Improved gradient boosting tree-based method and system for calculating startup cost of thermal power generating unit

Abstract

The application provides a coal-fired, gas-fired and biomass-fired power unit start-up cost calculation method and system based on an improved gradient boosting tree, and relates to the technical field of cost calculation. Based on the differences among the coal-fired, gas-fired and biomass-fired power units, corresponding unit characteristic models are established; according to the unit characteristic models, the start-up cost is divided into fuel consumption cost, equipment wear cost, auxiliary service cost and environmental protection treatment cost, and a start-up cost multi-dimensional subdivision model is constructed; the length of time before start-up and the start-up load climbing rate are quantified as dynamic influence factors, and are associated with each cost item; the improved gradient boosting tree algorithm is adopted to fit the nonlinear relationship between each cost item and the dynamic influence factor, and the total start-up cost and the calculation results of each subdivision cost item are obtained based on the multi-dimensional subdivision model, so that the accurate calculation of the start-up cost of different types of thermal power units under multiple working conditions is realized.

Description

Technical Field

[0001] This invention relates to the field of cost calculation technology, and in particular to a method and system for calculating the start-up cost of thermal power units based on an improved gradient boosting tree. Background Technology

[0002] The startup cost of a thermal power unit refers to the total cost of all kinds of consumption and losses incurred during the entire process of gradually heating up, pressurizing, and connecting to the grid and bearing load from a static shutdown state. It is one of the core technical indicators for evaluating the economic efficiency of unit operation and controlling costs throughout the entire life cycle. In terms of composition, startup cost mainly covers multiple dimensions such as fuel consumption cost, equipment wear and aging loss cost, plant power consumption cost, environmental protection facility start-up and shutdown loss cost, and auxiliary system operating cost. Its formation mechanism is closely related to the unit type, equipment structure, shutdown status, startup method, and operating conditions. As a key component of the operating cost of a thermal power unit, startup cost not only directly affects the economic benefits of the unit itself, but also serves as an important basis for power dispatching agencies to formulate start-up and shutdown plans and for power market participants to make bidding decisions. The accuracy and rationality of its calculation results are directly related to the safe and stable operation of the unit and the overall economic operation level of the power system.

[0003] In recent years, the installed capacity and power supply share of new energy sources, represented by wind power and photovoltaics, have continued to increase in the power system. The power grid operation mode has gradually shifted from the traditional rigid dispatch mode dominated by thermal power to a flexible dispatch mode with a high proportion of new energy integration. Affected by the randomness, volatility, and intermittency of new energy output, the pressure on power grid peak shaving and frequency regulation has increased significantly. Thermal power units have gradually shifted from being the main body of traditional base load operation to undertaking ancillary service tasks such as deep peak shaving and rapid start-up and shutdown. The frequency of unit start-up and shutdown has increased significantly, and the fluctuation of operating conditions has intensified. Frequent start-up and shutdown has become the normalized operation mode of thermal power units under the new situation. Against this background, accurate, rapid, and comprehensive calculation of the start-up cost of thermal power units has increasingly important engineering value and practical significance for optimizing unit start-up and shutdown strategies, improving the precision of power system dispatch, improving the electricity market transaction pricing mechanism, and ensuring the economy and reliability of power supply. It also puts forward higher requirements for the adaptability, accuracy, and practicality of start-up cost calculation methods.

[0004] Current methods and technologies for calculating the startup cost of thermal power units still have many limitations and shortcomings, making it difficult to meet the calculation needs of complex operating conditions under the new circumstances. First, traditional models often use a uniform formula to calculate the startup cost of different types of thermal power units, failing to fully consider the essential differences between coal-fired, gas-fired, and biomass units in terms of fuel characteristics, startup process, and equipment losses, resulting in large calculation errors. Second, existing models have a coarse classification of startup cost components, often only including fuel costs and simple equipment losses, ignoring sub-cost items such as preheating losses of denitrification catalysts, additional consumption of biomass feedstock pretreatment, and ignition energy losses of gas-fired units during startup, resulting in incomplete cost coverage. At the same time, traditional methods do not consider the dynamic impact of unit downtime before startup and the startup load ramp-up rate on costs, and often use fixed coefficients for calculation, which cannot adapt to cost changes under different start-up and shutdown conditions. In addition, existing calculation models are mostly linear simplification models, which are difficult to handle the nonlinear relationship between fuel consumption and load changes during startup, as well as the comprehensive impact of multiple factors coupled on costs, resulting in insufficient calculation accuracy. Summary of the Invention

[0005] To address the aforementioned issues, this invention proposes a method and system for calculating the startup cost of thermal power units based on an improved gradient boosting tree. Based on the solution strategy of the improved gradient boosting tree, it enables accurate calculation of the startup cost of different types of thermal power units under multiple operating conditions.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a method for calculating the start-up cost of thermal power units based on an improved gradient boosting tree, comprising:

[0008] Based on the differences between coal-fired, gas-fired, and biomass power generation units, corresponding unit characteristic models are established for each type.

[0009] Based on the unit characteristic model, the start-up cost is divided into four categories: fuel consumption cost, equipment wear and tear cost, auxiliary service cost, and environmental treatment cost, and a multi-dimensional breakdown model of start-up cost is constructed.

[0010] The pre-startup downtime and startup load ramp-up rate are quantified as dynamic influencing factors and associated with each cost item.

[0011] An improved gradient boosting tree algorithm is used to fit the nonlinear relationship between each cost item and the dynamic influence factor. Based on the multi-dimensional subdivision model, the total start-up cost and the calculation results of each subdivision cost item are obtained.

[0012] Preferably, the unit characteristic model is used to quantify the total energy consumption of coal-fired, gas-fired, and biomass generator units under differences in fuel characteristics, equipment configuration, and start-up procedures.

[0013] Preferably, the fuel consumption cost is specifically:

[0014] ;

[0015] in, Let k be the fuel consumption cost of the k-th type of unit, where k = coal for coal, k = gas for gas, and k = bio for biomass. The total energy consumption for starting up the k-th type of unit; For the unit price of energy corresponding to the k-th type of unit, the unit price of coal is used for coal-fired power, the unit price of natural gas is used for gas-fired power, and the unit price of raw material purchase is used for biomass.

[0016] The specific equipment depreciation cost is as follows:

[0017] ;

[0018] ;

[0019] ;

[0020] in, Cost of heat exchange surface loss; The starting loss cost of rotating equipment; Let be the unit starting loss coefficient of the rotating equipment of the k-th type of unit. This represents the cumulative number of launches corresponding to this launch. Costs associated with starting losses of electrical equipment; Let be the loss coefficient of the electrical equipment of the k-th type of unit. This represents the average current during the startup phase. Startup time;

[0021] The specific cost of the ancillary services is as follows:

[0022] ;

[0023] in, Power consumption of auxiliary equipment during the startup phase; For grid connection electricity price; Expenditure to initiate external ancillary services;

[0024] The specific environmental treatment cost is as follows:

[0025] ;

[0026] in, Pollutants in the start-up phase For coal-fired power units, the additional emissions from the inefficient initial emissions during the startup phase of the desulfurization and denitrification systems must be considered. Biomass units need to calculate dust levels. With nitrogen oxides emission; pollutants The unit treatment cost is determined based on the type of environmental protection equipment in the unit.

[0027] Preferably, the dynamic influencing factor of pre-start-up downtime affects fuel consumption costs and equipment wear and tear costs; specifically:

[0028] ;

[0029] in, The outage duration influencing factor for type k generating units. ≥1; The total downtime of the unit before startup; This is the shutdown impact coefficient for unit k. Different units have different shutdown impact coefficients, reflecting the differences in insulation and cooling characteristics of different units. is the characteristic time constant of the k-th type of unit, reflecting the unit's temperature decay rate.

[0030] Preferably, the dynamic influencing factor of the start-up load ramp-up rate affects fuel consumption costs and ancillary service costs; specifically:

[0031] ;

[0032] in, For the load ramp-up rate influencing factor of the k-th type of unit, ≥1; This represents the actual start-up load ramp-up rate. Let K be the climb impact coefficient for the k-th type of unit; The optimal load ramp-up rate for the k-th type of unit is determined by the unit's design parameters and operating experience.

[0033] Preferably, the step of employing an improved gradient boosting tree algorithm to fit the nonlinear relationship between each cost item and the dynamic influencing factor, and obtaining the calculation results of the total start-up cost and each subdivided cost item based on the multi-dimensional subdivision model, specifically includes:

[0034] The input features are filtered by mutual information entropy to remove redundant features; the input features include unit type, pre-start downtime, start-up load ramp-up rate, energy price, pollutant treatment cost and equipment rated parameters.

[0035] The Huber loss function is used instead of the traditional mean squared error loss function;

[0036] Using dynamic influencing factors as feature inputs, a nonlinear mapping relationship between them and each cost item is established through gradient boosting tree training, and the prediction results of total startup cost and each sub-cost item are output.

[0037] Secondly, the present invention provides a system for calculating the start-up cost of thermal power units based on an improved gradient boosting tree, comprising:

[0038] The characteristic modeling module is used to establish corresponding unit characteristic models based on the differences between coal-fired, gas-fired, and biomass generator sets.

[0039] The cost breakdown module is used to divide the start-up cost into four categories of cost items: fuel consumption cost, equipment depreciation cost, auxiliary service cost, and environmental treatment cost, based on the unit characteristic model, and to construct a multi-dimensional breakdown model of start-up cost.

[0040] The dynamic factor introduction module is used to quantify the pre-startup downtime and startup load ramp-up rate as dynamic influencing factors and associate them with each cost item.

[0041] The cost calculation module is used to fit the nonlinear relationship between each cost item and the dynamic influence factor using an improved gradient boosting tree algorithm, and obtain the calculation results of the total start-up cost and each subdivided cost item based on the multi-dimensional subdivision model.

[0042] Preferably, the unit characteristic model is used to quantify the total energy consumption of coal-fired, gas-fired, and biomass generator units under differences in fuel characteristics, equipment configuration, and start-up procedures.

[0043] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the method for calculating the start-up cost of thermal power units based on an improved gradient boosting tree as described in the first aspect.

[0044] Fourthly, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the method for calculating the start-up cost of thermal power units based on an improved gradient boosting tree as described in the first aspect.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] (1) This invention constructs characteristic models of coal-fired, gas-fired, and biomass units in a differentiated manner to achieve adaptation analysis of different types of units; it subdivides start-up costs into four categories of cost items: fuel, equipment loss, auxiliary services, and environmental protection, and establishes a multi-dimensional cost architecture to avoid the one-sidedness of cost accounting; it introduces downtime and load ramp-up rate as dynamic influencing factors to solve the defect of traditional calculation ignoring dynamic changes in operating conditions; and it combines an improved gradient boosting tree algorithm to fit nonlinear relationships, which greatly improves the accuracy and efficiency of cost calculation, and provides scientific and reliable data support for the economic dispatch, cost control, and energy conservation and consumption reduction of thermal power units.

[0047] (2) This invention effectively solves the technical problems of traditional calculation methods that use fixed coefficients and have poor adaptability to operating conditions by quantifying two core dynamic influencing factors: the downtime before startup and the startup load ramp-up rate, and establishing a precise correlation correction mechanism between the factors and each subdivided cost item. Differentiated influence coefficients and characteristic parameters are set for the characteristic differences of coal-fired, gas-fired, and biomass units, so that the factor correction is more in line with the startup rules of different units. It can dynamically adapt to different startup scenarios after frequent startup and shutdown and long-term downtime of units, and can also respond to the cost changes of different load ramp-up strategies under the power dispatch demand. This allows the cost calculation results to accurately match the dynamic changes of actual operating conditions, fills the gap in the traditional model's consideration of the dynamic impact of operating conditions, and greatly improves the adaptability and calculation fit of the calculation model to complex and variable startup operating conditions.

[0048] (3) The cost calculation method based on the improved gradient boosting tree designed in this invention solves the problem of nonlinear correlation of multiple factors and coupling of dynamic factors in the calculation of startup costs, and significantly improves the calculation accuracy compared with the traditional linear model. By optimizing the feature through mutual information entropy, core input features are selected and redundant noise is eliminated, which effectively improves the model training efficiency and generalization ability. The Huber loss function is used to replace the traditional mean square error function, which balances the robustness of the model and the accuracy of the error, and significantly reduces the interference of extreme startup conditions on the calculation results. At the same time, the quantified dynamic factors are integrated into the model as features, realizing the nonlinear mapping between dynamic factors and each cost item, which can accurately fit the complex relationship of multi-dimensional cost items and coupling of multiple influencing factors, and provides an efficient and accurate solution for startup costs of different types of units under multiple operating conditions.

[0049] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0050] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute a limitation thereof.

[0051] Figure 1 The main flowchart of a method for calculating the start-up cost of thermal power units based on an improved gradient boosting tree, provided in an embodiment of the present invention, is shown below. Detailed Implementation

[0052] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0053] In power system operation, the calculation of startup costs for thermal power units is a crucial link concerning the fairness of the grid's economic dispatch and ancillary service market compensation mechanism. Existing technologies have made various attempts at cost calculation for thermal power units, but their core focus is often on the operating losses and energy consumption compensation when the unit participates in peak-shaving services, rather than on refined cost decoupling specifically for the unit startup process.

[0054] For example, some studies have divided the peak shaving stage into basic peak shaving, deep peak shaving, and deep oil injection peak shaving based on the output of thermal power units during the peak shaving process, and have calculated the coal consumption cost, start-up and shutdown cost, unit life loss cost, and oil injection cost respectively.

[0055] Although this method provides a relatively detailed breakdown of cost components in peak-shaving scenarios, its technical solution remains essentially limited to the compensation mechanism of peak-shaving services and fails to model the unique characteristics of the unit startup process. Specifically, the study does not consider the differences in combustion characteristics of units with different fuel types (coal, gas, biomass) during startup, nor does it introduce the quantitative impact of dynamic operating condition factors such as pre-startup shutdown duration and startup load ramp-up rate on startup costs. Its start-up and shutdown cost calculation relies solely on the binary switching of unit operating states (start-up or shutdown), lacking a deep decoupling of internal energy consumption and equipment wear during startup, making it difficult to accurately calculate startup costs under multiple operating conditions.

[0056] On the other hand, some studies use deep neural network models to predict energy data at the user end and use real-time detection parameters at the pipeline end to dynamically update the model. At the same time, the gradient boosting tree algorithm is used to optimize the compression factor compensation parameter in the preset algorithm.

[0057] However, this research is limited to natural gas energy metering, and its model update mechanism only addresses changes in natural gas composition and fluctuations in pipeline operating conditions. It fails to address the multi-dimensional cost components during the start-up of thermal power units (such as equipment wear and tear, environmental treatment, and ancillary services), and does not consider differentiated modeling of unit types or the embedding of dynamic factors related to start-up conditions. Therefore, despite its algorithmic advancements, it cannot be directly applied to the calculation of thermal power unit start-up costs to solve the problem of low accuracy caused by differences in unit types and poor adaptability to operating conditions.

[0058] In summary, conventional technologies for calculating the startup costs of thermal power units share the following common shortcomings: First, they lack differentiated modeling of the startup characteristics of coal-fired, gas-fired, and biomass units, resulting in the inability of the basic parameters for cost calculation to reflect the essential differences between different units. Second, the cost components are incomplete, failing to cover all expenditures during the startup process, including fuel consumption, equipment wear and tear, ancillary services, and environmental treatment. Third, they have poor adaptability to operating conditions, failing to consider the quantitative impact of dynamic factors such as pre-startup downtime and startup load ramp-up rate on costs. Fourth, the calculation methods often employ linear models or simple superposition, making it difficult to fit the nonlinear coupling relationships between multiple cost items.

[0059] Based on this, the present invention provides a method, system, medium, and equipment for calculating the start-up cost of thermal power units based on an improved gradient boosting tree. By constructing a four-layer model system—differentiated modeling of unit characteristics, detailed breakdown of start-up costs, quantification of dynamic influencing factors, and optimization of nonlinear algorithms—it first establishes differentiated characteristic models for three types of units: coal-fired, gas-fired, and biomass units, accurately depicting their fuel combustion characteristics and equipment wear patterns. Then, it constructs a multi-dimensional detailed model of start-up costs, breaking down the total start-up cost into four major categories: fuel consumption, equipment wear, ancillary services, and environmental treatment, comprehensively covering all expenditures during the start-up process. On this basis, it quantifies two types of dynamic influencing factors: pre-start-up downtime and start-up load ramp-up rate, establishing the correlation between these factors and each cost item, enabling the model to adapt to multiple operating conditions. Finally, it designs a cost solution method based on an improved gradient boosting tree, using feature engineering optimization and Huber loss function improvement to accurately fit the nonlinear coupling relationship between multiple cost items, significantly improving the accuracy of start-up cost calculation. By addressing the technical bottlenecks of existing technologies from four dimensions—differentiated modeling of unit types, comprehensive decoupling of cost composition, dynamic factor quantification and embedding, and nonlinear algorithm optimization—the present invention systematically solves these technical bottlenecks.

[0060] Example 1

[0061] like Figure 1 As shown in the figure, this embodiment discloses a method for calculating the start-up cost of thermal power units based on an improved gradient boosting tree, including the following steps:

[0062] S1: Based on the differences between coal-fired, gas-fired, and biomass generator sets, establish corresponding unit characteristic models respectively;

[0063] S2: Based on the unit characteristic model, the start-up cost is divided into four categories: fuel consumption cost, equipment wear and tear cost, auxiliary service cost, and environmental treatment cost, and a multi-dimensional breakdown model of start-up cost is constructed.

[0064] S3: Quantify the downtime before startup and the startup load ramp-up rate as dynamic influencing factors, and associate them with each cost item;

[0065] S4: An improved gradient boosting tree algorithm is used to fit the nonlinear relationship between each cost item and the dynamic influence factor. Based on the multi-dimensional subdivision model, the total start-up cost and the calculation results of each subdivision cost item are obtained.

[0066] Next, combined Figure 1 This embodiment provides a detailed explanation of a method for calculating the start-up cost of thermal power units based on an improved gradient boosting tree.

[0067] In S1, a differentiated characteristic model for three types of thermal power units is established.

[0068] This step is used to accurately characterize the essential differences between coal-fired, gas-fired, and biomass power generation units in terms of fuel characteristics, equipment configuration, and start-up procedures, and to identify the core influencing factors of start-up costs for different units, providing data support for subsequent cost breakdown modeling and dynamic factor quantification.

[0069] 1. Characteristic Modeling of Coal-fired Units

[0070] The key features are the fuel combustion characteristics and equipment wear characteristics during the startup process of coal-fired power units:

[0071] (1) Fuel combustion model of coal-fired power unit:

[0072] (1)

[0073] in, This refers to the effective calorific value of fuel during the startup process of a coal-fired power unit. The lower heating value of coal is calculated based on the base calorific value. This represents the total coal consumption during startup. This refers to the coal combustion efficiency during the start-up phase, which differs from the rated operating efficiency. Combustion is incomplete during the start-up phase, and the value ranges from 0.65 to 0.85, reflecting the combustion characteristics during the start-up period. This represents the ash content of coal, reflecting the impact of ash on the heat released during combustion.

[0074] Traditional models directly apply the rated operating condition combustion efficiency without considering the incomplete combustion caused by low furnace temperature and unstable combustion conditions during the unit startup phase, resulting in a large deviation between the calculated results and the actual results.

[0075] This model incorporates a specific combustion efficiency for the start-up phase. This allows for matching the combustion state under startup conditions, while also incorporating the ash content influence coefficient. The quantification of ash content's inhibitory effect on combustion heat release. These two improvements make the calculation of effective calorific value of fuel more closely reflect actual startup scenarios, significantly reducing errors caused by ignoring the unique combustion patterns and fuel composition effects of startup, and improving the accuracy of the basic data for cost calculation.

[0076] (2) Model of heat transfer surface loss in coal-fired power units:

[0077] (2)

[0078] in, Costs related to heat loss during startup; It is the unit loss coefficient of the heating surface of a coal-fired power unit, which is related to the steel material and the service life, and reflects the characteristics of the equipment itself. This refers to the temperature change of the heated surface during the startup phase. The duration of the startup; This represents the total area of ​​the heating surface of the coal-fired power unit.

[0079] Traditional models for calculating the start-up costs of coal-fired power units often focus on fuel consumption, neglecting the special operating conditions during the start-up phase: the heating surfaces need to be rapidly heated from room temperature to operating temperature, and the sudden temperature rise will cause thermal stress concentration in the materials, resulting in additional losses far exceeding those under normal operating conditions. This loss has long been excluded from cost accounting, leading to incomplete calculations of equipment loss costs.

[0080] This model constructs a dedicated heat-receiving surface loss model, which increases the unit loss coefficient. Start-up temperature variation range Startup duration Total area of ​​heated surface Quantitative correlation was performed, enabling accurate calculation of this specific type of loss. This not only fills the gap in traditional models for calculating specific equipment losses, but also makes the composition of start-up costs more comprehensive, allowing the equipment loss costs during the unit start-up phase to be quantified and traceable. This provides data support for the unit's full life cycle cost management and also provides a scientific basis for equipment maintenance planning and cost control.

[0081] In this embodiment, the characteristic modeling of coal-fired power units focuses on fuel combustion characteristics and equipment loss characteristics. The accurate characterization of these two characteristics comprehensively covers the core components of the start-up cost of coal-fired power units, providing reliable basic parameters for subsequent cost breakdown modeling and dynamic factor quantification. This effectively improves the comprehensiveness and accuracy of the start-up cost calculation of coal-fired power units and provides scientific support for unit start-up scheduling optimization and pricing decisions.

[0082] 2. Gas turbine unit characteristic modeling

[0083] Focusing on the fuel consumption and ignition energy characteristics of gas turbine units during rapid start-up:

[0084] (3)

[0085] in, This represents the total energy consumption during the startup of the gas turbine unit. The lower heating value per unit volume of natural gas; This refers to the amount of natural gas consumed during startup. The combustion efficiency of the gas turbine unit is as high as 0.92-0.96 during the start-up phase, demonstrating the characteristic of complete gas combustion. Energy consumption for ignition startup reflects the additional energy requirements of the gas turbine unit during the ignition process.

[0086] Traditional models fail to distinguish the fundamental differences between natural gas and coal when calculating the start-up costs of gas turbine units. They simply apply the combustion efficiency parameters of coal-fired units. However, natural gas, as a clean and efficient fuel, has a combustion reaction rate and energy release efficiency far exceeding that of coal. Blindly applying these parameters will lead to serious distortions in energy consumption calculations. At the same time, they ignore the ignition energy consumption during the start-up of gas turbine units. Gas turbine units rely on high-pressure ignition systems for rapid start-up, and this additional energy requirement is a significant component of start-up costs. Failure to calculate this separately will result in incomplete cost coverage.

[0087] Based on this, this model is specifically optimized, on the one hand by setting a dedicated combustion efficiency for the gas turbine unit. This is to match the characteristics of complete combustion of natural gas; on the other hand, it adds an ignition energy consumption item. The energy loss during the ignition process is quantified separately. These two improvements make the calculation of the total energy consumption for gas turbine startup more closely reflect its unique characteristics of rapid startup, efficient combustion, and significant ignition energy consumption, greatly improving the calculation accuracy and solving the problem of poor adaptability caused by the "one-size-fits-all" approach of traditional models. This provides support for the accurate calculation of gas turbine startup costs.

[0088] 3. Characteristic Modeling of Biomass Power Generator Units

[0089] Highlighting the unique characteristics of biomass feedstock pretreatment and combustion:

[0090] (4)

[0091] in, This refers to the total energy consumption during the startup of the biomass unit. The lower heating value of biomass feedstock is received on a basis. This refers to the amount of biomass raw materials consumed during the startup process; The biomass combustion efficiency is 0.6-0.75 during the start-up phase, and is greatly affected by the characteristics of the feedstock. Energy consumption for raw material drying before startup is specifically designed for the pretreatment needs of biomass raw materials with high moisture content; This is the moisture content coefficient.

[0092] Biomass power plants and coal-fired power plants differ fundamentally in their feedstock characteristics and start-up processes. Traditional models treat them as equivalent, leading to significant cost errors. Biomass feedstocks (such as agricultural and forestry waste) naturally have high moisture content and must undergo drying pretreatment before startup, a process that consumes a large amount of energy. Coal-fired power plants do not require such pretreatment, and traditional models do not include this energy consumption, directly resulting in omissions in cost calculations. Furthermore, the moisture content of the feedstock significantly affects combustion efficiency; higher moisture content leads to less complete combustion and lower energy conversion efficiency. Traditional models fail to quantify this impact, using only a fixed combustion efficiency, further exacerbating calculation errors.

[0093] Based on this, in order to meet the needs of biomass units, this model firstly adds a drying energy consumption item. First, the energy consumption of the raw material pretreatment stage is calculated separately; second, a moisture content coefficient is introduced. Quantifying the inhibitory effect of moisture content on combustion heat release, combined with biomass-specific combustion efficiency This enables precise calculation of energy consumption. It fills the gap in traditional models regarding the calculation of specific energy consumption items and the impact of raw material characteristics on biomass power units, making cost structure more comprehensive and data more accurate. This provides a scientific basis for the rational pricing and optimized dispatch of biomass power units in the electricity market.

[0094] In this embodiment, the constructed model breaks through the limitations of the traditional model of "uniform formula and ignoring differences", accurately depicting the essential differences between coal-fired, gas-fired and biomass units in terms of fuel characteristics (the influence of coal ash content, the clean and efficient nature of natural gas, and the high moisture content of biomass), equipment configuration (the heating surface scale of coal-fired units and the ignition system of gas-fired units), and start-up process (pre-treatment of biomass raw materials drying), and clarifies the core influencing factors of start-up costs for various types of units.

[0095] By designing dedicated parameters and quantitative models for each type of unit, a shift from general fitting to precise adaptation has been achieved. This not only fills the gaps in the traditional model's calculation of specific energy consumption, equipment losses, and other detailed items, but also solves the problem of calculation distortion caused by type confusion. It provides reliable basic data support for subsequent multi-dimensional breakdown of start-up costs and quantification of dynamic influencing factors, thereby improving the comprehensiveness, adaptability, and accuracy of the entire cost calculation system from the source. It also provides a scientific basis for the optimization of start-up scheduling of various thermal power units and electricity market pricing.

[0096] In S2, a multi-dimensional breakdown model of startup costs is constructed.

[0097] Based on the differentiated characteristic models of the three types of thermal power units, the start-up cost is broken down into four major categories: fuel consumption cost, equipment wear and tear cost, auxiliary service cost, and environmental treatment cost. This comprehensively covers all cost expenditures during the start-up process and provides a complete framework for accurate calculation.

[0098] 1. Fuel consumption cost

[0099] Integrate energy consumption data from the unit characteristic model and combine it with energy price calculations:

[0100] (5)

[0101] in, Let k be the fuel consumption cost of the k-th type of unit, where k = coal for coal, k = gas for gas, and k = bio for biomass. The total energy consumption for starting up the k-th type of unit; For the k-th type of generating unit, the unit price is the converted value of coal, the unit price of natural gas, and the unit price of raw material purchase for biomass.

[0102] 2. Equipment depreciation costs

[0103] It is further broken down into three sub-items: heat transfer surface loss, rotating equipment loss, and electrical equipment loss.

[0104] (6)

[0105] (7)

[0106] (8)

[0107] in, Cost of heat exchange surface loss; Starting losses for rotating equipment such as steam turbines and fans; Let be the unit starting loss coefficient of the rotating equipment of the k-th type of unit. This represents the cumulative number of launches corresponding to this launch. Starting losses for electrical equipment such as transformers and switches; Let be the loss coefficient of the electrical equipment of the k-th type of unit. This represents the average current during the startup phase. This refers to the startup duration.

[0108] Traditional models simply estimate equipment losses without breaking them down into smaller components. This model, however, precisely quantifies the startup losses of different devices through multiple sub-items, making equipment cost calculations more realistic.

[0109] Traditional models lack detailed consideration of equipment loss costs during the startup phase of thermal power units, often employing single fixed coefficients or general estimation methods. They fail to differentiate the loss mechanisms and influencing factors of different types of equipment, such as heating surfaces, rotating equipment, and electrical equipment, leading to significant discrepancies between equipment loss costs and actual expenditures. Furthermore, the loss logic of different equipment during startup varies significantly: heating surface losses are directly related to temperature changes and startup duration; rotating equipment losses accumulate with the number of startups; and electrical equipment losses are closely related to startup current and duration.

[0110] This model specifically breaks down equipment loss costs into four sub-items. This multi-item detailed quantification method accurately captures the loss characteristics of different equipment, solves the problem of the crude "one-size-fits-all" estimation of traditional models, and makes the calculation of equipment loss costs more in line with actual operating scenarios. It not only improves the accuracy of the overall calculation of start-up costs, but also provides refined data support for equipment life cycle management, maintenance plan optimization and loss cost control.

[0111] 3. Ancillary service costs

[0112] Calculate the energy consumption and service expenses of auxiliary equipment during the startup process:

[0113] (9)

[0114] in, Power consumption of auxiliary equipment such as water pumps and oil pumps during the start-up phase; For grid connection electricity price; The expenditure for external auxiliary services such as compressed air and steam for startup is flexibly determined based on the configuration of the unit's own auxiliary systems.

[0115] 4. Environmental treatment costs

[0116] Considering the costs of pollutant treatment during the start-up phase, with particular emphasis on the environmental characteristics of coal-fired and biomass units:

[0117] (10)

[0118] in, Pollutants in the start-up phase For coal-fired power units, the additional emissions from the inefficient initial emissions during the startup phase of the desulfurization and denitrification systems must be considered. Biomass units need to calculate dust levels. With nitrogen oxides emission; pollutants The unit treatment cost is determined based on the type of environmental protection equipment in the unit.

[0119] With increasingly stringent environmental protection requirements, the cost of pollutant emissions and treatment during the start-up phase of thermal power units has become a significant and unavoidable expense. However, traditional models generally overlook this crucial cost item, resulting in an incomplete composition of start-up costs and a disconnect from current green development requirements. During the start-up phase, the unit's operating conditions are unstable, and environmental protection systems (such as desulfurization, denitrification, and dust removal equipment) often operate inefficiently, leading to fluctuating pollutant emission concentrations and unique treatment costs.

[0120] This model incorporates environmental treatment costs into its separate accounting, and precisely quantifies SO2 and NO emissions for units with significant pollutant emission characteristics, such as those from coal-fired or biomass-fired power plants. x The emissions of various pollutants, such as particulate matter and dust, are calculated in conjunction with the corresponding unit treatment costs. This calculation fully considers the inefficient emission increments during the initial startup of desulfurization and denitrification systems in coal-fired units, as well as the unique characteristics of dust and nitrogen oxide emissions from biomass units, making the accounting more targeted. This not only fills the gap in traditional models for environmental cost calculation, making the startup cost structure more comprehensive and realistic, guiding units to prioritize environmental protection investment during startup and scheduling, and promoting the green and low-carbon operation of thermal power units, but also provides a scientific basis for electricity market pricing that includes environmental costs.

[0121] 5. Total startup cost

[0122] The total startup cost is obtained by summing the four categories of subdivided costs:

[0123] (11)

[0124] in, This represents the total startup cost of the k-th type of generator unit.

[0125] This embodiment breaks through the limitations of the traditional model's coarse cost division, breaking down the start-up cost system into four major subcategories: fuel consumption, equipment depreciation, ancillary services, and environmental treatment. These are further refined into multiple sub-items with corresponding quantitative formulas. Fuel consumption costs integrate characteristic model data and energy prices to ensure accurate basic cost calculations; equipment depreciation costs are subdivided by equipment type to capture differences in depreciation across different equipment; ancillary service costs cover auxiliary equipment energy consumption and external service expenditures, balancing flexibility; and environmental treatment costs comply with regulatory requirements, filling gaps in traditional calculations. This multi-dimensional breakdown provides clear accounting basis for all start-up cost expenditures, comprehensively covering all cost sources during the start-up process. It solves the problem of incomplete cost coverage in traditional models and provides a structured and complete calculation framework for subsequent dynamic influencing factor embedding and nonlinear algorithm solutions, significantly improving the comprehensiveness, refinement, and practical adaptability of start-up cost calculations.

[0126] In S3, the dynamic influencing factors of the quantified start-up condition are determined.

[0127] Based on the cost breakdown model, two core dynamic influencing factors, namely the downtime before startup and the startup load ramp-up rate, are introduced to establish the correlation between the factors and each breakdown cost item, so that the model can adapt to the calculation needs of different startup conditions.

[0128] 1. Factors affecting the duration of downtime before startup

[0129] Characterizing the impact of downtime on unit startup energy consumption and losses:

[0130] (12)

[0131] in, The outage duration influencing factor for unit k is... ≥1, the longer the downtime, the larger the factor; The total downtime of the unit before startup; The shutdown impact coefficient for the k-th type of unit is 0.3-0.5 for coal-fired units, 0.1-0.2 for gas-fired units, and 0.2-0.35 for biomass units, reflecting the differences in insulation and cooling characteristics of different units; is the characteristic time constant of the k-th type of unit, reflecting the unit's temperature decay rate.

[0132] This factor primarily affects fuel consumption cost C1 and equipment wear and tear cost C2, and the correction formula is as follows:

[0133] (13)

[0134] (14)

[0135] in, It is the first Factors affecting the downtime of similar generating units.

[0136] Traditional models for calculating the startup costs of thermal power units do not consider the differences in downtime before startup and use fixed parameters, making them unsuitable for real-world scenarios involving frequent start-ups and shutdowns as well as long-term downtime. This embodiment addresses this by quantifying the relationship between downtime and fuel consumption and equipment wear costs, and by setting specific coefficients based on the insulation and cooling characteristics of different units. The longer the downtime, the greater the factor correction. This allows the model to dynamically match the startup energy consumption and wear patterns under different downtime conditions, solving the problem of insufficient adaptability in traditional models and significantly improving the accuracy of cost calculations in complex start-up and shutdown scenarios.

[0137] 2. Factors affecting the start-up load ramp-up rate

[0138] The dynamic impact of quantified load ramp-up rate on startup costs:

[0139] (15)

[0140] in, For the load ramp-up rate influencing factor of the k-th type of unit, ≥1, the greater the deviation from the optimal climb rate, the larger the factor; This represents the actual start-up load ramp-up rate. The climbing influence coefficient is 0.25-0.4 for gas-fired units, 0.15-0.3 for coal-fired units, and 0.2-0.38 for biomass units, reflecting the climbing flexibility advantage of gas-fired units. The optimal load ramp-up rate for the k-th type of unit is determined by the unit's design parameters and operating experience.

[0141] This factor primarily affects fuel consumption cost C1 and ancillary service cost C3, and the correction formula is as follows:

[0142] (16)

[0143] (17)

[0144] in, It is the factor affecting the start-up load ramp-up rate of the k-th type of unit.

[0145] Traditional models assume a fixed ramp rate, neglecting the differences in ramp strategies resulting from variations in grid load demand during actual dispatch. This embodiment introduces a load ramp rate influence factor, quantifying the deviation between the actual and optimal rates. Combined with ramp flexibility coefficients for various generating units, the correction is stronger the deviation, primarily affecting fuel consumption and ancillary service costs. This allows the model to flexibly respond to various ramp demands, including fast, medium, and slow ramps, breaking through the static calculation limitations of traditional models and significantly improving adaptability to diverse dispatch conditions.

[0146] In this embodiment, based on the cost breakdown model, two core factors are introduced: pre-startup downtime and load ramp-up rate. By establishing dedicated quantitative formulas and correction rules, the correlation between these factors and the breakdown cost items is constructed. This breaks through the limitations of traditional models that rely on fixed parameters, enabling the model to accurately capture cost variation patterns under different downtime states and scheduling requirements. It flexibly adapts to diverse startup conditions, effectively solving the bottleneck of poor adaptability in traditional models. This further improves the practicality and accuracy of startup cost calculation, providing scientific support for unit scheduling optimization under complex operating conditions.

[0147] In S4, a cost solution method based on an improved gradient boosting tree is designed.

[0148] To address the nonlinear correlations and dynamic factor coupling characteristics in the startup cost model, an improved gradient boosting tree algorithm is adopted. Through feature engineering optimization and loss function improvement, the total startup cost can be accurately calculated.

[0149] 1. Feature Engineering Optimization

[0150] Select core input features such as unit type, pre-start downtime, start-up load ramp-up rate, energy price, pollutant treatment cost, unit operating years, and equipment rated parameters to construct a high-dimensional feature matrix.

[0151] By using mutual information entropy to filter key features and eliminate redundant information, the efficiency of model training is improved. Traditional algorithms do not perform feature filtering and are easily affected by noisy features.

[0152] 2. Improve the design of the loss function

[0153] The Huber loss function is used instead of the traditional mean squared error loss function:

[0154] (18)

[0155] Where y is the actual value of the startup cost. These are the model's predicted values. The threshold is set between 0.1 and 0.3.

[0156] Traditional mean squared error loss functions are sensitive to outliers. This embodiment uses the Huber loss function to balance robustness and error accuracy, reducing the impact of extreme startup conditions on the model.

[0157] 3. Model Training and Solving

[0158] (1) Data preprocessing: Collect actual operating data of three types of thermal power units under different start-up conditions, including energy consumption, equipment wear and tear, pollutant emissions, etc., and perform normalization and outlier processing.

[0159] (2) Gradient boosting tree training: Set hyperparameters such as decision tree depth, learning rate, and number of iterations, optimize parameter combinations through cross-validation, and build a basic model.

[0160] (3) Dynamic factor embedding: the quantized factor embedding , Factors are incorporated into the model as input features to establish a nonlinear mapping relationship between factors and each cost item.

[0161] (4) Predictive output: Input the type, operating parameters and characteristic data of the unit to be measured, and the model outputs the calculation results of the total start-up cost and each sub-cost item. Traditional linear models are difficult to fit the nonlinear relationship of multiple coupled factors.

[0162] This embodiment optimizes the model through feature engineering, selects core input features such as unit type, pre-start downtime, start-up load ramp-up rate, energy price, and pollutant treatment cost to construct a high-dimensional feature matrix, and uses mutual information entropy to filter key features, eliminate redundant information and noise interference, effectively reduce model complexity, improve training efficiency, and avoid the defects of traditional algorithms that are easily affected by noise features.

[0163] Meanwhile, the Huber loss function is used instead of the traditional mean squared error loss function, which balances the robustness and error accuracy of the model, reduces the impact of extreme start-up conditions on the model, and enables the model to remain stable and reliable when facing complex and fluctuating real-world operating data.

[0164] Furthermore, by quantifying the dynamic factors , An embedded model establishes a nonlinear mapping relationship between factors and various cost items, enabling cost calculation to dynamically respond to real-time operating condition changes during unit start-up and shutdown. This solves the problem that traditional linear models struggle to fit the coupled nonlinear relationships of multiple factors, providing more timely decision support for the economic dispatch and cost control of thermal power units. Furthermore, by improving the gradient boosting tree algorithm, the accuracy of start-up cost calculation is significantly enhanced compared to traditional models, making it adaptable to complex and diverse start-up scenarios.

[0165] This specific embodiment proposes three differentiated characteristic models for thermal power units. It designs exclusive modeling parameters and formulas for the essential differences between coal-fired, gas-fired, and biomass units, breaking through the traditional one-size-fits-all calculation mode and accurately depicting the startup characteristics of different units. Simultaneously, it constructs a multi-dimensional cost breakdown system, decomposing startup costs into four major sub-items, covering all cost components including fuel, equipment, ancillary services, and environmental protection, solving the problem of incomplete cost accounting in traditional models. Secondly, it introduces dynamic influencing factors to quantify the impact of operating conditions, establishing the correlation between pre-startup downtime, load ramp-up rate, and cost items, enabling the model to adapt to multiple operating conditions and improving the flexibility and applicability of the calculation. Furthermore, it employs an improved gradient boosting tree algorithm for solving the problem, and through feature engineering and loss function optimization, efficiently handles nonlinear coupling relationships, improving the calculation accuracy compared to traditional linear models. This provides technical support for accurate startup cost accounting, enabling accurate calculation of startup costs for different types of thermal power units under multiple operating conditions.

[0166] Example 2

[0167] This embodiment provides a system for calculating the start-up cost of thermal power units based on an improved gradient boosting tree, including:

[0168] The characteristic modeling module is used to establish corresponding unit characteristic models based on the differences between coal-fired, gas-fired, and biomass generator sets.

[0169] The cost breakdown module is used to divide the start-up cost into four categories of cost items: fuel consumption cost, equipment depreciation cost, auxiliary service cost, and environmental treatment cost, based on the unit characteristic model, and to construct a multi-dimensional breakdown model of start-up cost.

[0170] The dynamic factor introduction module is used to quantify the pre-startup downtime and startup load ramp-up rate as dynamic influencing factors and associate them with each cost item.

[0171] The cost calculation module is used to fit the nonlinear relationship between each cost item and the dynamic influence factor using an improved gradient boosting tree algorithm, and obtain the calculation results of the total start-up cost and each subdivided cost item based on the multi-dimensional subdivision model.

[0172] This embodiment achieves refined, dynamic, and intelligent cost calculation through modular design. First, the characteristic modeling module constructs characteristic models for different types of generator units, such as coal-fired, gas-fired, and biomass units, ensuring the relevance and adaptability of cost calculation. Next, the cost breakdown module decomposes start-up costs into four dimensions: fuel consumption, equipment wear and tear, ancillary services, and environmental treatment, constructing a multi-dimensional breakdown model to clarify and trace the cost composition. The dynamic factor introduction module quantifies and correlates key dynamic factors such as pre-startup downtime and load ramp-up rate to each cost item, enabling cost calculation to accurately respond to changes in operating conditions. Finally, the cost solution module uses an improved gradient boosting tree algorithm to fit nonlinear relationships, obtaining accurate calculation results. Overall, this system effectively solves the problems of traditional calculation methods being difficult to adapt to multiple types of units and unable to dynamically reflect the impact of operating conditions. It effectively improves the accuracy, efficiency, and practicality of start-up cost calculation for thermal power units, providing reliable technical support for unit economic scheduling and cost control.

[0173] Example 3

[0174] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the method for calculating the start-up cost of thermal power units based on an improved gradient boosting tree as described in Embodiment 1 above.

[0175] This invention can take the form of a computer program product on a computer-readable storage medium having computer-readable program instructions (e.g., computer software) contained therein. More specifically, this invention can take the form of network-implemented computer software. Any suitable computer-readable storage medium can be used, including but not limited to: non-transitory computer-readable media such as hard disk drives, CD-ROMs, DVD-ROMs, optical storage devices, magnetic storage devices (e.g., magnetic tapes, floppy disks), flash memory storage devices (e.g., USB flash drives, SD cards, solid-state drives, SSDs), persistent storage in cloud storage services, etc. These storage media can be locally deployed physical storage devices or distributed storage nodes accessed via a network, as long as they can persistently store the computer program and be invoked and executed by a processor.

[0176] In some embodiments, the computer program stored on the computer-readable storage medium includes all or part of the instructions for implementing the method for calculating the startup cost of thermal power units. These instructions can be divided into multiple functional modules, such as: characteristic modeling module instructions, cost breakdown module instructions, dynamic factor introduction module instructions, cost solution module instructions, etc. When a processor (including CPU, GPU, FPGA, or dedicated acceleration chip, etc.) reads and executes these instructions from the computer-readable storage medium, it can sequentially complete steps such as thermal power unit characteristic modeling, multi-dimensional cost breakdown, dynamic influence factor quantification and correlation, improved gradient boosting tree model training, and cost calculation, thereby accurately outputting the calculation results of the total startup cost and each breakdown cost item.

[0177] In some embodiments, the computer-readable storage medium may also store auxiliary data related to the calculation of start-up costs of thermal power units, including but not limited to: historical operating datasets of different types of units (coal-fired, gas-fired, biomass, etc.), hyperparameter configuration files of the improved gradient boosting tree model, and dynamic influencing factors (…). , The system includes a quantification rule table and cost correction rules for abnormal operating conditions. These auxiliary data work in conjunction with the computer program, enabling the processor to directly call pre-stored feature matrices, loss function parameters, and model weights when executing the measurement method, without having to repeatedly perform data preprocessing and model initialization, thereby improving measurement efficiency and response speed.

[0178] In some embodiments, the computer program on the computer-readable storage medium can also be configured to support multi-scenario deployment and multi-device collaborative execution: when the program is deployed in the storage medium of a terminal device (such as an industrial control computer or a portable computing terminal), the processor can perform local real-time calculations to meet the needs of rapid on-site decision-making; when the program is deployed in the distributed storage medium of a server cluster, it can be executed in parallel by multiple nodes to achieve batch calculations of large-scale historical data and iterative optimization of models; when the program is deployed in a cloud storage service, it can be remotely invoked through a network interface to support cost control and economic scheduling analysis across regions and multiple units.

[0179] It should be noted that the computer-readable storage medium of the present invention is not simply a signal carrier or transmission medium, but rather carries program instructions and auxiliary data that are deeply bound to the method for calculating the start-up cost of thermal power units. These instructions and data together constitute the core carrier for realizing the technical solution of the present invention, enabling innovations such as improved gradient boosting tree algorithm, feature engineering optimization, and Huber loss function to be implemented in specific hardware environments. This solves the technical problems that traditional linear models are difficult to fit nonlinear correlations and cannot dynamically respond to changes in operating conditions, thereby improving the accuracy, robustness, and practicality of calculating the start-up cost of thermal power units.

[0180] Example 4

[0181] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the method for calculating the start-up cost of thermal power units based on an improved gradient boosting tree as described in Embodiment 1 above. Specifically:

[0182] Typically, computer equipment includes a processor and memory.

[0183] The processor can include one or more processing cores, such as a 4-core processor or a 16-core processor. In scenarios involving large-scale historical data batch calculations, multi-core processors with 32 or more cores can be used to improve parallel computing capabilities and model training efficiency. The processor can be implemented using at least one hardware form of Digital Signal Processing (DSP), Field Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). FPGA can be used to accelerate and improve the decision tree splitting and feature selection process in gradient boosting tree algorithms, significantly reducing model inference latency. The processor can also include a main processor and coprocessors. The main processor, also known as the central processing unit (CPU), is used to process data in the wake-up state and is mainly responsible for overall process scheduling, data preprocessing, and result output. The coprocessor is a low-power processor used to process data in the standby state and can be used to collect real-time thermal power unit operating data and perform preliminary noise reduction processing. In some embodiments, the processor may include a graphics processing unit (GPU), which is responsible for performing computationally intensive tasks in the method steps provided by the present invention, such as high-dimensional feature matrix operations, Huber loss function gradient calculation, parallel training of decision trees, etc., and achieves large-scale parallel acceleration through CUDA or OpenCL frameworks, so that the model can still maintain high efficiency when processing massive amounts of unit operation data.

[0184] The memory may include one or more computer-readable storage media, which may be non-transitory. The memory may also include high-speed random access memory (RAM) for temporarily storing intermediate data, feature matrices, and gradient information during model training, ensuring the processor can quickly read and write to improve computational efficiency; and non-volatile memory, such as one or more disk storage devices (e.g., HDDs, SSDs), flash storage devices (e.g., USB flash drives, eMMC, NVMe storage), etc., for persistently storing computer programs, historical unit operation datasets, model weight files, and hyperparameter configurations. In some embodiments, the non-transitory computer-readable storage media in the memory is used to store a computer program configured to be executed by one or more processors (e.g., GPUs) to implement the aforementioned method for calculating the startup cost of thermal power units based on an improved gradient boosting tree. Specifically, this includes: characteristic modeling module instructions, cost breakdown module instructions, dynamic factor introduction module instructions, cost solution module instructions, etc. These module instructions work collaboratively, enabling the processor to sequentially complete the entire process of unit characteristic modeling, cost dimension breakdown, dynamic factor quantization, improved gradient boosting tree model fitting, and cost calculation result output.

[0185] In some embodiments, the computer device may further include a communication interface for data interaction with the SCADA system, DCS system, or data acquisition terminal of the thermal power unit, to obtain real-time unit start-up and shutdown status data, energy consumption data, and equipment status information; it may also include an input / output module for receiving configuration information such as unit type and start-up parameters input by the user, and outputting calculation results such as total start-up cost, proportion of each sub-cost item, and influence weight of dynamic factors, which can be presented in a visual form such as reports and charts.

[0186] In different application scenarios, the computer equipment can take various hardware forms: when deployed locally in a thermal power plant, an industrial control computer (ICC) or server can be used, possessing high stability and anti-interference capabilities, and adapting to complex on-site environments; when deployed in the cloud, a cloud server instance can be used, supporting batch cost calculation and model iteration for multiple units and across regions through elastic computing power scheduling; when used for rapid on-site decision-making, a portable computing terminal or mobile workstation can be used, meeting the requirements of portability and real-time performance. Regardless of the hardware form, the computer equipment, through the collaborative work of the processor and memory, implements improved gradient boosting tree algorithms, feature engineering optimization, and Huber loss functions into executable technical solutions, effectively solving the problems of traditional calculation methods being difficult to adapt to multiple types of units and unable to dynamically respond to changes in operating conditions, significantly improving the accuracy, efficiency, and practicality of thermal power unit start-up cost calculation.

[0187] The steps or modules involved in Embodiments 2 to 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0188] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for calculating the start-up cost of thermal power units based on an improved gradient boosting tree, characterized in that, include: Based on the differences between coal-fired, gas-fired, and biomass power generation units, corresponding unit characteristic models are established for each type. Based on the unit characteristic model, the start-up cost is divided into four categories: fuel consumption cost, equipment wear and tear cost, auxiliary service cost, and environmental treatment cost, and a multi-dimensional subdivision model of start-up cost is constructed. The pre-startup downtime and start-up load ramp-up rate are quantified as dynamic influencing factors and linked to various cost items; among them, the pre-startup downtime dynamic influencing factor affects fuel consumption costs and equipment wear and tear costs; specifically: ； in, The outage duration influencing factor for type k generating units. ≥1; The total downtime of the unit before startup; This is the shutdown impact coefficient for unit k. Different units have different shutdown impact coefficients, reflecting the differences in insulation and cooling characteristics of different units. The characteristic time constant of the k-th type of unit reflects the temperature decay rate of the unit; The dynamic influencing factor of start-up load ramp-up rate affects fuel consumption costs and ancillary service costs; specifically: ； in, For the load ramp-up rate influencing factor of the k-th type of unit, ≥1; This represents the actual start-up load ramp-up rate. Let K be the climb impact coefficient for the k-th type of unit; The optimal load ramp-up rate for the k-th type of unit is determined by the unit's design parameters and operating experience. An improved gradient boosting tree algorithm is used to fit the nonlinear relationship between each cost item and the dynamic influence factor. Based on the multi-dimensional subdivision model, the total start-up cost and the calculation results of each subdivision cost item are obtained.

2. The method for calculating the start-up cost of thermal power units based on an improved gradient boosting tree as described in claim 1, characterized in that, The unit characteristic model is used to quantify the total energy consumption of coal-fired, gas-fired, and biomass generator units under differences in fuel characteristics, equipment configuration, and start-up procedures.

3. The method for calculating the start-up cost of thermal power units based on an improved gradient boosting tree as described in claim 1, characterized in that, The specific fuel consumption cost is as follows: ； in, Let k be the fuel consumption cost of the k-th type of unit, where k = coal for coal, k = gas for gas, and k = bio for biomass. The total energy consumption for starting up the k-th type of unit; For the unit price of energy corresponding to the k-th type of unit, the unit price of coal is used for coal-fired power, the unit price of natural gas is used for gas-fired power, and the unit price of raw material purchase is used for biomass. The specific equipment depreciation cost is as follows: ； ； ； in, Cost of heat exchange surface loss; The cost of starting up rotating equipment; Let be the unit starting loss coefficient of the rotating equipment of the k-th type of unit. This represents the cumulative number of launches corresponding to this launch. Costs associated with starting losses of electrical equipment; Let be the loss coefficient of the electrical equipment of the k-th type of unit. This represents the average current during the startup phase. Startup time; The specific cost of the ancillary services is as follows: ； in, Power consumption of auxiliary equipment during the startup phase; For grid connection electricity price; Expenditure to initiate external ancillary services; The specific environmental treatment cost is as follows: ； in, Pollutants in the start-up phase For coal-fired power units, the additional emissions from the inefficient initial emissions during the startup phase of the desulfurization and denitrification systems must be considered. Biomass units need to calculate dust levels. With nitrogen oxides emission; pollutants The unit treatment cost is determined based on the type of environmental protection equipment in the unit.

4. The method for calculating the start-up cost of thermal power units based on an improved gradient boosting tree as described in claim 1, characterized in that, The improved gradient boosting tree algorithm is used to fit the nonlinear relationship between each cost item and the dynamic influence factor. Based on the multi-dimensional segmentation model, the total start-up cost and the calculation results of each segmented cost item are obtained, specifically including: The input features are filtered by mutual information entropy to remove redundant features; the input features include unit type, pre-start downtime, start-up load ramp-up rate, energy price, pollutant treatment cost and equipment rated parameters. The Huber loss function is used instead of the traditional mean squared error loss function; Using dynamic influencing factors as feature inputs, a nonlinear mapping relationship between them and each cost item is established through gradient boosting tree training, and the prediction results of total startup cost and each sub-cost item are output.

5. A system for calculating the start-up cost of thermal power units based on an improved gradient boosting tree, characterized in that, include: The characteristic modeling module is used to establish corresponding unit characteristic models based on the differences between coal-fired, gas-fired, and biomass generator sets. The cost breakdown module is used to divide the start-up cost into four categories of cost items: fuel consumption cost, equipment depreciation cost, auxiliary service cost, and environmental treatment cost, based on the unit characteristic model, and to construct a multi-dimensional breakdown model of start-up cost. The dynamic factor introduction module quantifies the pre-startup downtime and start-up load ramp-up rate as dynamic influencing factors and associates them with various cost items. Specifically, the pre-startup downtime dynamic influencing factor affects fuel consumption costs and equipment wear and tear costs. ； in, The outage duration influencing factor for type k generating units. ≥1; The total downtime of the unit before startup; This is the shutdown impact coefficient for unit k. Different units have different shutdown impact coefficients, reflecting the differences in insulation and cooling characteristics of different units. The characteristic time constant of the k-th type of unit reflects the temperature decay rate of the unit; The dynamic influencing factor of start-up load ramp-up rate affects fuel consumption costs and ancillary service costs; specifically: ； in, For the load ramp-up rate influencing factor of the k-th type of unit, ≥1; This represents the actual start-up load ramp-up rate. Let K be the climb impact coefficient for the k-th type of unit; The optimal load ramp-up rate for the k-th type of unit is determined by the unit's design parameters and operating experience. The cost calculation module is used to fit the nonlinear relationship between each cost item and the dynamic influence factor using an improved gradient boosting tree algorithm, and obtain the calculation results of the total start-up cost and each subdivided cost item based on the multi-dimensional subdivision model.

6. The thermal power unit start-up cost calculation system based on an improved gradient boosting tree as described in claim 5, characterized in that, The unit characteristic model is used to quantify the total energy consumption of coal-fired, gas-fired, and biomass generator units under differences in fuel characteristics, equipment configuration, and start-up procedures.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the method for calculating the start-up cost of thermal power units based on an improved gradient boosting tree as described in any one of claims 1-4.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the method for calculating the start-up cost of thermal power units based on an improved gradient boosting tree as described in any one of claims 1-4.

Images