A multi-working-condition adaptive power plant real-time full cost accounting method and system

Abstract

The application discloses a multi-working condition adaptive power plant real-time full cost accounting method and system. The method comprises the following steps: acquiring unit operation parameter flow in real time, extracting statistical characteristics through a sliding time window and inputting a working condition classification model, and outputting a working condition dynamic trajectory vector and a transition state membership degree vector; generating a fixed cost real-time apportioning amount by adopting a three-level closed-loop dynamic apportioning model of prediction-apportioning-backtracking; calculating real-time variable and environmental comprehensive costs; calculating creep loss cost and fatigue loss cost based on variable load rate components; adopting an equivalent enthalpy drop method to perform heat-electricity cost apportioning when combined heat and power is generated; and taking the working condition dynamic trajectory vector and the transition state membership degree vector as independent variables to query a dynamic weighting function, weighting and summing each cost element to generate a real-time full cost. The application solves the problems of working condition static description, data time scale fragmentation and aggregation weight fixation in the prior art, and the accounting precision is significantly improved.

Description

Technical Field

[0001] This invention relates to the field of electronic information technology, specifically to a real-time full-cost accounting method and system for power plants that adapts to multiple operating conditions. Background Technology

[0002] With the accelerated construction of a unified national electricity market system, the frequency of electricity spot market transactions is increasing daily, and price volatility is significantly intensifying, leading to a fundamental shift in the operational decision-making models of power generation companies. Under the traditional planned electricity generation model, power generation companies primarily focused on annual or monthly average cost control, with relatively relaxed requirements for the timeliness and accuracy of cost accounting. However, in the electricity spot market environment, generating units need to participate in bidding with a time granularity of fifteen minutes or even five minutes. The marginal cost of each trading session directly determines the rationality of the bidding strategy and the profit or loss outcome after winning the bid. Therefore, power generation companies urgently need an accounting method and system that can accurately reflect the "true full cost per kilowatt-hour" of the unit under current operating conditions in real time to support scientific bidding and operational optimization decisions.

[0003] Currently, there is a certain foundation for research on online monitoring and accounting of thermal power unit costs both domestically and internationally, but several key technical bottlenecks remain unresolved.

[0004] Firstly, in terms of operating condition processing, existing technologies generally adopt a static analysis model of "offline segmentation of typical scenarios and subsequent selection of representative operating conditions." For example, some existing methods pre-define several typical load rate ranges or operating modes (such as pure condensing, heating, and start-up / shutdown conditions), categorize real-time operating data into a discrete operating condition, and then call the corresponding static cost parameters for calculation. This approach simplifies the complex continuous operating state of the unit into a finite number of discrete points, failing to depict the continuous evolution of costs during operating condition switching. In actual operation, thermal power units frequently need to perform rapid load increases / decreases, deep peak shaving, and even start-up / shutdown operations according to grid dispatch instructions. During these transient processes, the unit's main steam parameters, reheat steam parameters, turbine efficiency, boiler combustion status, and thermal stress levels of key components are all in continuous dynamic change, and corresponding cost drivers such as fuel consumption characteristics, plant power consumption rate, and equipment lifespan wear rate also change continuously. Existing discrete operating condition classification methods inevitably produce step changes or discontinuities in cost curves at the operating condition switching boundaries, leading to underestimation or overestimation of the actual cost during load changes, and failing to provide a reliable cost benchmark for real-time pricing.

[0005] Secondly, regarding multi-source data fusion, the cost data of power generation companies exhibits inherent heterogeneity across time scales. Production operation data obtained from distributed control systems and plant-level monitoring information systems can achieve refresh rates at the second or even sub-second level, precisely depicting the dynamic changes in the unit's thermal processes. However, fixed cost data obtained from financial management systems (such as monthly depreciation, labor costs, operation and maintenance expenses, management fees, and financial expenses) are typically collected and calculated on a monthly or annual basis, representing typical low-frequency, long-cycle data. How to organically integrate high-frequency operational data with low-frequency financial data within a unified real-time computing framework is a key challenge in achieving real-time full-cost accounting. Existing technologies often address this issue by simply averaging or allocating according to planned electricity volume, distributing the total monthly fixed costs evenly or according to a preset ratio to the power generation in each time period. This method is inherently open-loop and static, unable to dynamically correct for deviations between actual and predicted loads. When the actual operating curve of the generating unit deviates significantly from the pre-predicted load curve, the fixed cost allocated to each time period will be mismatched with the actual power generation value created by the unit in that time period. This results in the cost being underestimated during peak hours and overestimated during off-peak hours, severely distorting the true economic cost signal of electricity in each time period and causing the bidding decision to deviate from the optimal strategy.

[0006] Secondly, in terms of calculating equipment lifespan loss costs, existing technologies largely rely on post-hoc statistics and assessments. A common practice is to accrue equipment depreciation or overhaul costs at the end of each month or year based on the unit's cumulative operating hours and number of start-ups and shutdowns. This accounting method, lagging behind actual operation, prevents the transient thermal stress damage during load changes and start-ups / shutdowns from being monetized in real time and included in the marginal cost of the current period. For example, a rapid load increase may cause significant fatigue damage to the turbine rotor and thick-walled boiler components in a short period, shortening their remaining lifespan. However, this cost cannot be reflected in the traditional cost accounting system in a timely manner, potentially inducing operators to blindly pursue high load change rates during peak electricity price periods while ignoring the hidden losses caused by accelerated lifespan loss. Furthermore, even if some existing technologies attempt to introduce the concepts of equivalent operating hours or equivalent start-ups / shutdowns to quantify lifespan loss, their calculations still rely on offline statistics and fail to achieve deep coupling with real-time operating condition identification. They cannot adjust the weighting of lifespan loss costs online based on the magnitude and direction of the actual load change rate.

[0007] Finally, regarding the total cost aggregation logic, existing technologies typically assign fixed weights to each cost element, or use simple linear adjustments based solely on a single static variable such as load factor. This fixed or quasi-static weighting fails to reflect the dynamic changes in the relative importance of each cost element under different operating conditions. For example, at the same 40% load level, the stress state, coal consumption characteristics, and share of fixed costs of a unit operating stably in a deep peak-shaving state differ significantly from those of a unit rapidly reducing load from high load to this state. Using the same cost aggregation weights fails to distinguish between these two fundamentally different operating scenarios, leading to a systematic deviation between the final calculated total cost and the unit's actual economic performance.

[0008] In summary, existing technologies suffer from problems such as static and discretized operating condition characterization, fragmented time scales of multi-source data lacking closed-loop correction mechanisms, lag in lifetime loss accounting compared to real-time operation, and fixed weights in full-cost aggregation. These issues have become major technical obstacles hindering power generation companies from achieving accurate cost control and scientific pricing decisions in the electricity spot market environment. Therefore, there is an urgent need for a real-time full-cost accounting solution that can achieve a leapfrog improvement from "static operating condition classification" to "dynamic trajectory recognition," from "open-loop data allocation" to "closed-loop optimization and fusion," from "post-life statistics" to "real-time damage monetization," and from "simple summation of elements" to "trajectory weighted aggregation." Summary of the Invention

[0009] This invention provides a real-time full-cost accounting method and system for power plants that adapts to multiple operating conditions. It aims to solve key problems in existing technologies, such as static operating condition characterization, fixed cost aggregation weights, fragmented time scales of multi-source data, and lag in lifetime loss calculation. This provides power generation companies with high-precision, dynamic full-cost data that is fully synchronized with the electricity spot market pricing cycle. The following detailed description of the method steps and system structure of this invention, in conjunction with specific embodiments, provides a detailed explanation.

[0010] In the method provided by this invention, the operating condition identification step is performed first. This step acquires the unit's operating parameter stream in real time. These operating parameters should include at least the unit load, main steam temperature, main steam pressure, reheat steam temperature, reheat steam pressure, feedwater flow rate, plant power consumption rate, and fuel input. The real-time value of the unit load directly determines the current power generation output level and is the basic input for subsequent calculations of fuel consumption, plant power consumption rate, and various variable costs. Main steam temperature and pressure are key thermodynamic parameters characterizing the quality of boiler outlet steam; their values ​​directly affect the turbine's heat-to-work conversion efficiency, and thus the fuel consumption level per unit of power generation. Reheat steam temperature and pressure reflect the state of steam returning to the boiler for reheating after performing work in the turbine's high-pressure cylinder. Fluctuations in these two parameters also cause changes in thermodynamic cycle efficiency and need to be considered within the scope of operating condition characteristics. Feedwater flow rate is an important monitoring parameter of the boiler feedwater system. It has a close thermal balance relationship with the unit load and fuel input, and plays an important verification role in accurately calculating fuel back-balance heat consumption. The plant power consumption rate represents the proportion of electrical energy consumed by the generating unit to its total power generation in order to maintain its own operation. This value fluctuates dynamically with changes in the operating combination of auxiliary equipment and load levels, and is a key factor in the cost of purchased power or the overall power supply cost. Fuel input is the most direct driver of variable cost calculation, and its measurement accuracy directly affects the accuracy of fuel cost accounting. The above operating parameters are continuously acquired in the form of real-time data streams from the unit's distributed control system or the plant-level monitoring information system, with the acquisition frequency preferably at the second or sub-second level, thus providing a high-temporal-resolution raw data foundation for subsequent operating condition identification and cost calculation.

[0011] To extract feature vectors reflecting the overall operating status within a short time window from a continuously flowing stream of operating parameters, facilitating pattern recognition by machine learning models, this invention employs a sliding time window technique. The width of the sliding time window is set to a first preset duration, typically between twenty and sixty seconds. This timescale selection considers both the time constant of the thermal power unit's thermal process and the requirements of the electricity spot market for cost data refresh frequency. If the window width is too short, the extracted features are easily affected by measurement noise and transient disturbances, leading to high-frequency jitter in the output of the operating condition classification model. If the window width is too long, important transient changes will be smoothed out, weakening the ability to capture dynamic operating conditions such as load variations in a timely manner. The sliding step size is set to a second preset duration, typically between two and ten seconds, and is smaller than the window width, thus forming an overlapping sliding window processing mode. This overlapping processing method ensures good temporal continuity between adjacent feature extraction results, allowing the operating condition dynamic trajectory vector and transition state membership vector output by the operating condition classification model to evolve smoothly, avoiding abrupt jumps. Within each sliding time window, time-domain statistical analysis and frequency-domain transform analysis are performed on the sequence of operating parameters contained in the window to extract representative and discriminative statistical features.

[0012] The mean calculation in time-domain statistical analysis involves taking the arithmetic mean of all operating parameters within a window to characterize the overall level of the parameters over that time period. Let the window contain... The time series composed of sampling points is Its arithmetic mean The calculation formula is The standard deviation reflects the dispersion of parameter values ​​within a window, and its calculation formula is: For unit load signals, the standard deviation is typically small during steady-state operation, but increases significantly during variable load periods due to continuous changes in load commands. The rate of change is calculated based on the difference between adjacent sampling points. The sampling point and the first The rate of change between the sampling points can be expressed as: ,in The sampling time interval is defined as the average or maximum rate of change of all adjacent points within the window. This average or maximum rate of change serves as a statistical characteristic representing the speed of parameter change. Frequency domain transformation analysis uses the Fast Fourier Transform (FFT) method to process the unit load sequence within the window. Its core purpose is to convert the time-domain signal to the frequency domain, thereby analyzing the energy proportion of different frequency components in the signal. By calculating the proportion of energy of each frequency component in the low-frequency, mid-frequency, and high-frequency bands relative to the total signal energy, the spectral energy distribution characteristics can be obtained. For example, if the signal energy is mainly concentrated in the low-frequency band, it indicates that the unit load change is relatively gradual; if the high-frequency band energy proportion is significant, it indicates that the unit is undergoing rapid load adjustment or is experiencing strong random disturbances. All the features obtained from the above time-domain statistics and frequency domain transformation are combined into a fixed-dimensional statistical feature vector, which is then fed into a pre-constructed operating condition classification model for inference calculations, thereby obtaining a refined description of the current operating condition.

[0013] The operating condition classification model is built upon a machine learning algorithm based on a gradient boosting decision tree framework, which exhibits excellent generalization ability and interpretability in processing structured tabular data. The model training process is completed offline, and the historical operating data used must cover the unit's operating records for at least one full annual cycle to ensure the model can learn the operating characteristics under different seasonal ambient temperature conditions. Simultaneously, the training data must cover operating data across the entire load range from minimum technical output to rated output, including operating data for pure condensing operation with power generation only, operating data for combined heat and power operation with simultaneous external heat supply, operating data when burning a single design coal type, and operating data when blending multiple non-design coal types. It is particularly important to emphasize that transient operating data during unit start-up and shutdown are crucial for the model to learn the characteristics of non-steady-state operating conditions, because during start-up and shutdown, parameters such as main steam temperature, main steam pressure, and turbine metal wall temperature undergo large-scale changes, and their thermodynamic behavior and material stress state are significantly different from those during steady-state operation. By covering the rich and diverse historical operating scenarios mentioned above, the trained operating condition classification model can accurately identify various operating states during the online inference phase.

[0014] The output of the operating condition classification model contains two multi-dimensional vectors: a dynamic trajectory vector representing the operating condition and a membership vector representing the transition state. The dynamic trajectory vector is a multi-dimensional array composed of multiple dimensional components with clear physical or engineering meanings. The load rate component is defined as the ratio of the current actual load to the unit's rated load. This value continuously ranges from zero to one, intuitively reflecting the relative level of the unit's output and serving as the most fundamental indicator for measuring the unit's operational economy. The load change rate component is obtained by smoothing and filtering the first derivative of the load with respect to time. Its unit is megawatts per minute (MW / min). A positive sign indicates that the unit is in the process of increasing load, while a negative sign indicates that the unit is in the process of decreasing load. This component is a core parameter characterizing the unit's dynamic characteristics, affecting not only the rate of change in fuel input but also directly related to the thermal stress level of the turbine rotor and thick-walled boiler components, thus influencing the unit's lifespan and wear costs. The operating mode dimension component uses discrete coding to distinguish the current operating mode of the unit. For example, pure condensing operation mode means that all the exhaust steam from the turbine enters the condenser for condensation, and the unit operates solely for power generation. Extraction heating operation mode means that a portion of the steam is extracted from the intermediate stage of the turbine for external heating, while the remaining steam continues to expand and generate electricity. Back pressure heating operation mode means that the exhaust steam pressure of the turbine is higher than atmospheric pressure, and all the exhaust steam is used for heating without entering the condenser. Deep peak shaving operation mode means that the unit operates at a load level significantly lower than the conventional minimum technical output. The fuel composition dimension component is used to identify the coal composition ratio or blending ratio of the current fuel fed into the furnace. Different coal types have different coal quality characteristics such as calorific value, ash content, volatile matter, and sulfur content, which directly affect the calculation of fuel costs, carbon emission factors, and environmental taxes and fees. The time dimension component is used to identify the electricity market time period to which the current moment belongs. In the electricity spot market, a day is usually divided into peak period, high period, flat period and low period. The price level of electricity in each period is significantly different. The introduction of this time dimension component enables cost accounting to form a better correspondence with the price signals of the electricity market.

[0015] The transition state membership vector is a set of continuous values ​​ranging from zero to one, with a vector dimension equal to the total number of predefined steady-state operating condition types. Each element of this vector represents the degree of similarity or membership between the current operating state and a certain preset steady-state operating condition. Unlike traditional hard classification methods that simply categorize the current state into a specific operating condition, the membership vector allows the current state to be associated with multiple typical operating conditions simultaneously with a certain degree of membership. This design concept originates from the basic idea of ​​fuzzy set theory. In actual unit operation, operating condition switching is a continuous evolutionary process, rather than a sudden jump at a certain moment. For example, when a unit starts receiving a load increase command from a low-load steady-state operating condition of deep peak shaving and gradually transitions to a rated load steady-state operating condition, the unit's thermodynamic parameters need to undergo a period of dynamic change before stabilizing at the new load level. During this transition process, the unit's state is neither completely equivalent to the original deep peak shaving operating condition nor completely equivalent to the target rated load operating condition, but rather an intermediate state between the two. The transition state membership vector precisely characterizes this intermediate state through continuous numerical values: Let the transition state membership vector be... ,in This indicates that the current running state belongs to the first... The membership degree of a typical steady-state operating condition is defined. When the unit switches from the first steady-state operating condition to the second steady-state operating condition, the membership degree value corresponding to the first steady-state operating condition continuously decreases from one to zero, while the membership degree value corresponding to the second steady-state operating condition continuously increases from zero to one, with the sum of the two remaining at one. This continuous and smooth change in membership degree provides a high-resolution dynamic correction basis for the subsequent full cost aggregation step, allowing the weight of each cost element to be finely adjusted synchronously with the operating condition switching process, avoiding discontinuous jumps in cost accounting results at the boundary of operating condition classification.

[0016] After completing the operating condition identification, the data fusion and allocation steps are performed. This step involves real-time acquisition of multi-source heterogeneous data, which includes at least second-level or minute-level production operation data obtained from the power plant's plant-level monitoring information system, coal quantity and quality data from the fuel management system, pollutant emission concentration and carbon emission factor data from the environmental monitoring system, and monthly fixed cost data from the financial system. These data differ significantly in terms of acquisition frequency, data format, transmission protocol, and semantic definition, requiring the establishment of a unified data access adaptation layer to perform preprocessing operations such as time alignment, unit conversion, outlier removal, and missing value imputation on various types of data. For the production operation data, a sliding window aggregation method is used to aggregate high-frequency second-level data into a minute-level mean sequence that matches the calculation period, thereby compressing the data volume and smoothing random noise while retaining the main trend information.

[0017] To address the inherent time-scale discrepancy between monthly fixed cost data from the financial system and real-time operational data from the production system, this invention constructs a three-level closed-loop dynamic allocation model—prediction-allocation-backtracking—to generate real-time fixed cost allocations for each calculation period. In the fixed cost structure of power generation enterprises, monthly depreciation expenses originate from the monthly depreciation amount formed by amortizing unit construction investment over its estimated service life; this amount remains relatively fixed between months. Monthly labor costs include salaries, bonuses, and welfare expenses for operating, maintenance, and management personnel, and do not fluctuate with unit load levels in the short term. Monthly operation and maintenance costs cover expenses such as routine maintenance and troubleshooting, periodic maintenance allocation, and spare parts consumption. Monthly management expenses involve public expenditures such as power plant-level administration and logistical support. Monthly financial expenses mainly consist of interest expenses incurred from infrastructure loans or working capital loans. In traditional cost accounting, the aforementioned fixed costs are often simply averaged out based on monthly power generation. However, in the context of the electricity spot market, the value of power generation varies greatly at different times. Evenly allocating fixed costs to each kilowatt-hour would severely distort the true economic cost of electricity at different times.

[0018] In the forecasting stage, the intraday rolling load forecast curve and time-of-use (TOU) price signals are acquired. The intraday rolling load forecast curve is generated by the load forecasting system every preset update cycle, with a time resolution of minutes. The TOU price signals include peak-hour, mid-hour, flat-hour, and off-hour prices as defined by the electricity market. Based on the forecasted load values ​​for each time period in the intraday rolling load forecast curve and the TOU price signals, the time period weighting factor for each time period is determined. A typical method for determining the weighting factor is to use the product of the forecasted load and the forecasted price for that time period as the weighting basis for that time period. Its economic meaning is the proportion of the expected generation revenue realized in that time period to the total daily revenue. Allocating the total monthly fixed costs to each time period according to this proportion ensures that the allocation of fixed costs matches the revenue value created by each time period. Based on this, the total monthly fixed costs are multiplied by the proportion of the time period weighting factor of each time period to the sum of all time period weighting factors to obtain the initial allocation coefficient for each time period.

[0019] In the real-time allocation level, at the start of each calculation period, the total monthly fixed cost is multiplied by the initial allocation factor for that period to obtain the initial fixed cost allocation for that period. The initial fixed cost allocation is then divided by the number of calculation cycles within that period to obtain the real-time fixed cost allocation for each calculation cycle. Simultaneously, starting from the start of the current calculation cycle, the actual unit load is continuously acquired. The single-point deviation between the actual load and the corresponding predicted load on the daily rolling load forecast curve is calculated, and this single-point deviation is integrated and accumulated over time to obtain the cumulative deviation value from the start of the current calculation cycle to the current time. Let the actual load be... The predicted load is The cumulative deviation is .

[0020] In the retrospective correction stage, the absolute value of the accumulated deviation is compared with a preset deviation threshold. The preset deviation threshold needs to be set considering both the accuracy requirements of fixed cost allocation and the limitations of system computing resources. A typical value can be set to 1% to 3% of the total monthly fixed cost, or 5% to 10% of the initial allocation amount for the current period. When the absolute value of the accumulated deviation is less than or equal to the preset deviation threshold, the initial allocation coefficient for the current period remains unchanged, and allocation continues according to the real-time fixed cost allocation amount. When the absolute value of the accumulated deviation is greater than the preset deviation threshold, online retrospective correction is triggered. This online retrospective correction uses minimizing the total allocation deviation for the remaining periods of the current calculation cycle as the objective function, and the total fixed cost to be allocated for the remaining periods as the equality constraint. A quadratic programming optimization algorithm is used to recalculate the optimized allocation coefficients for the remaining periods. Its mathematical expression is: finding a set of adjustment amounts for the allocation coefficients of the remaining periods. , so that the objective function Minimize, while satisfying equality constraints ,in Represents the set of remaining time periods. These are the original initial allocation coefficients for each remaining time period. The ideal allocation factor is reassessed based on the latest load forecasts and actual operating information. This represents the total monthly fixed costs. This represents the remaining total fixed costs that have not yet been allocated. The optimized allocation factor replaces the corresponding original initial allocation factor and serves as the basis for fixed cost allocation in subsequent periods. This closed-loop correction mechanism allows the fixed cost allocation scheme to continuously self-correct based on actual operating conditions, dynamically optimizing the allocation amount for each period while ensuring the total monthly amount remains unchanged.

[0021] The calculation steps for variable and environmental costs are performed synchronously based on real-time data from the operating parameter stream. Real-time fuel cost calculation employs the inverse balance method, a widely used coal consumption rate calculation method in thermal power plant operation monitoring. Real-time turbine heat rate values, typically expressed in kilojoules per kilowatt-hour, are obtained from the unit's distributed control system and calculated in real-time by the thermal performance calculation module based on various factors such as unit load rate, main steam parameters, reheat steam parameters, and exhaust pressure. Real-time boiler efficiency values ​​are obtained from the boiler performance monitoring system, calculated online based on parameters such as flue gas temperature, flue gas oxygen content, and fly ash carbon content. Pipeline efficiency design values ​​are obtained from the pipeline characteristic parameter database. Real-time plant power consumption rate is obtained from the plant power monitoring system. The formula for calculating the standard coal consumption rate for power supply is: Standard coal consumption rate for power supply equals the difference between the turbine heat rate divided by the boiler efficiency, divided by the pipeline efficiency, divided by one minus the plant power consumption rate, and finally divided by the lower calorific value constant of standard coal, 29308 kilojoules per kilogram. The physical meaning of this formula lies in calculating the standard coal mass required to generate a unit of electricity by reversing the heat consumption rate at the turbine end along the energy transfer path to the fuel input end. The real-time standard coal price is obtained by the fuel management system based on the received lower heating value of the coal fed into the furnace that day and the lower heating value of standard coal. The calculation logic is as follows: divide the actual delivered unit price of the coal into its received lower heating value, and then multiply by the lower heating value of standard coal. Finally, multiplying the real-time unit load, the standard coal consumption rate for power generation, and the real-time standard coal price, and then dividing by one thousand, yields the real-time fuel cost in yuan per hour.

[0022] The calculation of purchased power cost addresses the scenario where the generating unit absorbs electricity from the external power grid. Real-time values ​​of purchased power are obtained from the meter readings at the connection point between the unit and the external power grid, along with the power purchase contract price or real-time spot market node price. Multiplying the purchased power by the price yields the purchased power cost. Real-time carbon emission cost is calculated by obtaining the unit's real-time carbon emission factor from the environmental monitoring system. This factor is adjusted in real-time based on the carbon content of the coal fed into the furnace, the unit load rate, and the boiler combustion efficiency. The real-time carbon market price is obtained, and multiplying the unit's real-time load, real-time carbon emission factor, and real-time carbon market price yields the real-time carbon emission cost. Real-time environmental tax and fee cost is calculated by obtaining real-time monitoring values ​​of sulfur dioxide, nitrogen oxide, and particulate matter emission concentrations in the flue gas from the environmental monitoring system, combined with the flue gas flow rate, to calculate the real-time emission volume of each pollutant. Dividing the real-time emission volume of each pollutant by its corresponding pollution equivalent value yields the pollution equivalent number, which is then multiplied by the unit equivalent tax / fee amount. The sum of these values ​​yields the real-time environmental tax and fee cost. The real-time fuel cost, purchased power cost, real-time carbon emission cost, and real-time environmental tax cost calculated within the same calculation period are summed to obtain the real-time change and comprehensive environmental cost for that calculation period.

[0023] The lifespan loss cost calculation step is deeply coupled with the operating condition identification module. It extracts the variable load rate component from the dynamic trajectory vector of the operating condition and calculates creep loss cost and fatigue loss cost separately. Creep refers to the phenomenon of slow plastic deformation of metallic materials over time under constant temperature and stress. For critical components such as the high-pressure rotor, intermediate-pressure rotor, boiler superheater header, and boiler reheater header, which are subjected to long-term high-temperature and high-pressure environments, creep damage is one of the main factors determining their design life. The design life parameters of each critical component are obtained. The design life hours are calculated and determined by the equipment manufacturer based on the material creep characteristics under design temperature and pressure conditions. The equivalent operating hour increment of each critical component within the current calculation period is obtained. The equivalent operating hour increment uses one hour of stable operation of the unit under rated parameters as the baseline equivalent unit. It is calculated based on the deviation of the actual main steam temperature and pressure from the rated values ​​within the current calculation period. The higher the temperature or pressure, the greater the amplification factor of the equivalent operating hour increment relative to the actual operating time. Dividing the equivalent operating hour increment by the design life hours yields the creep life loss ratio within the current calculation period. Multiplying this creep life loss ratio by the cost of a single overhaul gives the creep loss cost. The cost of a single overhaul includes not only the labor, material, and spare parts replacement costs incurred directly during the overhaul, but also the opportunity cost of the power generation lost due to downtime for maintenance.

[0024] The calculation of fatigue loss cost focuses on the cumulative damage caused to components by thermal stress cycles during variable load processes. Online cycle counting is performed on the time series of the variable load rate component, preferably using the rainflow counting method to identify complete stress cycles formed during the variable load process in real time. The rainflow counting method is a widely used two-parameter cycle counting method in the field of materials fatigue analysis. By pairing the peaks and valleys of the load time history, it can accurately identify stress cycles at various levels nested within complex load histories. For each identified complete stress cycle, the stress amplitude of that cycle is recorded. This stress amplitude is obtained by converting the peak and valley values ​​of the variable load rate through a preset mapping relationship between variable load rate and thermal stress amplitude. Based on the stress amplitude, a pre-established curve showing the correspondence between stress amplitude and allowable cycle number is consulted to obtain the allowable cycle number corresponding to that stress cycle. This curve typically follows a power function relationship; let the allowable cycle number be... The stress amplitude is The two satisfy the relationship ,in and Where is a material constant, and The value is typically between three and ten. The fatigue damage degree of this stress cycle is calculated as the reciprocal of the allowable number of cycles corresponding to that stress cycle, i.e. Multiply the fatigue damage degree by the cost of a single overhaul to obtain the fatigue loss cost per cycle corresponding to that stress cycle. Accumulate the fatigue loss costs per cycle for all stress cycles identified within the current calculation period, and then distribute the accumulated results evenly over the duration of the calculation period to obtain the fatigue loss cost for the current calculation period.

[0025] The creep loss cost and fatigue loss cost calculated within the same calculation cycle are added together to obtain the lifetime loss cost for that calculation cycle. The fatigue loss component of the lifetime loss cost exhibits a superlinear functional relationship with the absolute value of the variable load rate. This is due to the stress amplitude... The absolute value of the variable load rate There is a positive correlation between the fatigue damage degree in a single cycle and the fatigue damage degree in a single cycle. Proportional to .because When the value is greater than 1, the rate of increase in fatigue damage with the absolute value of the load change rate exceeds the linear growth rate of the load change rate itself, exhibiting an accelerated growth characteristic. This means that increasing the load change rate of the unit from four megawatts per minute to eight megawatts per minute, although the load change time is halved, will result in a much greater increase in fatigue life loss costs than just doubling, but potentially reaching three to ten times.

[0026] When the unit operates in combined heat and power (CHP) mode, a heat-power allocation process is executed. The extraction steam flow rate, extraction steam pressure, and extraction steam temperature of the heating extraction steam header are obtained in real time from the unit's distributed control system. Simultaneously, the fresh steam pressure, fresh steam temperature, reheat steam pressure, reheat steam temperature, exhaust steam pressure, and actual power generation of the unit under the current operating conditions are also obtained. The design efficiency values ​​and rated power generation of each stage of the unit under pure condensing conditions are retrieved from the thermodynamic characteristic database. The equivalent enthalpy drop method is used for heat-electricity cost allocation: First, the overall enthalpy drop of the turbine before steam extraction for heating is calculated, which is the total usable heat energy released by a unit mass of new steam expanding from the turbine inlet to the exhaust outlet. Second, the residual enthalpy drop of the turbine after steam extraction for heating is calculated, which is the enthalpy drop corresponding to the continued expansion and work done by the remaining steam in the turbine after deducting the enthalpy carried by the steam extracted for heating. Then, the difference between the overall enthalpy drop and the residual enthalpy drop is calculated to obtain the enthalpy drop lost per unit mass of extracted steam due to steam extraction for heating. Multiplying the enthalpy drop lost per unit mass of extracted steam by the steam extraction flow rate yields the power generation loss caused by steam extraction for heating. Adding the power generation loss to the actual power generation of the unit yields the equivalent power generation under pure condensing conditions. The heating cost allocation ratio is defined as the power generation loss divided by the equivalent power generation under pure condensing conditions. The power generation cost allocation ratio is obtained by subtracting the heating cost allocation ratio from one. The various cost elements, including fuel cost, purchased power cost, carbon emission cost, environmental tax cost, creep loss cost, and fatigue loss cost, calculated in real time, are multiplied by the heating cost allocation ratio to obtain the share of each cost to be borne by the heating side; these are then multiplied by the power generation cost allocation ratio to obtain the share of each cost to be borne by the power generation side. The various costs borne by the power generation side are then summed to obtain the variable costs, environmental costs, and lifetime loss costs borne by the power generation side.

[0027] The full cost aggregation step obtains the real-time allocation of fixed costs, real-time changing and environmentally integrated costs, lifetime depreciation costs, and generation-side costs after thermal-electrical allocation generated in the preceding steps. Based on the current operating condition dynamic trajectory vector and transition state membership vector, a pre-built operating condition-weight mapping function library is queried to determine the dynamic weighting function corresponding to each cost element. The construction of this function library combines offline simulation analysis and historical operating data regression analysis: offline simulation analysis utilizes a high-precision thermal system simulation model to perform systematic simulation calculations for different load rates, load change rates, operating modes, and fuel combinations, analyzing the sensitivity of each cost element to changes in total cost; historical operating data regression analysis utilizes massive amounts of data accumulated from long-term unit operation, employing multiple regression or machine learning methods to uncover the inherent statistical regularities between the weights of each cost element and operating condition parameters. Through function fitting methods, the weight values ​​at discrete operating point points are extended to the operating condition dynamic trajectory vector. and transition state membership vector Let be a continuous function of the independent variable. The core formula for total cost aggregation can be expressed as:

[0028] ;

[0029] in For the first The original values ​​of each cost element This is the value of its corresponding dynamic weighting function under the current operating conditions.

[0030] For the real-time allocation of fixed costs, its dynamic weighting function uses the membership values ​​of the load factor component and the corresponding deep peak-shaving conditions in the transition state membership vector as independent variables. When the load factor decreases or the membership value of the deep peak-shaving condition increases, the value of the dynamic weighting function for fixed costs increases accordingly, reflecting the economic principle that a larger share of the unit's fixed costs should be borne by the unit's electricity consumption during low-load operation. For lifetime loss costs, its dynamic weighting function uses the absolute value of the variable load rate component as the main independent variable. The typical form of this function is a piecewise function: when the absolute value of the variable load rate component is less than or equal to a preset first rate threshold, the dynamic weighted function value of the life loss cost remains at the baseline value one; when the absolute value of the variable load rate component is greater than the preset first rate threshold, the dynamic weighted function value of the life loss cost increases superlinearly with the increase of the absolute value of the variable load rate component, and the slope of the tangent line of the growth curve gradually increases with the increase of the absolute value of the variable load rate component; when the absolute value of the variable load rate component reaches a preset second rate threshold, the dynamic weighted function value of the life loss cost reaches a preset maximum weighting coefficient, and thereafter maintains this maximum weighting coefficient unchanged. Its mathematical form can be expressed as:

[0031] ;

[0032] in , This is the scaling factor. and These are the first and second rate thresholds, respectively. For real-time variable and environmental comprehensive costs, the dynamic weighting function uses the membership values ​​of the load rate component and the corresponding rated load condition in the transition state membership vector as independent variables. When the load rate approaches the rated value, the weighting function value approaches one; when the load rate deviates from the rated value, it is adjusted accordingly according to the preset correction curve. The values ​​of each cost element are multiplied by their corresponding dynamic weighting function values ​​to obtain the weighted costs for each element. The weighted costs for each element are then summed to generate the unit-level real-time total cost at the current calculation moment.

[0033] Finally, the cost output step is executed. At the end of each calculation cycle, the unit-level real-time total cost is divided by the average power generation of the unit within the current calculation cycle to obtain the real-time total cost per unit of electricity, expressed in yuan per kilowatt-hour. The preset cycle is a time interval synchronized with the electricity spot market trading cycle: for the day-ahead spot market, the preset cycle is fifteen minutes, corresponding to the 96-point daily trading session; for the real-time spot market, the preset cycle is five minutes, corresponding to the real-time market clearing cycle. The real-time total cost per unit of electricity and the unit-level real-time total cost calculated within each preset cycle are pushed to the spot market pricing decision system in real time via an application programming interface. The pushed data also includes a detailed breakdown of costs within the preset cycle, including the values ​​and percentages of fixed cost allocation, fuel cost, purchased power cost, carbon emission cost, environmental tax cost, creep loss cost, and fatigue loss cost. The spot market pricing decision system receives the real-time full cost per unit of electricity and uses it as a reference value for the lower limit of the price. When the generating unit is in an adjustable state within the current preset period and plans to participate in the bidding for the next trading session, the pricing decision system generates a suggested price based on the real-time full cost per unit of electricity, plus a preset target profit margin. Simultaneously, the cost data is output to the operation optimization system. The operation optimization system compares the real-time full cost per unit of electricity for the current preset period with adjacent historical periods, calculating the cost change rate and the contribution of each cost component. When the real-time full cost per unit of electricity exceeds a preset cost alarm threshold, a cost anomaly alarm signal is generated, and key operating parameters and cost composition details for the corresponding period are retrieved and pushed to the operator's terminal. The operation optimization system also compares the current operating condition dynamic trajectory vector and the corresponding real-time full cost per unit of electricity with the pre-established operating condition-cost optimal mapping table. When the current cost is higher than the historical optimal cost level under the same or similar operating conditions in the mapping table and the excess exceeds the preset optimization trigger threshold, the system pushes operation optimization suggestions to the operator's terminal. The operation optimization suggestions include the target value of the main steam parameter to be adjusted, the operating combination of the plant power equipment to be adjusted, or the coal blending ratio to be adjusted.

[0034] Corresponding to the above method, this invention also provides a real-time full-cost accounting system for power plants adapted to multiple operating conditions. This system comprises an operating condition identification module, a data fusion and allocation module, a variable and environmental cost calculation module, a lifespan loss cost calculation module, a thermal-electrical allocation module, a full-cost aggregation module, and a cost output module. The operating condition identification module is responsible for acquiring the unit's operating parameter stream in real time, extracting statistical features through a sliding time window, and calling a pre-trained operating condition classification model to output the operating condition dynamic trajectory vector and transition state membership vector. The data fusion and allocation module is responsible for collecting multi-source heterogeneous data and performing time alignment and aggregation, and uses a built-in prediction-allocation-backtracking three-level closed-loop dynamic allocation model to generate real-time fixed cost allocation. The variable and environmental cost calculation module calculates real-time fuel costs, purchased power costs, real-time carbon emission costs, and real-time environmental tax costs based on the inverse balance method. The lifespan loss cost calculation module extracts the variable load rate component from the operating condition dynamic trajectory vector and calculates creep loss costs and fatigue loss costs respectively. When the unit operates in combined heat and power (CHP) mode, the heat-power allocation module uses the equivalent enthalpy drop method to allocate various costs between the power generation and heating sides. The full cost aggregation module acquires various cost elements and real-time operating condition information, queries the built-in operating condition-weighting mapping function library to determine the dynamic weighting function, and performs weighted summation on each cost element to generate the unit-level real-time full cost. The cost output module converts the unit-level real-time full cost into the real-time full cost per unit of electricity and outputs it to the downstream system at a preset period synchronized with the market.

[0035] The full cost aggregation module further includes a cost element acquisition unit, an operating condition information acquisition unit, an operating condition-weight mapping function library, and a weighted summation unit. The cost element acquisition unit acquires cost element values ​​from various modules and verifies their validity within each calculation cycle. The operating condition information acquisition unit obtains the operating condition dynamic trajectory vector and transition state membership vector corresponding to the current calculation cycle from the operating condition identification module, and extracts key features such as the variable load rate component and load factor component. The operating condition-weight mapping function library stores dynamic weighting functions corresponding to each cost element. These functions are pre-built through offline simulation and historical data analysis, and continuous weighting functions with load factor, absolute value of variable load rate, and membership degree as independent variables are set for fixed costs, lifetime loss costs, and variable environmental costs, respectively. The weighted summation unit multiplies the value of each cost element by the function value of its corresponding dynamic weighting function within the current calculation cycle to obtain the weighted cost of each element, and then sums them to generate the unit-level real-time full cost for the current calculation cycle. The system's calculation cycle is shorter than or equal to the clearing cycle of the electricity spot market, ensuring that the refresh frequency of full-cost data meets the timeliness requirements of market quotations. Through the collaborative work of the above modules, the system of this invention can realize a complete calculation process from multi-source cost data aggregation, operating condition information parsing, dynamic weight determination to final weighted summation, providing high-precision real-time full-cost data support for power generation companies to participate in electricity spot market competition.

[0036] The beneficial effects of this invention are as follows:

[0037] The accuracy of cost calculation is significantly improved. By introducing dynamic trajectory vectors for operating conditions and membership vectors for transitional states, this invention can continuously and smoothly depict the cost evolution process of generating units under unsteady operating conditions such as rapid load changes, deep peak shaving, and start-up and shutdown, avoiding the step distortion of cost curves at the boundary of operating condition switching caused by traditional methods. In actual operation data verification, for periods with large load change rates or drastic changes in operating conditions, the calculation results of this invention differ from those of existing technologies by as much as 15% to 44%, significantly improving the authenticity and reliability of cost data.

[0038] The dynamic adaptability has been significantly enhanced. The full-cost aggregation model, built upon dynamic trajectory vectors of operating conditions and membership vectors of transition states, allows for continuous and dynamic adjustment of the weights of various cost elements based on real-time load change rates, load rates, operating modes, and transition states. In particular, the weight of life-cycle loss costs exhibits a superlinear growth with the absolute value of the load change rate, accurately reflecting the nonlinear accelerated damage to equipment caused by rapid load changes. This mechanism ensures that the cost accounting results consistently maintain a high degree of consistency with the specific operating stress path and economic characteristics of the unit at its current stage.

[0039] This effectively resolves the fundamental contradiction of time scale mismatch in multi-source data. The proposed three-level closed-loop dynamic allocation model of prediction-allocation-backtracking monitors the cumulative deviation between actual and predicted loads online and triggers quadratic programming-based optimization correction when the deviation exceeds limits, achieving dynamic optimization allocation of fixed costs across different time periods within the day. This mechanism ensures the matching degree between the fixed costs allocated in each time period and the actual power generation value of the units, providing a reliable cost basis for accurate pricing.

[0040] It achieves online, real-time monetization of lifespan loss costs. By deeply coupling lifespan loss calculation with operating condition identification, it can identify complete stress cycles online based on real-time load change rate sequences and quantify creep damage and fatigue damage into monetary costs in real time, incorporating them into the marginal cost of the current period. This allows operators and pricing decision-making systems to intuitively perceive the economic cost of each rapid load change operation, effectively preventing the blind pursuit of short-term electricity gains at the expense of long-term equipment lifespan loss.

[0041] The invention provides significant support for spot market pricing decisions. By using the calculation results of this invention as a reference value for the lower limit of pricing, compared with the use of existing technology cost data, approximately 8% of "invalid pricing" situations can be avoided in day-ahead spot market pricing simulations, effectively improving the market competitiveness and overall profitability of power generation companies.

[0042] The system architecture is reasonable and highly practical. This invention adopts a modular design with clear boundaries and standardized interfaces for each functional module, facilitating integration and deployment within the existing information systems of power generation enterprises. It also supports edge-cloud collaborative computing, meeting the spot market's requirements for high timeliness and reliability of cost data, and possesses significant potential for widespread adoption. Attached Figure Description

[0043] Figure 1 : Flowchart of the method of the present invention. Detailed Implementation

[0044] The present invention will be further described below with reference to the accompanying drawings and specific embodiments:

[0045] Example: Figure 1As shown, the core of the technical solution provided by this invention lies in constructing a full-cost online calculation system capable of dynamically adapting to various operating conditions of thermal power units and aggregating fixed costs, variable costs, environmental costs, and lifespan depreciation costs in real time. This system is suitable for large thermal power generating units participating in the electricity spot market and can be deployed on the edge computing nodes of the power plant's plant-level monitoring information system or on independent real-time data servers. The overall data flow and processing logic of the system are as follows: First, multi-source heterogeneous raw operating data and business data are acquired from data sources such as the unit's distributed control system, plant-level monitoring information system, fuel management system, environmental monitoring system, and financial system at different acquisition frequencies. After these data enter the system, they are first processed by the operating condition identification module. This module extracts statistical features through a sliding time window and uses a pre-trained offline operating condition classification model to output the current operating condition dynamic trajectory vector and transition state membership vector in real time. Meanwhile, the data fusion and allocation module receives data from different systems, aggregates high-frequency production data over time, and uses a three-level closed-loop dynamic allocation model of prediction-allocation-backtracking to dynamically allocate the total monthly fixed cost to each calculation cycle, generating a real-time fixed cost allocation. The variable and environmental cost calculation module, based on real-time operating parameters, uses an inverse balance method to calculate real-time fuel costs, purchased power costs, real-time carbon emission costs, and real-time environmental tax costs, and then sums these four costs to obtain the real-time variable and environmental comprehensive cost. The lifespan loss cost calculation module extracts the variable load rate component from the dynamic trajectory vector of the operating conditions and calculates the monetized cost of lifespan loss caused by high-temperature creep and thermal stress fatigue. When the unit operates in cogeneration mode, the heat-power allocation module is triggered, using the equivalent enthalpy drop method to calculate the impact of heating on power generation and rationally allocates various costs between the power generation and heating sides. Finally, the full cost aggregation module gathers all cost elements from the aforementioned modules, combines the current operating condition dynamic trajectory vector and transition state membership vector, queries the built-in operating condition-weight mapping function library to determine the dynamic weighting coefficient of each cost element under the current operating condition, and performs a weighted summation of each cost element to generate the unit-level real-time full cost. The cost output module divides this full cost by the current generating power to obtain the real-time full cost per unit of electricity, and pushes the cost data and its component details to the downstream pricing decision system and operation optimization system at a frequency synchronized with the electricity spot market trading clearing cycle. The above steps constitute a complete closed loop from data acquisition, feature extraction, operating condition identification, cost item calculation, cost allocation to dynamic aggregation and output.

[0046] Regarding specific hardware deployment and software architecture, this invention recommends a three-layer architecture of "end-edge-cloud" collaboration to achieve efficient and reliable data processing. At the device end, through interface machines deployed within the unit's distributed control system network or via the standard OPC data access protocol, real-time data such as unit load, main steam temperature, main steam pressure, reheat steam temperature, reheat steam pressure, feedwater flow rate, plant power consumption, fuel input, and heating extraction steam flow rate and parameters are read in seconds or sub-seconds. At the edge computing layer, one or more high-performance edge computing servers are typically set up in the power plant's local information center or control room, deploying the core computing engine of this invention, including a condition identification module, various cost calculation modules, and a full cost aggregation module. The edge computing layer is responsible for completing all real-time computing tasks to ensure low latency and high reliability in data processing, avoiding cost accounting interruptions due to wide area network communication outages or cloud service failures. At the cloud or power plant-level application server layer, deploy predictive and retrospective correction-level optimization computing services, historical data storage and analysis services, and visualization and reporting systems for operations and management decision-makers within the data fusion and allocation module. This architecture allows the most real-time-critical operating condition identification and second-level cost calculation to be completed locally at the edge, while the computationally intensive but less real-time-critical rolling prediction and quadratic planning optimization correction are executed asynchronously on servers with stronger computing power. This ensures both system responsiveness and accommodates the computational needs of complex algorithms.

[0047] The specific implementation process of the operating condition identification module first requires offline training of the operating condition classification model based on the unit's long-term historical operating data. Constructing the training data is a crucial step in ensuring the model's accuracy. Historical operating records for at least one full calendar year should be selected to cover the impact of ambient temperature changes during the four seasons (spring, summer, autumn, and winter) on the unit's vacuum and efficiency. The data must include continuous operating data across the entire load range from minimum technical output to rated output, as well as typical data for the unit in pure condensing operation mode, extraction steam heating operation mode, and back-pressure heating operation mode. Particularly important is that the training set must include a sufficient amount of data on unit start-up and shutdown processes and operating data for deep peak shaving to below 30% of rated load, because the statistical characteristics of operating parameters under these transient or extreme conditions are fundamentally different from those under steady-state conditions. During the data cleaning phase, outliers and dead values ​​caused by sensor failures or communication interruptions should be removed, and missing data should be appropriately imputed. For each sliding time window, the window width can be set to 30 seconds, and the sliding step size to 5 seconds. Within each window, the arithmetic mean, standard deviation, and average rate of change based on the difference between adjacent sampling points are calculated for key parameters such as unit load, main steam temperature, main steam pressure, reheat steam temperature, reheat steam pressure, and feedwater flow rate. Specifically, let the window contain... The time series composed of sampling points is If the sampling frequency is 1Hz, then the 30-second window contains One sampling point. Arithmetic mean. The calculation formula is Standard deviation The calculation formula is The rate of change is calculated based on the first-order difference between adjacent sampling points. For the ... The sampling point and the first The rate of change between the sampling points can be expressed as: ,in Seconds, average rate of change within the window Simultaneously, a Fast Fourier Transform (FFT) is performed on the unit load sequence within the window, extracting the proportion of energy in three frequency bands—0Hz to 0.01Hz, 0.01Hz to 0.1Hz, and 0.1Hz to 1Hz—as the spectral energy distribution characteristics. Let the square of the modulus of each frequency component obtained after the FFT be the power spectral density. then frequency band The energy percentage inside is All the above features are concatenated into a fixed-length feature vector, which is then used as the input to the model.

[0048] The model is trained using the XGBoost algorithm based on a gradient boosting decision tree framework. The training label design is unique: in addition to encoding the dominant operating condition category for each sample, fuzzy C-means clustering is used to analyze the membership degree of each sample, generating a membership value vector for each typical steady-state operating condition. This vector serves as one of the objectives of the model's multi-output regression. The objective function of fuzzy C-means clustering is...

[0049] ,in For the first The feature vector of each sample For the first Cluster centers, For the sample Cluster centers membership degree The fuzzy factor is typically set to 2. The membership value of each sample to each typical working condition is obtained through iterative optimization and used as one of the objectives of supervised learning. The trained model is exported as a model file and deployed to the online inference engine. During online runtime, the working condition recognition module reads the latest operating parameters once per second, constructs a sliding window and extracts features, invokes model inference, and outputs the dynamic trajectory vector of the working condition and the membership vector of the transition state in real time.

[0050] The implementation of the data fusion and allocation module involves the access and time alignment of multi-source data. For the second-level production operation data obtained from the plant-level monitoring information system, the module has a built-in data aggregator that uses a sliding window with a width of 1 minute and a step size of 1 minute to calculate the average value of each parameter within the window, generating a minute-level average sequence for alignment with the minute-level cost calculation cycle. For the coal quantity and quality data from the fuel management system, since coal quality testing usually has a certain time lag, this embodiment adopts a rolling update method. That is, whenever a new coal testing report is generated, the real-time coal quality parameters in the system are updated until the next testing report arrives. Let the received lower heating value of the coal be... The unit is kilojoules per kilogram, and the lower heating value of standard coal is... The price of coal delivered to the furnace per kilogram, including tax, is [price missing]. The unit is yuan per ton, which represents the real-time standard coal price. The calculation formula is: The unit is yuan per ton of standard coal. For carbon emission factor data, in addition to the baseline value of carbon content in the coal fed into the boiler, real-time corrections are required based on boiler load rate and flue gas oxygen content. Correction coefficients are obtained through table lookup or regression formulas, which can be pre-determined through performance tests of the boiler under varying load conditions. Assume the received baseline carbon content of the coal fed into the boiler is... Boiler load rate The oxygen content of the flue gas is Real-time carbon emission factor It can be represented as ,in This is the incomplete combustion correction factor determined based on the load rate and oxygen content.

[0051] Fixed cost allocation is the core function of this module. The following example of monthly fixed cost allocation will illustrate the workflow of the three-level closed-loop dynamic allocation model: prediction-allocation-backtracking. Assume the total monthly fixed cost of a 600MW supercritical unit is... The fixed cost is 24 million yuan. Since the month has 30 days, the benchmark value for the total fixed cost allocated daily is... Ten thousand yuan. At 0:00 each day, the forecasting level obtains the intraday rolling load forecast curve for the next 24 hours from the load forecasting system, with a time resolution of 15 minutes and a total of 96 time periods. Let the first... The predicted load for each time period is The predicted electricity price is Then the time period weighting factor for that time period Defined as .

[0052] The sum of the weighting factors for all time periods on that day is Then the first Initial allocation coefficient for each time period for For example, if the predicted load for a peak period on a certain day (let's say period 37 to 40) is 550MW and the predicted electricity price is 0.8 yuan per kilowatt-hour, then its period weighting factor is: During off-peak hours (assumed to be periods 1 to 8), the predicted load is 200MW and the predicted electricity price is 0.2 yuan per kilowatt-hour. The period weighting factor is: Clearly, the cost-sharing factor during peak hours is much larger than that during off-peak hours, reflecting the principle of "beneficiaries bearing the costs."

[0053] At the real-time allocation level, each calculation cycle is 1 minute. At the start of each 15-minute period, the initial allocation coefficient corresponding to that period is multiplied by the total fixed cost of the day, 800,000 yuan, to obtain the initial fixed cost allocation for that 15-minute period. This value is then divided by 15 to obtain the real-time fixed cost allocation for each 1-minute calculation cycle. Let the... Within the first 15-minute period The fixed cost per minute is amortized in real time. At the same time, the system continuously acquires the actual load of the generating units. And compare it with the value at the corresponding moment on the predicted load curve. Subtracting the individual points gives the single-point deviation, and integrating the single-point deviation gives the cumulative deviation. In digital systems, integration is achieved through numerical accumulation: ,in The calculation period is 1 minute.

[0054] When the absolute value of the accumulated deviation exceeds a preset threshold, a backtracking correction is triggered. The preset threshold can be set to 10% of the initial allocated amount for the current period, i.e., if the accumulated allocated fixed cost for the current period is... The threshold is The correction stage invokes a quadratic programming solver to minimize the total allocation deviation for the remaining time slots of the day. Under the constraint that the total allocation amount for the remaining time slots equals the total remaining fixed costs for the day, it re-optimizes the allocation coefficients for each remaining 15-minute time slot. The standard form of the quadratic programming problem is: Find... make Minimum, and satisfying and In this problem, the optimization variables are... The apportionment coefficient for the remaining time periods The objective function is The equality constraint is Since quadratic programming solutions converge quickly, calculations can typically be completed within one second, and the optimized allocation coefficients are updated for use at the real-time allocation level.

[0055] The implementation of the variable and environmental cost calculation module relies on real-time data interfaces with the unit's thermal performance monitoring system, plant power monitoring system, and continuous emission monitoring system. Calculating real-time fuel costs first requires obtaining the standard coal consumption rate for power supply. Within a specific 1-minute calculation cycle, the real-time value of the turbine heat rate obtained from the distributed control system. The real-time boiler efficiency is 7850 kJ per kilowatt-hour. The pipeline efficiency is 93.5%. Take 99% of the design value, and the real-time value of the plant power consumption rate. It is 4.8%. The formula for calculating the standard coal consumption rate for power supply is: ;

[0056] Substitute the numerical values ​​into the calculation:

[0057] Assuming the fuel management system provides the daily lower heating value of the coal fed into the furnace... The price is 20,500 kJ per kilogram, including tax, delivered to the factory. The price is 850 yuan per ton, which translates to the real-time standard coal price. for:

[0058] If the average load of the unit in the current calculation period For a capacity of 450MW, the real-time fuel cost within that minute is... for:

[0059] Note that the division by 60 here is because the load unit is MW (i.e., kilowatts), while the cost is usually calculated in yuan per hour and then converted to minutes. Here, the cost per minute is calculated directly.

[0060] Regarding the cost of purchased power, the purchased power is obtained through the metering system. If the current meter reading shows 0 purchased electricity, then this cost is 0. The calculation of real-time carbon emission costs is based on the real-time carbon emission factor output by the environmental monitoring system. The figure is 0.82 tons per megawatt-hour, assuming the current carbon market price. The cost is 65 yuan per ton, so the carbon emission cost per minute is... for: Environmental tax costs are calculated based on the real-time emission rates of sulfur dioxide, nitrogen oxides, and particulate matter, along with their corresponding pollution equivalent values ​​and tax rates. Assume the total environmental tax cost for these three pollutants is 25 yuan per minute. Then, the real-time changes in the current period and the overall environmental cost... for:

[0061] ;

[0062] The creep loss cost calculation module and the operating condition identification module are tightly coupled. The creep loss cost calculation takes a specific key component—the high-pressure rotor of a steam turbine—as an example. The rotor's design life... Cost of a single major overhaul (including downtime losses) is 250,000 hours. The estimated cost is 80 million yuan. The average main steam temperature during the current calculation period... The average temperature is 566℃, and the average main steam pressure is... It is 24.2 MPa, compared to the rated parameters. , The temperature was 6°C higher than normal. Based on the creep characteristics of materials, the accelerating effect of temperature on the creep rate can be described by the Larssen-Miller parameter. Larssen-Miller parameter Defined as ,in The absolute temperature is K. The fracture time (h) is the time between fractures. This is a material constant (usually taken as 20). Under the same Larssen-Miller parameters, higher temperatures correspond to shorter fracture times. Assuming that the ratio of equivalent operating time to calendar time is 1:1 under rated parameters, the amplification factor of the equivalent operating time increases by 6°C (i.e., from 833K to 839K).

[0063] It can be estimated by the following formula:

[0064] ;

[0065] Substituting the numerical values, we get This means that the actual running time of 1 minute is equivalent to 1.06 minutes of operation under rated parameters. Therefore, the creep life loss ratio for that 1 minute is... for:

[0066] Creep loss cost during this period for:

[0067] ;

[0068] Calculating fatigue loss costs requires online rainflow counting of the variable load rate component sequence. Let the variable load rate component be... The unit is measured in MW / min. Over the past 30 minutes, the unit experienced a load increase from 300MW to 500MW, with an average load change rate of... The flow rate was initially set at MW / min, then stabilized, followed by a load reduction. Rainflow counting is used to measure... The peak and valley values ​​are paired to identify a complete stress cycle. The stress amplitude of the cycle... The relationship with the variable load rate is determined by thermal stress analysis; under a simplified model, it can be assumed that... Assume the allowable number of cycles corresponds to the stress amplitude of this cycle after conversion. The curve shows 12,000 cycles. This indicates the fatigue damage caused by this cycle. for:

[0069] ;

[0070] Fatigue loss cost per cycle for:

[0071] ;

[0072] The cycle lasts 45 minutes from the start of load increase to the end of load decrease, therefore the fatigue wear cost is averaged per minute. for:

[0073] ;

[0074] If no new stress cycles are detected during stable operation, the fatigue loss cost is 0. This represents the life loss cost over the current period. The total cost is the sum of creep loss cost of 5.66 yuan and fatigue loss cost of 148 yuan, i.e.:

[0075] ;

[0076] It is evident that fatigue loss costs account for the vast majority of lifespan loss costs during rapid load changes.

[0077] The superlinear relationship between fatigue loss and variable load rate can be rigorously derived using the material fatigue characteristic formula. Let the allowable number of cycles be... With stress amplitude satisfy ,in The fatigue index of the material, typically ranging from 3 to 10. Stress amplitude. absolute value of variable load rate Positive correlation can be written as ,in If it is close to 1, then the fatigue damage degree per cycle is... It can be represented as:

[0078] ;

[0079] because Therefore, the degree of damage Follow The growth of exhibits a superlinear acceleration. For example, if , If the variable load rate increases by 100%, the fatigue damage will increase. This quantitative relationship is fully reflected in the lifetime loss cost calculation of this invention, and is also the physical basis for the superlinear growth of the lifetime loss cost weighting function in the total cost aggregation.

[0080] The heat-power sharing mechanism is activated when the unit supplies heat to the outside. Taking a single extraction steam heating unit as an example, the current actual power generation... For 350MW, the steam extraction flow rate for heating is... The extraction steam pressure is 280 tons per hour (77.78 kg / s), the extraction pressure is 0.4 MPa, and the extraction temperature is 260℃. The turbine's efficiency design values ​​for each stage are retrieved from the thermodynamic characteristics database. First, the overall enthalpy drop of the fresh steam from the main steam valve to the condenser under pure condensing conditions is calculated. Assume the new steam enthalpy. The exhaust enthalpy is 3420 kJ / kg. If it is 2170 kJ / kg, then kJ / kg. Enthalpy of steam at the extraction port after extraction. The enthalpy drop is 2980 kJ / kg, and the extracted steam no longer participates in subsequent expansion. Calculate the residual enthalpy drop of the turbine after steam extraction for heating. The impact of steam extraction on the flow rates of each stage needs to be considered. In simplified calculations, steam extraction can be considered as directly removing a portion of energy from the thermodynamic cycle. The enthalpy drop corresponding to the remaining steam continuing to expand and do work in the turbine is calculated to be 1080 kJ / kg based on heat balance. Therefore, the enthalpy drop lost per unit mass of steam extracted due to steam extraction is... for:

[0081] ;

[0082] Multiply by extraction steam mass flow rate kg / s, to obtain the power loss in electricity generation due to steam extraction. for:

[0083] ;

[0084] Equivalent power generation under pure condensing conditions for:

[0085] ;

[0086] Heating cost sharing ratio for:

[0087] ;

[0088] Electricity generation cost sharing ratio Then it is: ;

[0089] By multiplying each of the aforementioned calculated costs by these two ratios, cost separation between the power generation and heating sides can be achieved. For example, the variable and environmental comprehensive costs to be borne by the power generation side are... Yuan / min, lifespan loss cost is Yuan / min.

[0090] The implementation of the full-cost aggregation module is one of the key aspects of this invention, and its core lies in the establishment and online invocation of the operating condition-weight mapping function library. The function library establishment process is as follows: In the offline stage, a full-range high-precision thermal simulation model of the unit is used to simulate different steady-state operating points within the range of 30% to 100% load, as well as the dynamic process of changing load from 1% rated load per minute to 5% rated load per minute at different load change rates. At each simulated operating point, the values ​​of various cost elements are recorded, and the marginal contribution rate of each cost element to the change in total cost is determined through sensitivity analysis, which serves as a reasonable weight reference value for each cost at that operating point. Subsequently, combined with the actual cost data and operating condition data of the unit in the past year, a multivariate adaptive regression spline function is used for fitting to generate a continuous and smooth dynamic weighting function.

[0091] For the real-time allocation of fixed costs, its weighting function By load factor and deep peaking membership As the independent variable, the fitted function can be in the form of, for example: This function indicates that when the load factor... When it is 100%, if If the load factor is 1.0, the weight is 1.0; when the load factor drops to 40%, the weight increases to 1.0. If the membership degree of deep peak shaving is 1 at the same time, the weight will further increase by 0.8 to 4.05. This accurately reflects the economic reality of a sharp increase in the fixed cost per unit of electricity under low load conditions.

[0092] For lifetime attrition cost, its dynamic weighting function absolute value of variable load rate The first rate threshold is the primary independent variable, following a piecewise superlinear model. Set to 3MW / min, second rate threshold Set to 10 MW / min, scaling factor The value is 0.02, the exponent. The value is 1.8, the maximum weighting coefficient. Set to 5.0. The specific function expression is:

[0093] ;

[0094] when At MW / min, ;when At MW / min, This demonstrates superlinear acceleration characteristics. For real-time changes and comprehensive environmental costs, its weighting function... By load factor Membership of rated operating conditions As the independent variable, a quadratic function of the following form can be used:

[0095] ;in This is the optimal load factor for the unit (usually 100%). and These are the fitting coefficients. For example, take... , Then when and hour, This reflects that variable costs should be given higher weight due to decreased efficiency under low load conditions.

[0096] When running online, the weighted summation unit performs calculations every 1 minute. Let the original values ​​of each cost element in the current period be: real-time allocation of fixed costs. Yuan / min, changes in power generation and comprehensive environmental costs Yuan / min, generation-side lifespan loss cost Yuan / min. The dynamically weighted function values ​​calculated based on the current operating conditions are as follows:

[0097] (Due to low load factor) (Due to deviation from rated operating conditions). (Since the load rate is 8 MW / min). The weighted costs are as follows:

[0098] Yuan / min Yuan / min Yuan / min. This is the unit-level real-time full cost per minute. for:

[0099] Dividing by the average power generation per minute of 450MW, the real-time total cost per unit of electricity is:

[0100] (Note that multiplying by 60 converts the cost per minute to the cost per hour, and then dividing by the power).

[0101] The implementation of the cost output step requires seamless integration with the electricity spot market trading system. In the day-ahead market, the trading target is the electricity volume in each 15-minute timeframe. Therefore, at the end of each 15-minute timeframe, the cost output module summarizes the sum of the unit-level real-time total costs calculated over 15 one-minute cycles within that timeframe, and divides it by the total power generation within that timeframe to obtain the average real-time total cost per unit of electricity for that 15-minute timeframe. Simultaneously, it compiles the average values ​​and percentages of various cost elements within that timeframe. The data is pushed to the spot market pricing decision system in JSON format via a RESTful API interface based on the HTTP protocol. The pushed data structure includes a timestamp, unit number, unit-level total cost, unit-level total cost, and a sub-object detailing cost components, listing the values ​​and percentages of fixed costs, fuel costs, purchased power costs, carbon emission costs, environmental tax costs, creep loss costs, and fatigue loss costs. After parsing this data, the pricing decision system uses it as the marginal cost benchmark for the pricing strategy for the remaining timeframe of the day or the same timeframe of the following day. In the real-time market, the transaction clearing cycle is 5 minutes, and the cost output module performs similar processing and push notifications every 5 minutes. After receiving cost data, the operation optimization system stores it in the time-series database and performs a month-on-month analysis with adjacent historical periods. For example, if it finds that the total cost per unit of electricity in the current period has increased by 12% compared to the previous comparable period, the system automatically calculates the contribution of each cost change, finding that the main contribution comes from an 85% increase in fatigue loss costs, further tracing back to several rapid load increases and decreases experienced by the unit during this period. Based on this, the operation optimization system displays optimization suggestions on the operator's terminal, suggesting that the load change rate can be appropriately reduced to control lifespan loss, or that a higher load change rate can be accepted only during periods with sufficiently high electricity prices.

[0102] To further verify the technical effect of the present invention, the method of the present invention (Method A) is compared with two existing technical methods. Method B is an improved cost method based on daily accounting and post-operational condition correction, while Method C is the traditional monthly average cost method. Actual operating data for a 600MW supercritical unit within one day was selected. The daily operating curve includes deep peak shaving to 180MW in the early morning, rapid ramp-up to 420MW during the morning peak, full load operation of 600MW at noon, and the nighttime shutdown process. The comparison results show that during the deep peak shaving period from 2:00 AM to 2:15 AM, the unit electricity cost calculated by Method A is 0.452 yuan / kWh, Method B is 0.398 yuan / kWh, and Method C is 0.385 yuan / kWh. Method A is about 13.6% higher than Method B because Method A real-time incorporates the increased coal consumption for power generation under low load and the increased weight of lifespan loss due to deviation from design conditions, truly reflecting the depth of loss in power generation during the deep peak shaving period. During the rapid ramp-up period from 6:30 AM to 6:45 AM, Method A calculates a cost of 0.371 yuan / kWh, while Method B calculates 0.308 yuan / kWh, a difference of 20.5%. This is attributed to Method A capturing the excessive fuel consumption and significant fatigue life loss costs during rapid load changes. During the midday full-load period, the calculation results for both Method A and Method B are lower than those for Method C, accurately reflecting the cost advantage of full-load operation. During the shutdown period from 11:00 PM to 11:15 PM, Method A calculates a cost of 0.585 yuan / kWh, while Method B calculates only 0.405 yuan / kWh, a difference of 44.4%. This is because Method A accurately captures the significant cost of fuel and auxiliary equipment consumption during shutdowns, which result in a complete start-stop life loss cycle. Using the cost data provided by Method A in the simulation of day-ahead spot market pricing strategies, compared to using the cost data from Method B as the basis for pricing, can avoid approximately 8% of "invalid bids," i.e., bids lower than the actual marginal cost of the unit, leading to actual losses after winning the bid. This comparison fully demonstrates the significant superiority of the various technical features of this invention in terms of cost accounting accuracy and pricing decision support under dynamic operating conditions, achieved through the synergistic effect of these features.

[0103] In the system provided by this invention, each module can be implemented using a modular software design approach. For example, the working condition identification module can be encapsulated as an independent Windows service or Linux daemon, communicating with other modules through shared memory or message queues. The quadratic programming solver in the data fusion and allocation module can call open-source numerical optimization libraries such as OSQP or commercial solvers such as CPLEX. The working condition-weight mapping function library can be serialized into JSON or XML configuration files, facilitating dynamic updates during operation without requiring a system restart. The entire system supports primary-backup redundant deployment; when the primary computing node fails, the backup node can automatically take over the data interface and continue calculations, ensuring the continuity and high availability of cost accounting.

[0104] In summary, this invention organically integrates advanced machine learning operating condition identification, multi-scale data fusion closed-loop optimization, online fatigue life monetization calculation, and a weighted aggregation model based on dynamic trajectories, providing thermal power generation companies participating in the electricity spot market with a practical, high-precision, and robust real-time full-cost accounting solution.

[0105] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-condition adaptable real-time full-cost accounting method for power plants, applied to real-time cost accounting and pricing decision support for generator units in an electricity spot market environment, characterized in that... include: Operating condition identification steps: Real-time acquisition of unit operating parameter streams, including at least unit load, main steam temperature, main steam pressure, reheat steam temperature, reheat steam pressure, feedwater flow rate, plant power consumption rate, and fuel input; extraction of statistical features from the operating parameter streams based on a sliding time window, including the mean, standard deviation, rate of change, and spectral energy distribution within the window; input of the statistical features into a pre-constructed operating condition classification model, outputting the current operating condition dynamic trajectory vector and transition state membership vector; wherein, the operating condition dynamic trajectory vector is a multi-dimensional vector containing load rate dimension components, load change rate dimension components, operating mode dimension components, fuel combination dimension components, and time dimension components; the transition state membership vector is a set of continuous values ​​ranging from 0 to 1, representing the similarity between the current operating state and multiple preset steady-state operating conditions, and each component of the transition state membership vector changes continuously and smoothly over time during the operating condition switching process; Data fusion and allocation steps: Real-time collection of multi-source heterogeneous data, including at least second-level / minute-level production operation data from the power plant's plant-level monitoring information system, coal quantity and quality data from the fuel management system, pollutant emission concentration and carbon emission factor data from the environmental monitoring system, and monthly fixed cost data from the financial system; the second-level / minute-level production operation data is aggregated into a minute-level average sequence using a sliding window; the monthly fixed cost data is allocated using a three-level closed-loop dynamic allocation model of prediction-allocation-backtracking. The model generates real-time fixed cost allocation for each calculation period. The three-level closed-loop dynamic allocation model includes: a prediction level, which generates initial allocation coefficients for each period based on the intraday rolling load forecast curve and time-of-use electricity price signals; a real-time allocation level, which allocates the total monthly fixed cost to each calculation period according to the initial allocation coefficients and continuously monitors the cumulative deviation between the actual load and the predicted load; and a backtracking correction level, which, when the cumulative deviation exceeds a preset threshold, optimizes and updates the allocation coefficients for the remaining periods online, using minimizing the total allocation deviation for the remaining monthly periods as the objective function. The steps for calculating variable and environmental costs are as follows: Based on the real-time fuel input, real-time fuel calorific value, and real-time fuel price in the operating parameter stream, the real-time fuel cost is calculated using the inverse balance method; the cost of purchased power is calculated based on the plant power consumption rate; the real-time carbon emission cost is calculated based on the carbon emission factor data and the real-time carbon market price; the real-time environmental tax cost is calculated based on the pollutant emission concentration and the corresponding environmental tax standard; and the real-time fuel cost, purchased power cost, real-time carbon emission cost, and real-time environmental tax cost are summed to generate the real-time variable and comprehensive environmental cost. The steps for calculating life loss cost are as follows: Extract the variable load rate component from the dynamic trajectory vector of the operating condition; calculate the creep loss cost and fatigue loss cost based on the variable load rate component; wherein, the creep loss cost is obtained by multiplying the ratio of the equivalent operating hour increment to the design life hours by the cost of a single overhaul; the fatigue loss cost is obtained by performing online cyclic counting on the variable load rate component sequence to identify effective stress cycles, and calculating the fatigue damage degree based on the allowable number of cycles corresponding to the stress amplitude of each stress cycle, multiplying the fatigue damage degree by the cost of a single overhaul and allocating it to the duration of that stress cycle; the life loss cost is the sum of the creep loss cost and the fatigue loss cost, and the fatigue loss component related to the variable load rate in the life loss cost has a superlinear functional relationship with the absolute value of the variable load rate. The heat-power cost allocation process is as follows: When the unit is operating in cogeneration mode, the steam extraction flow rate, extraction parameters, and actual power generation of the unit are obtained; the power generation loss caused by steam extraction for heating is calculated using the equivalent enthalpy drop method; the heating cost allocation ratio is determined based on the ratio of the power generation loss to the rated power under pure condensing mode; based on the heating cost allocation ratio, the calculated costs are allocated between the power generation side and the heating side to obtain the variable costs, environmental costs, and lifespan depreciation costs that the power generation side should bear; The full cost aggregation steps are as follows: Obtain the real-time fixed cost allocation, the real-time variable and environmental comprehensive cost, the lifetime loss cost, and the allocated generation-side cost; based on the current operating condition dynamic trajectory vector and transition state membership vector, query the pre-built operating condition-weight mapping function library to determine the dynamic weighting function corresponding to each cost element; wherein, the dynamic weighting function is a continuous function with the operating condition dynamic trajectory vector and transition state membership vector as independent variables, and for lifetime loss cost, its dynamic weighting function value is positively correlated with the absolute value of the load change rate; use the dynamic weighting function to perform weighted summation on each cost element to generate the unit-level real-time full cost at the current calculation time; Cost output steps: Divide the unit-level real-time total cost by the current power generation to obtain the unit-level real-time total cost; output the unit-level real-time total cost and the unit-level real-time total cost to the spot market pricing decision system and the operation optimization system at a preset period.

2. The multi-condition adaptive real-time full-cost accounting method for power plants according to claim 1, characterized in that, The operating condition identification step specifically includes: The operating parameter stream is acquired from the distributed control system of the unit at a preset first acquisition frequency, wherein the first acquisition frequency is on the order of seconds or sub-seconds. A sliding time window with a width of a first preset duration and a sliding step size of a second preset duration is used to continuously capture time segments of the running parameter stream, wherein the first preset duration is within the range of 20 seconds to 60 seconds, the second preset duration is within the range of 3 seconds to 10 seconds, and the sliding step size is less than the window width; Time-domain statistical analysis and frequency-domain transform analysis are performed on the operating parameters within each sliding time window. The time-domain statistical analysis includes calculating the arithmetic mean, standard deviation, and rate of change of the difference between adjacent sampling points for each operating parameter within the window. The frequency-domain transform analysis includes performing a fast Fourier transform on the unit load sequence within the window and extracting the energy proportion of the main frequency components as the spectral energy distribution. All features extracted by the time-domain statistical analysis and frequency-domain transform analysis are combined into a fixed-dimensional statistical feature vector, which is then input into a pre-trained operating condition classification model. The operating condition classification model employs a machine learning algorithm based on a gradient boosting decision tree framework and is obtained through offline training using historical operating data of the unit. The historical operating data covers the unit's operating records for at least one complete annual cycle, including operating data under different seasonal ambient temperature conditions, operating data across the entire load range from minimum technical output to rated output, operating data for pure condensing and cogeneration conditions, operating data for single-coal combustion and blended coal combustion conditions, and transient operating data during unit start-up and shutdown. The operating condition classification model simultaneously outputs two multi-dimensional vectors. The first multi-dimensional vector is the dynamic trajectory vector of the operating condition, which contains at least the following five dimensions: The first dimension is the load rate dimension, determined by the ratio of the current actual load to the rated load of the unit, with a value between 0 and 1; the second dimension is the load change rate dimension, determined by the first derivative of the load with respect to time after smoothing and filtering, with the unit being megawatts per minute, and a positive sign indicating a load increase process and a negative sign indicating a load decrease process; the third dimension is the operating mode dimension, which uses discrete coding to distinguish between pure condensing operation mode, extraction steam heating operation mode, back pressure heating operation mode, and deep peak shaving operation mode; the fourth dimension is the fuel combination dimension, used to identify the coal composition ratio or blending ratio of the current fuel fed into the furnace; and the fifth dimension is the time dimension, used to identify the time period to which the current moment belongs, including peak period, high period, flat period, and low period as defined by the electricity market. The second multidimensional vector is the transition state membership vector. This vector is a set of continuous values ​​ranging from 0 to 1. The vector dimension is equal to the total number of predefined steady-state operating condition types. Each element in the vector represents the similarity between the current operating state and a corresponding preset steady-state operating condition. When the unit switches from the first steady-state operating condition to the second steady-state operating condition, the membership element value corresponding to the first steady-state operating condition decreases continuously from 1 to 0, while the membership element value corresponding to the second steady-state operating condition increases continuously from 0 to 1, and the sum of the two remains 1.

3. The multi-condition adaptive real-time full-cost accounting method for power plants according to claim 1, characterized in that, In the data fusion and allocation step, the monthly fixed cost data is processed using a three-level closed-loop dynamic allocation model (prediction-allocation-backtracking) to generate real-time fixed cost allocation amounts for each calculation period. Specifically, this includes: In the forecasting stage, an intraday rolling load forecast curve is obtained. The time resolution of the intraday rolling load forecast curve is at the minute level, and it is generated by the load forecasting system every preset update cycle. Time-of-use (TOU) electricity price signals are obtained, which include peak hour prices, mid-peak hour prices, flat hour prices, and off-peak hour prices in the electricity market. Based on the forecasted load values ​​of each time period in the intraday rolling load forecast curve and the TOU electricity price signals, the time period weighting factor for each time period is determined. The monthly fixed cost is multiplied by the proportion of the time period weighting factor of each time period to the sum of all time period weighting factors to obtain the initial allocation coefficient for each time period. In the real-time allocation level, at the beginning of each calculation period, the total monthly fixed cost is multiplied by the initial allocation coefficient for that period to obtain the initial fixed cost allocation amount for that period; the initial fixed cost allocation amount is divided by the number of calculation cycles within that period to obtain the real-time fixed cost allocation amount for each calculation cycle; simultaneously, starting from the beginning of the current calculation cycle, the actual load of the unit is continuously acquired, the single-point deviation between the actual load and the predicted load at the corresponding moment in the intraday rolling load forecast curve is calculated, and the single-point deviation is integrated and accumulated over time to obtain the cumulative deviation value from the beginning of the current calculation cycle to the current moment; In the retrospective correction stage, the absolute value of the accumulated deviation is compared with a preset deviation threshold. When the absolute value of the accumulated deviation is less than or equal to the preset deviation threshold, the initial allocation coefficient for the current period remains unchanged, and the fixed cost is allocated in real time. When the absolute value of the accumulated deviation is greater than the preset deviation threshold, online retrospective correction is triggered. The online retrospective correction uses minimizing the total allocation deviation for the remaining periods of the current calculation cycle as the objective function, and the total fixed cost to be allocated for the remaining periods as the equality constraint. A quadratic programming optimization algorithm is used to recalculate the optimized allocation coefficients for the remaining periods. The optimized allocation coefficients replace the corresponding original initial allocation coefficients and are used as the basis for fixed cost allocation in subsequent periods. The total monthly fixed cost includes monthly depreciation expenses, monthly labor costs, monthly operation and maintenance expenses, monthly management expenses, and monthly financial expenses; the calculation period is a time segment divided according to the electricity spot market trading cycle, and the calculation period is the calculation time granularity for performing full cost aggregation.

4. The multi-condition adaptive real-time full-cost accounting method for power plants according to claim 1, characterized in that, The specific steps for calculating the changes and environmental costs include: The real-time fuel cost calculation process is as follows: The real-time value of the turbine heat rate is obtained from the unit's distributed control system; the real-time value of the boiler efficiency is obtained from the boiler performance monitoring system; the design value of the pipeline efficiency is obtained from the pipeline characteristic parameter database; and the real-time value of the plant power consumption rate is obtained from the plant power monitoring system. The standard coal consumption rate for power supply is calculated by dividing the turbine heat rate by the boiler efficiency, dividing by the pipeline efficiency, dividing by 1 and subtracting the plant power consumption rate, and then dividing by the standard coal lower heating constant of 29308 kJ per kilogram. The real-time standard coal price is obtained, calculated by the fuel management system based on the received lower heating value of the coal fed into the furnace that day and the standard coal lower heating value. The real-time unit load, the standard coal consumption rate for power supply, and the real-time standard coal price are multiplied together and then divided by 1000 to obtain the real-time fuel cost in yuan per hour. The calculation process for the cost of purchased power is as follows: obtain real-time data from the metering at the connection between the generating unit and the external power grid to determine the purchased power; obtain the electricity price under the power purchase contract or the real-time spot market node price signed with the external power grid; multiply the purchased power by the electricity price under the power purchase contract or the real-time spot market node price to obtain the cost of purchased power; The real-time carbon emission cost calculation process is as follows: The real-time carbon emission factor of the unit is obtained from the environmental monitoring system. This real-time carbon emission factor is adjusted in real time based on the carbon content of the coal fed into the furnace, the unit load rate, and the boiler combustion efficiency. The real-time carbon market price is obtained from the national carbon emission trading market or the carbon market data interface of pilot areas. The real-time carbon emission cost is obtained by multiplying the unit's real-time load, the real-time carbon emission factor, and the real-time carbon market price. The real-time environmental tax and fee cost calculation process is as follows: real-time monitoring values ​​of sulfur dioxide emission concentration, nitrogen oxide emission concentration, and particulate matter emission concentration in flue gas are obtained from the environmental monitoring system; the real-time emission amount of each pollutant is calculated based on the real-time monitoring values ​​and flue gas flow rate; the environmental tax and fee collection standards corresponding to sulfur dioxide, nitrogen oxides, and particulate matter are obtained; the real-time emission amount of each pollutant is multiplied by the corresponding environmental tax and fee collection standard and then summed to obtain the real-time environmental tax and fee cost. The real-time fuel cost, purchased power cost, real-time carbon emission cost, and real-time environmental tax cost calculated within the same calculation period are summed to obtain the real-time change and comprehensive environmental cost for that calculation period.

5. The multi-condition adaptive real-time full-cost accounting method for power plants according to claim 1, characterized in that, The specific steps for calculating the lifespan loss cost include: The variable load rate component is extracted from the second dimension component of the dynamic trajectory vector of the operating condition. The variable load rate component is the first derivative of the load with respect to time after smoothing and filtering, with the unit being megawatts per minute. Its absolute value ranges from zero to the maximum variable load rate designed for the unit. The creep loss cost calculation process is as follows: First, obtain the design life parameters of key components of the unit. These key components include the high-pressure turbine rotor, the intermediate-pressure turbine rotor, the boiler superheater header, and the boiler reheater header. The design life parameters include the design life hours of each key component, which are calculated and determined by the equipment manufacturer based on the material creep characteristics under design temperature and pressure conditions. Then, obtain the equivalent operating hour increment for each key component within the current calculation period. The equivalent operating hour increment uses one hour of stable operation of the unit under rated parameters as the baseline equivalent unit, based on the current calculation period. The deviations of the actual main steam temperature and pressure from the rated values ​​during the period are converted. The higher the temperature or pressure, the greater the amplification factor of the equivalent operating hours increment relative to the actual operating time. The cost of a single overhaul of the unit is obtained, which includes labor costs, material costs, spare parts replacement costs, and power generation loss due to shutdown for maintenance during the overhaul. The equivalent operating hours increment is divided by the design life hours to obtain the creep life loss ratio in the current calculation period. The creep life loss ratio is multiplied by the cost of a single overhaul to obtain the creep loss cost. The fatigue loss cost calculation process is as follows: Online cycle counting is performed on the time series of the variable load rate component. This online cycle counting uses the rainflow counting method to identify complete stress cycles formed during the variable load process in real time. For each complete stress cycle identified, the stress amplitude of that cycle is recorded. The stress amplitude is obtained by converting the peak and trough values ​​of the variable load rate through a preset mapping relationship between variable load rate and thermal stress amplitude. Based on the stress amplitude, a pre-established correspondence curve between stress amplitude and allowable cycle count is queried to obtain the allowable cycle count corresponding to that stress cycle. This correspondence curve is determined by fitting fatigue life test data of key components of the unit. The fatigue damage degree of that stress cycle is counted as the reciprocal of the allowable cycle count corresponding to that stress cycle. The fatigue damage degree is multiplied by the single overhaul cost to obtain the single-cycle fatigue loss cost corresponding to that stress cycle. The single-cycle fatigue loss costs of all stress cycles identified within the current calculation period are accumulated, and the accumulated result is averaged out according to the duration of the calculation period to obtain the fatigue loss cost for the current calculation period. The creep loss cost and fatigue loss cost calculated within the same calculation cycle are added together to obtain the life loss cost for that calculation cycle. Among them, the fatigue loss component in the life loss cost has a superlinear functional relationship with the absolute value of the variable load rate. Specifically, in the fatigue life characteristics of the materials of key components of the unit, the stress amplitude and the allowable number of cycles have a linear negative correlation in a double logarithmic coordinate system. That is, the increase in stress amplitude leads to an exponential decrease in the allowable number of cycles, which causes the increase in the absolute value of the variable load rate to cause the growth rate of fatigue loss cost to be greater than the growth rate of the absolute value of the variable load rate. When the absolute value of the variable load rate exceeds the preset rate threshold, the fatigue loss cost shows an accelerated growth trend.

6. The multi-condition adaptive real-time full-cost accounting method for power plants according to claim 1, characterized in that, The heat-electricity sharing step specifically includes: The system obtains the extraction steam flow rate, extraction steam pressure, and extraction steam temperature of the heating extraction steam header in real time from the distributed control system of the unit. At the same time, it obtains the fresh steam pressure, fresh steam temperature, reheat steam pressure, reheat steam temperature, exhaust steam pressure, and actual power generation of the unit under the current operating conditions. It also obtains the efficiency design values ​​and rated power generation of each group of the unit under pure condensing conditions from the thermodynamic characteristic database. The process of allocating heat and electricity costs using the equivalent enthalpy drop method is as follows: First, based on the parameters of new steam, reheat steam, extraction steam, and exhaust steam, and combined with the efficiency characteristics of each stage of the turbine, the overall enthalpy drop of the turbine before extraction steam heating is calculated. The overall enthalpy drop represents the total usable heat energy released per unit mass of new steam from the inlet to the exhaust outlet. Second, the residual enthalpy drop of the turbine after extraction steam heating is calculated, that is, after deducting the enthalpy carried by the steam extracted for heating, the enthalpy drop corresponding to the continued expansion and work done by the remaining steam in the turbine. Then, the difference between the overall enthalpy drop and the residual enthalpy drop is calculated to obtain the enthalpy drop lost per unit mass of extraction steam due to extraction steam heating. Multiplying the enthalpy drop lost per unit mass of extraction steam by the extraction steam flow rate yields the power generation loss caused by extraction steam heating. The process for determining the heating cost allocation ratio is as follows: Add the power generation loss to the actual power generation of the unit to obtain the equivalent power generation under pure condensing conditions; divide the power generation loss by the equivalent power generation under pure condensing conditions to obtain the heating cost allocation ratio. This ratio represents the proportion of power generation capacity sacrificed due to heating under combined heat and power (CHP) conditions to the total power generation capacity; the power generation cost allocation ratio is obtained by subtracting the heating cost allocation ratio from 1. The calculated real-time fuel cost, purchased power cost, real-time carbon emission cost, real-time environmental tax cost, creep loss cost, and fatigue loss cost are multiplied by the heating cost allocation ratio to obtain the share of each cost to be borne by the heating side. The above costs are then multiplied by the power generation cost allocation ratio to obtain the share of each cost to be borne by the power generation side. The costs to be borne by the power generation side are then summed to obtain the variable cost, environmental cost, and lifetime loss cost to be borne by the power generation side, for use in the subsequent full cost aggregation step.

7. The multi-condition adaptive real-time full-cost accounting method for power plants according to claim 1, characterized in that, The full-cost aggregation step specifically includes: Within each calculation cycle, the following cost element values ​​are obtained: the real-time fixed cost allocation amount from the prediction-allocation-backtracking three-level closed-loop dynamic allocation model, the real-time variable and environmental comprehensive cost from the variable and environmental cost calculation, the life loss cost from the life loss cost calculation, and the variable cost of the power generation side, the environmental cost of the power generation side, and the life loss cost of the power generation side after heat-electricity allocation. The operating condition dynamic trajectory vector and transition state membership vector corresponding to the current calculation cycle are obtained from the operating condition classification model; the variable load rate component and load rate component are extracted from the operating condition dynamic trajectory vector; and the membership values ​​between the current operating state and each preset steady-state operating condition are obtained from the transition state membership vector. The pre-built working condition-weight mapping function library is constructed as follows: For each cost element, a weight mapping function is established with the relevant components in the working condition dynamic trajectory vector and the relevant membership degree values ​​in the transition state membership degree vector as independent variables; the weight mapping function is determined by combining offline simulation analysis and historical operation data regression analysis. In the offline stage, thermal simulation and cost sensitivity analysis are performed for different working condition combinations to obtain reasonable weight values ​​for each cost element under different working conditions. A continuous weight mapping function is established through function fitting method. For the real-time allocation of fixed costs, its dynamic weighting function uses the membership degree values ​​of the load factor component and the transition state membership degree vector corresponding to the deep peak shaving condition as independent variables; when the load factor decreases or the membership degree of the deep peak shaving condition increases, the value of the dynamic weighting function of fixed costs increases accordingly to reflect the economic law that the fixed costs of the unit should be borne by a larger share of the electricity volume when the unit is operating at low load. For lifetime loss cost, its dynamic weighting function uses the absolute value of the variable load rate component as the main independent variable. When the absolute value of the variable load rate component is less than or equal to a preset first rate threshold, the dynamic weighting function value of lifetime loss cost remains at a baseline value of 1. When the absolute value of the variable load rate component is greater than the preset first rate threshold, the dynamic weighting function value of lifetime loss cost increases in a superlinear manner with the increase of the absolute value of the variable load rate component, and the slope of the tangent line of the growth curve gradually increases with the increase of the absolute value of the variable load rate component. When the absolute value of the variable load rate component reaches a preset second rate threshold, the dynamic weighting function value of lifetime loss cost reaches a preset maximum weighting coefficient, and thereafter the preset maximum weighting coefficient remains unchanged. For real-time changes and comprehensive environmental costs, the dynamic weighting function uses the membership degree values ​​of the load rate component and the transition state membership degree vector corresponding to the rated load condition as independent variables. When the load rate is close to the rated value, the weighting function value approaches 1. When the load rate deviates from the rated value, it is adjusted accordingly according to the preset correction curve. Multiply the value of each cost element by its corresponding dynamic weighting function value to obtain the weighted cost of each element; sum the weighted costs of each element to obtain the unit-level real-time full cost for the current calculation period.

8. The multi-condition adaptive real-time full-cost accounting method for power plants according to claim 1, characterized in that, The cost output step specifically includes: At the end of each calculation cycle, obtain the unit-level real-time total cost and the average unit power generation during the current calculation cycle; divide the unit-level real-time total cost by the average power generation to obtain the unit power real-time total cost per unit of electricity for that calculation cycle, in yuan per kilowatt-hour. The preset period is a time interval synchronized with the electricity spot market trading cycle; for the day-ahead spot market, the preset period is 15 minutes, corresponding to the 96-point daily trading session; for the real-time spot market, the preset period is 5 minutes, corresponding to the real-time market clearing cycle. The real-time full cost per unit of electricity and the real-time full cost at the generator level calculated within each preset period are pushed to the spot market quotation decision system in real time through the application programming interface. The pushed data also includes the cost composition details within the preset period, which include the values ​​and proportions of fixed cost allocation, fuel cost, purchased power cost, carbon emission cost, environmental protection tax cost, creep loss cost and fatigue loss cost. After receiving the real-time full cost per unit of electricity, the spot market pricing decision system uses the real-time full cost per unit of electricity as a reference value for the lower limit of the price. When the unit is in an adjustable state within the current preset cycle and plans to participate in the bidding for the next trading session, the spot market pricing decision system generates a suggested price based on the real-time full cost per unit of electricity and adds a preset target profit rate. The real-time full cost per unit of electricity and the real-time full cost per unit of electricity calculated within each preset period are synchronously output to the operation optimization system. The operation optimization system compares the real-time full cost per unit of electricity in the current preset period with the real-time full cost per unit of electricity in adjacent historical periods, calculates the cost change rate and the contribution of each cost component. When the real-time full cost per unit of electricity exceeds the preset cost alarm threshold, the operation optimization system generates a cost anomaly alarm signal, retrieves the key operating parameters and cost composition details for the corresponding time period, and pushes them to the operator's terminal. The operation optimization system also compares the current operating condition dynamic trajectory vector and the corresponding real-time full cost per unit of electricity with a pre-established optimal operating condition-cost mapping table. When the current real-time full cost per unit of electricity is higher than the historical optimal cost level under the same or similar operating conditions in the mapping table and the excess exceeds the preset optimization trigger threshold, the system pushes operation optimization suggestions to the operator's terminal. The operation optimization suggestions include the target value of the main steam parameter to be adjusted, the operating combination of the plant power equipment to be adjusted, or the coal blending ratio to be adjusted.

9. A real-time full-cost accounting system for power plants that adapts to multiple operating conditions, characterized in that, include: Operating condition identification module: Used to acquire unit operating parameter stream in real time. The operating parameters include at least unit load, main steam temperature, main steam pressure, reheat steam temperature, reheat steam pressure, feedwater flow rate, plant power consumption rate, and fuel input. Statistical features are extracted from the operating parameter stream using a sliding time window processing method. The statistical features include the mean, standard deviation, rate of change, and spectral energy distribution within the window. The operating condition identification module has a built-in pre-trained operating condition classification model. The statistical features are input into the operating condition classification model, and the module outputs the current operating condition dynamic trajectory vector and transition state membership vector. The operating condition dynamic trajectory vector is a multi-dimensional vector containing load rate, load change rate, operating mode, fuel combination, and time dimensions. The transition state membership vector is a set of continuous values ​​ranging from 0 to 1, representing the similarity between the current operating state and multiple preset steady-state operating conditions. Each component of the transition state membership vector changes continuously and smoothly over time during the operating condition switching process. Data fusion and allocation module: Used for real-time acquisition of multi-source heterogeneous data. This multi-source heterogeneous data includes at least second-level / minute-level production operation data obtained from the power plant's plant-level monitoring information system, coal quantity and quality data from the fuel management system, pollutant emission concentration and carbon emission factor data from the environmental monitoring system, and monthly fixed cost data from the financial system. The second-level / minute-level production operation data is aggregated into a minute-level average sequence using a sliding window. The monthly fixed cost data is dynamically allocated using a three-level closed-loop system of prediction, allocation, and backtracking. The model generates real-time fixed cost allocation for each calculation period; wherein, the three-level closed-loop dynamic allocation model includes: a prediction level, which generates initial allocation coefficients for each period based on the intraday rolling load forecast curve and time-of-use electricity price signal; a real-time allocation level, which allocates the total monthly fixed cost to each calculation period according to the initial allocation coefficients and continuously monitors the cumulative deviation between the actual load and the predicted load; and a backtracking correction level, which optimizes and updates the allocation coefficients for the remaining periods online when the cumulative deviation exceeds a preset threshold, with the objective function of minimizing the total allocation deviation for the remaining periods of the month. The variable and environmental cost calculation module is used to calculate real-time fuel costs based on real-time fuel input, real-time fuel calorific value, and real-time fuel price in the operating parameter stream using an inverse balance method; calculate purchased power costs based on plant power consumption rate; calculate real-time carbon emission costs based on carbon emission factor data and real-time carbon market prices; calculate real-time environmental tax costs based on pollutant emission concentrations and corresponding environmental tax standards; and sum the real-time fuel costs, purchased power costs, real-time carbon emission costs, and real-time environmental tax costs to generate real-time variable and environmental comprehensive costs. The life loss cost calculation module is used to extract the variable load rate component from the dynamic trajectory vector of the operating condition; based on the variable load rate component, it calculates the creep loss cost and fatigue loss cost respectively; wherein, the creep loss cost is obtained by multiplying the ratio of the equivalent operating hour increment to the design life hours by the cost of a single overhaul; the fatigue loss cost is obtained by performing online cyclic counting on the variable load rate component sequence, identifying effective stress cycles, calculating the fatigue damage degree according to the allowable number of cycles corresponding to the stress amplitude of each stress cycle, multiplying the fatigue damage degree by the cost of a single overhaul and allocating it to the duration of the stress cycle; the life loss cost is the sum of the creep loss cost and the fatigue loss cost, and the fatigue loss component related to the variable load rate in the life loss cost has a superlinear functional relationship with the absolute value of the variable load rate. The heat-power cost allocation module is used to obtain the steam extraction flow rate, extraction parameters, and actual power generation of the unit from the distributed control system when the unit is operating in cogeneration mode; calculate the power generation loss caused by steam extraction for heating using the equivalent enthalpy drop method; determine the heating cost allocation ratio based on the ratio of the power generation loss to the rated power under pure condensing mode; and allocate the calculated costs between the power generation side and the heating side based on the heating cost allocation ratio to obtain the variable costs, environmental costs, and life loss costs to be borne by the power generation side. The full cost aggregation module is used to obtain the real-time allocation of fixed costs, the real-time variable and environmental comprehensive costs, the lifetime loss costs, and the allocated generation-side costs. Based on the current operating condition dynamic trajectory vector and transition state membership vector, it queries the operating condition-weighting mapping function library built into the full cost aggregation module to determine the dynamic weighting function corresponding to each cost element. The dynamic weighting function is a continuous function with the operating condition dynamic trajectory vector and transition state membership vector as independent variables, and for lifetime loss costs, its dynamic weighting function value is positively correlated with the absolute value of the load change rate. The dynamic weighting function is used to perform a weighted summation of each cost element to generate the unit-level real-time full cost at the current calculation moment. Cost output module: used to divide the unit-level real-time total cost by the current power generation to obtain the unit-level real-time total cost; and output the unit-level real-time total cost and the unit-level real-time total cost to the spot market quotation decision system and the operation optimization system at preset intervals.

10. The multi-condition adaptable real-time full-cost accounting system for power plants according to claim 9, characterized in that, The full-cost aggregation module further includes: The cost element acquisition unit is used to acquire the following cost element values ​​within the same calculation cycle: the real-time fixed cost allocation amount from the data fusion and allocation module, the real-time variable and environmental comprehensive cost from the variable and environmental cost calculation module, the life loss cost from the life loss cost calculation module, and the allocated power generation side variable cost, power generation side environmental cost, and power generation side life loss cost from the heat-electricity allocation module. The working condition information acquisition unit is used to acquire the working condition dynamic trajectory vector and transition state membership vector corresponding to the current calculation cycle from the working condition identification module, and extract the variable load rate component and load rate component from the working condition dynamic trajectory vector, and extract the membership value between the current operating state and each preset steady-state working condition from the transition state membership vector. The system includes a working condition-weighting mapping function library, which stores dynamic weighting functions corresponding to various cost elements. These dynamic weighting functions are continuous functions with the relevant components in the dynamic trajectory vector of the working condition and the relevant membership values ​​in the membership vector of the transition state as independent variables. The dynamic weighting functions are pre-constructed through a combination of offline simulation analysis and regression analysis of historical operating data. In the offline phase, thermal simulation and cost sensitivity analysis are performed for different working condition combinations to obtain the weight values ​​of each cost element under different working conditions. A continuous weighting mapping relationship is then established through function fitting methods. For the real-time allocation of fixed costs, its dynamic weighting function uses the membership degree values ​​of the load rate component and the corresponding deep peak shaving conditions in the transition state membership degree vector as independent variables. When the load rate decreases or the membership degree value of the deep peak shaving conditions increases, the value of the dynamic weighting function of fixed costs increases accordingly. For lifetime loss cost, its dynamic weighting function uses the absolute value of the variable load rate component as the main independent variable; when the absolute value of the variable load rate component is less than or equal to a preset first rate threshold, the dynamic weighting function value of lifetime loss cost remains at a baseline value of one; when the absolute value of the variable load rate component is greater than the preset first rate threshold, the dynamic weighting function value of lifetime loss cost increases in a superlinear manner with the increase of the absolute value of the variable load rate component, and the slope of the tangent line of the growth curve gradually increases with the increase of the absolute value of the variable load rate component; when the absolute value of the variable load rate component reaches a preset second rate threshold, the dynamic weighting function value of lifetime loss cost reaches a preset maximum weighting coefficient, and thereafter the preset maximum weighting coefficient remains unchanged; For real-time changes and comprehensive environmental costs, the dynamic weighting function uses the membership degree values ​​of the load rate component and the corresponding rated load conditions in the transition state membership vector as independent variables. When the load rate is close to the rated value, the weighting function value approaches one, and when the load rate deviates from the rated value, it is adjusted accordingly according to the preset correction curve. The weighted summation unit is used to multiply the value of each cost element by the function value of its corresponding dynamic weighting function in the current calculation period to obtain the weighted cost of each element, and to sum up the weighted costs of each element to generate the unit-level real-time full cost for the current calculation period. The calculation period is the time granularity for performing full cost aggregation, and the length of the calculation period is less than or equal to the transaction clearing period of the electricity spot market.

Images