, 0.35 and 0.2 respectively; K represents the scaling factor, K>0; the specific value is set according to experience. In this embodiment, K is set to 500. In this embodiment, the sum of the absolute values ​​of the deviation rates corresponding to several sub-indexes is less than or equal to 0.2.

[0032] Please see Figure 2 In this embodiment, the construction of a multidimensional energy efficiency correlation map based on multidimensional energy efficiency indicators and the dynamic optimization to obtain the optimal analysis result include: Obtain multi-dimensional energy efficiency indicators and adjustable operating parameters; in this embodiment, the adjustable operating parameters include total air volume, primary air pressure / air volume, secondary air ratio, coal feed distribution, coal mill outlet temperature, oxygen setpoint, burner swing angle, main steam temperature setting, reheat steam temperature adjustment method, number of circulating water pumps in operation, and fan frequency conversion command, etc. Construct a multidimensional energy efficiency correlation map; the multidimensional energy efficiency correlation map includes a key adjustable parameter table and a SHAP dependency graph; Extract at least U key adjustable parameters that coexist in the key adjustable parameter table of the multidimensional energy efficiency index, and use them as adjustment and optimization parameters, where U>0; the specific value is set according to experience, and in this embodiment it is set to 2; in another embodiment it can be set to 3; that is, the key adjustable parameter exists in at least U key adjustable parameter tables; The values ​​of the adjustment and optimization parameters are determined based on the adjustment and optimization parameters and their corresponding SHAP dependency graph; The optimal analysis result is determined based on the adjusted optimization parameters and their corresponding values.

[0033] The construction of the multidimensional energy efficiency correlation map in this embodiment includes: Obtain several historical multidimensional energy efficiency indicators and historical operating parameters; A sample dataset was obtained by time-aligning several historical multidimensional energy efficiency indicators and historical operating parameters. Train a LightGBM regression model for each of the three energy efficiency indicators: operational energy efficiency score, additional coal consumption cost in the economic dimension, and additional environmental protection cost in the environmental dimension. The input data of the LightGBM regression model is the operational parameters, and the output data is the predicted value of the corresponding energy efficiency indicator. The sample dataset is divided into training, validation, and test sets. A LightGBM regression model is trained on the training set, hyperparameters are adjusted using Bayesian optimization on the validation set, and the model is tested on the test set. When the test accuracy exceeds a test threshold, the predicted models for energy efficiency scores, additional coal consumption costs in the economic dimension, and additional environmental costs in the environmental dimension are obtained. The test threshold is set empirically; in this embodiment, it is set to 85%. In this embodiment, the ratio between the training, validation, and test sets is set to 70%:15%:15%. In the test set, the SHAP contribution value of each adjustable parameter in each sample to the output of the prediction model is calculated using the TreeSHAP algorithm. The key adjustable parameter table is determined based on the SHAP contribution value; the key adjustable parameter table includes several key adjustable parameters; A SHAP dependency graph is plotted for the key adjustable parameters and their corresponding SHAP contribution values; the horizontal axis of the SHAP dependency graph represents the key adjustable parameter values, and the vertical axis represents the SHAP contribution values; in this embodiment, the key adjustable parameter values ​​are represented by the numerical values ​​of the key adjustable parameters. A multidimensional energy efficiency correlation map was determined based on a table of key adjustable parameters and a SHAP dependency graph.

[0034] The key adjustable parameter table determined based on SHAP contribution values ​​in this embodiment includes: Extract the SHAP contribution values ​​for all samples; Statistical analysis was performed on the SHAP contribution values ​​of all samples, the mean absolute value of the SHAP value of each adjustable parameter was calculated, and the mean absolute values ​​of the adjustable parameters were sorted in descending order to obtain the adjustable parameter ranking table corresponding to each multidimensional energy efficiency index. Select the first M adjustable parameters in the adjustable parameter sorting table as the key adjustable parameters corresponding to the current multidimensional energy efficiency index, and form a key adjustable parameter table; where M>0, the specific value is set according to experience, and in this embodiment it is set to 4.

[0035] In this embodiment, determining the adjustment and optimization parameter values ​​based on the adjustment and optimization parameters and their corresponding SHAP dependency graph includes: Extract and adjust the optimization parameters and their corresponding SHAP dependency graphs; The comprehensive return is obtained by weighted summation of the ordinate values ​​corresponding to each adjusted optimization parameter value in several SHAP dependency graphs; in this embodiment, the formula for calculating the comprehensive return ZSY satisfies: Where A, B, and C represent the weighting coefficients corresponding to the operating energy efficiency score, the additional coal consumption cost in the economic dimension, and the additional environmental protection cost in the environmental protection dimension, respectively. The specific values ​​are set according to experience, and in this embodiment, they are set to 0.4, 0.4, and 0.2, respectively. , and These represent the vertical coordinate values ​​in the SHAP dependency graph corresponding to the operating energy efficiency score, the additional coal consumption cost in the economic dimension, and the additional environmental protection cost in the environmental dimension, respectively. In this embodiment, the adjustable operating parameters are optimized to reduce costs. Therefore, the additional coal consumption cost in the economic dimension and the additional environmental protection cost in the environmental dimension are calculated as negative values ​​when calculating the comprehensive benefits. The adjustment and optimization parameter value corresponding to the maximum comprehensive return is selected as the final adjustment and optimization parameter value; in this embodiment, the adjustment and optimization parameter value is represented by the numerical value of the adjustment and optimization parameter.

[0036] This embodiment first trains multiple LightGBM models using historical data to predict energy efficiency indicators in various dimensions. It then employs the TreeSHAP algorithm to quantify the impact of each adjustable operating parameter on the prediction results, thereby constructing a multi-dimensional energy efficiency correlation graph centered on a key parameter ranking table and a SHAP dependency graph. Next, by identifying common key parameters important in multiple dimensions as optimization variables, and weighting and fusing them based on their contribution values ​​in the SHAP dependency graphs of each dimension, it calculates the specific parameter settings that maximize overall benefits. This transforms the complex multi-objective optimization problem into an interpretable deterministic search process based on a data-driven quantified graph. This not only significantly reduces the computational complexity of online optimization and its dependence on the accuracy of the entire model, but also allows the optimization results to directly and transparently reflect the synergistic trade-offs between operational, economic, and environmental goals, providing efficient, reliable, and easily understood and executed precise operational guidance for on-site operations.

[0037] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0038] The working principle of this application is as follows: It acquires data on the quality of coal entering the furnace and data on operational indicators; it generates dynamic benchmark values ​​corresponding to the operational indicators based on the coal quality data; it generates multi-dimensional energy efficiency indicators based on the dynamic benchmark values ​​and operational indicator data; it constructs a multi-dimensional energy efficiency correlation graph based on the multi-dimensional energy efficiency indicators and performs dynamic optimization to obtain the optimal analysis results, correcting the evaluation benchmark from a fixed value to a dynamic value that adjusts in real time with coal quality, fundamentally eliminating the interference of coal quality fluctuations on the evaluation; it uses a data model to reveal the complex mapping relationship between operating parameters and multi-dimensional energy efficiency, and calculates the optimal operating strategy adapted to the current coal quality in real time, changing the evaluation distortion problem caused by inaccurate benchmarks in traditional methods, significantly improving the accuracy of small-indicator energy efficiency evaluation and the timeliness and accuracy of optimization decisions, avoiding the problem that existing technologies often ignore real-time fluctuations in coal quality entering the furnace and use fixed benchmark values ​​for evaluation and optimization, resulting in low accuracy of small-indicator energy efficiency evaluation and optimization.

[0039] The above embodiments are only used to illustrate the technical methods of this application and are not intended to limit it. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of this application without departing from the spirit and scope of the technical methods of this application.

Claims

1. A multi-dimensional energy efficiency analysis method based on online detection of coal quality entering the furnace, characterized in that, include: Obtain data on the quality of coal entering the furnace and key operational indicators; Dynamic benchmark values ​​corresponding to operational indicators are generated based on the coal quality data entering the furnace. Multidimensional energy efficiency indicators are generated based on dynamic benchmark values ​​and operational small-scale indicator data. A multidimensional energy efficiency correlation map is constructed based on multidimensional energy efficiency indicators, and the optimal analysis results are obtained through dynamic optimization.

2. The method for multidimensional energy efficiency analysis of small indicators based on online detection of coal quality entering the furnace, as described in claim 1, is characterized in that... The generation of dynamic benchmark values ​​corresponding to operational sub-indicators based on coal quality data includes: Obtain device status characteristics; Extract coal quality data for furnace feed; the coal quality data for furnace feed includes net calorific value, moisture content, volatile matter on dry ash-free basis, sulfur content and ash content on ash basis; Integrate equipment status characteristics and coal quality data into continuous small index analysis data; Input the continuous small indicator analysis data into the continuous small indicator benchmark prediction model to obtain the dynamic benchmark values ​​corresponding to several continuous small indicators. The continuous small indicators include flue gas temperature, oxygen content, and fly ash carbon content; The dynamic benchmark value corresponding to the theoretical minor index is calculated using an anti-balance calculation formula based on the principle of energy conservation; the theoretical minor index refers to coal consumption for power supply. The dynamic baseline values ​​for the control-related minor indicators are determined using a deterministic computational model based on chemical principles; the control-related minor indicators refer to sulfur dioxide concentration. Operational indicators are determined based on continuous small indicators, theoretical small indicators, and control small indicators.

3. The method for multidimensional energy efficiency analysis of small indicators based on online detection of coal quality entering the furnace, as described in claim 2, is characterized in that... The continuous small index benchmark prediction model is a multi-task branch prediction model, which consists of an input and feature preprocessing layer, a shared feature extraction layer, and a multi-task prediction branch. The input and feature preprocessing layer is used to receive continuous small index analysis data and perform feature preprocessing operations on the continuous small index analysis data to obtain preprocessed feature vectors. The feature preprocessing operations include feature standardization of continuous numerical features, encoding of categorical variables, and feature concatenation of all features to obtain a preprocessed feature vector; The shared feature extraction layer consists of several fully connected layers, each of which is followed by a ReLU activation function and a Dropout layer; its input data is a preprocessed feature vector, and its output data is a shared feature representation vector. The multi-task prediction branch consists of several parallel task branches; the task branches include a flue gas temperature branch, an oxygen branch, and a fly ash carbon content branch; the task branches employ several fully connected layers; the input data of the multi-task prediction branch is a shared feature representation vector, and the output data is the dynamic baseline values ​​corresponding to flue gas temperature, oxygen content, and fly ash carbon content.

4. The method for multidimensional energy efficiency analysis of small indicators based on online detection of coal quality entering the furnace, as described in claim 2, is characterized in that... The loss function of the continuous small index benchmark prediction model satisfies: Where LOSS represents the total loss; This is represented as multi-task loss. Represented as physical constraint loss; Represented as hyperparameters; The multi-task loss is the uncertainty-weighted sum of the prediction losses corresponding to several task branches; the prediction loss is calculated using the mean squared error loss. The multi-task loss satisfies: Where i represents the number corresponding to the task branch. Let the variance be the variance corresponding to the i-th task branch. Let be the prediction loss corresponding to the i-th task branch; The physical constraint loss satisfies: ; where max represents the maximum value operation. This is expressed as the current boiler efficiency. and These represent the theoretical minimum and theoretical maximum values ​​of boiler efficiency, respectively.

5. The method for multidimensional energy efficiency analysis of small indicators based on online detection of coal quality entering the furnace according to claim 1, characterized in that, The generation of multi-dimensional energy efficiency indicators based on dynamic benchmark values ​​and operational small-scale indicator data includes: Obtain operational small indicator data and dynamic benchmark values ​​within the time window; the operational small indicator data includes continuous small indicators, theoretical small indicators and control small indicators, as well as the actual values ​​corresponding to the continuous small indicators, theoretical small indicators and control small indicators; Calculate the relative deviation rate of each sub-indicator in the continuous sub-indicators; the relative deviation rate is expressed as the ratio between the sub-indicator difference and its corresponding dynamic benchmark value; the sub-indicator difference is expressed as the difference between the actual value and its corresponding dynamic benchmark value; The operational energy efficiency score is determined based on the relative deviation rate corresponding to several consecutive small indicators. The additional coal consumption cost from an economic perspective is calculated using a formula. The formula satisfies: ;in, This is expressed as the actual value of coal consumption for power generation; This represents the dynamic baseline value of coal consumption for power supply; Load represents the equipment load; T represents the size of the time window. This is expressed as the unit price of coal fed into the furnace; Calculate the additional environmental costs for environmental protection dimensions using a formula. The formula satisfies: ;in, This is expressed as additional emissions; Expressed as a pollution emission rate; the additional emissions satisfy: ;in, This is expressed as a dynamic baseline value for sulfur dioxide concentration; Expressed as the actual value of sulfur dioxide concentration; The value is expressed as flue gas flow rate; T represents the size of the time window. Multidimensional energy efficiency indicators are determined based on operational energy efficiency scores, additional coal consumption costs in the economic dimension, and additional environmental protection costs in the environmental dimension.

6. The method for multidimensional energy efficiency analysis of small indicators based on online detection of coal quality entering the furnace, as described in claim 5, is characterized in that... The determination of operational energy efficiency scores based on the relative deviation rates corresponding to several consecutive small indicators includes: Extract the relative deviation rate corresponding to several consecutive small indicators; Calculate the operating energy efficiency score using a formula. The formula satisfies: Where j represents the number of the sub-indicator in the continuous sub-indicators; n represents the total number of continuous sub-indicators; This is expressed as the relative deviation rate corresponding to the j-th sub-index. Let K represent the weight coefficient corresponding to the j-th sub-indicator; K represents the scaling factor, where K>

0.

7. The method for multidimensional energy efficiency analysis of small indicators based on online detection of coal quality entering the furnace according to claim 1, characterized in that, The process of constructing a multidimensional energy efficiency correlation map based on multidimensional energy efficiency indicators and dynamically optimizing it to obtain the optimal analysis results includes: Obtain multi-dimensional energy efficiency indicators and adjustable operating parameters; Construct a multidimensional energy efficiency correlation map; the multidimensional energy efficiency correlation map includes a key adjustable parameter table and a SHAP dependency graph; Extract at least U key adjustable parameters that coexist in the key adjustable parameter table of the multidimensional energy efficiency index, and use them as adjustment and optimization parameters, where U>0; The values ​​of the adjustment and optimization parameters are determined based on the adjustment and optimization parameters and their corresponding SHAP dependency graph; The optimal analysis result is determined based on the adjusted optimization parameters and their corresponding values.

8. The method for multidimensional energy efficiency analysis of small indicators based on online detection of coal quality entering the furnace according to claim 7, characterized in that, The construction of the multidimensional energy efficiency correlation map includes: Obtain several historical multidimensional energy efficiency indicators and historical operating parameters; A sample dataset was obtained by time-aligning several historical multidimensional energy efficiency indicators and historical operating parameters. A LightGBM regression model is trained for each of the multidimensional energy efficiency indicators: operational energy efficiency score, additional coal consumption cost in the economic dimension, and additional environmental protection cost in the environmental dimension. The input data of the LightGBM regression model is the operational parameters, and the output data is the predicted value of the corresponding energy efficiency indicator. The sample dataset is divided into training set, validation set and test set; the LightGBM regression model is trained on the training set, the hyperparameters are adjusted by Bayesian optimization on the validation set, and the model is tested on the test set. When the test accuracy is greater than the test threshold, the prediction model corresponding to the running energy efficiency score, the additional coal consumption cost in the economic dimension and the additional environmental protection cost in the environmental dimension is obtained. In the test set, the SHAP contribution value of each adjustable parameter in each sample to the output of the prediction model is calculated using the TreeSHAP algorithm. A table of key adjustable parameters is determined based on the SHAP contribution value; the table of key adjustable parameters includes several key adjustable parameters. A SHAP dependency graph is plotted for key adjustable parameters and their corresponding SHAP contribution values; the horizontal axis of the SHAP dependency graph represents the key adjustable parameter values, and the vertical axis represents the SHAP contribution values. A multidimensional energy efficiency correlation map was determined based on a table of key adjustable parameters and a SHAP dependency graph.

9. The method for multidimensional energy efficiency analysis of small indicators based on online detection of coal quality entering the furnace, as described in claim 8, is characterized in that... The table for determining key adjustable parameters based on SHAP contribution values ​​includes: Extract the SHAP contribution values ​​for all samples; Statistical analysis was performed on the SHAP contribution values ​​of all samples, the mean absolute value of the SHAP value of each adjustable parameter was calculated, and the mean absolute values ​​of the adjustable parameters were sorted in descending order to obtain the adjustable parameter ranking table corresponding to each multidimensional energy efficiency index. Select the top M adjustable parameters from the adjustable parameter sorting table as the key adjustable parameters corresponding to the current multidimensional energy efficiency index, and form a key adjustable parameter table; where M>0.

10. The method for multidimensional energy efficiency analysis of small indicators based on online detection of coal quality entering the furnace according to claim 7, characterized in that, The process of determining the adjustment and optimization parameter values ​​based on the adjustment and optimization parameters and their corresponding SHAP dependency graph includes: Extract and adjust the optimization parameters and their corresponding SHAP dependency graphs; The overall benefit is obtained by weighted summation of the ordinate values ​​corresponding to each adjusted and optimized parameter value in several SHAP dependency graphs. The adjustment and optimization parameter value corresponding to the maximum comprehensive return is selected as the final adjustment and optimization parameter value.

Images