Thermal power unit output performance evaluation method based on time-sharing energy consumption fitting

Abstract

This invention relates to the field of thermal power unit analysis and evaluation technology, specifically to a method for evaluating the output performance of thermal power units based on time-sharing energy consumption fitting. The method includes the following steps: dividing the operating cycle of the thermal power unit into multiple time periods, collecting independent energy consumption data within each time period, and simultaneously collecting actual load data; calculating the response time lag between energy consumption changes and load changes of each component to form a cross-component time-sharing energy consumption response trajectory; establishing a time-sharing energy consumption decoupling fitting framework including a phase adjustment function and a response elastic window, outputting quasi-steady-state energy consumption and purified dynamic additional energy consumption; generating a phase-purified output performance curve, and identifying abnormal response stages with high adjustment costs based on the proportion of the two types of energy consumption and performance deviation characteristics, outputting a comprehensive evaluation result including performance level, response burden ratio, and energy consumption deviation source. This invention achieves purification and fine decomposition of energy consumption signals, solving the pain point of confusing normal fluctuations with abnormal consumption in traditional evaluations.

Description

Technical Field

[0001] This invention relates to the field of thermal power unit analysis and evaluation technology, and in particular to a method for evaluating the output performance of thermal power units based on time-sharing energy consumption fitting. Background Technology

[0002] As the proportion of renewable energy installed capacity in the power system continues to rise, thermal power units, as the basic regulating power source, are undertaking more and more peak-shaving and frequency regulation tasks, and their operating state is shifting from traditional long-term stable load to frequent dynamic regulation. In this context, the energy efficiency level of the units no longer depends solely on steady-state operating conditions, but is more affected by dynamic regulation behavior. This change makes a precise assessment of the output performance of thermal power units a core foundation for improving operating efficiency and optimizing regulation strategies.

[0003] Existing energy consumption assessment methods primarily rely on regression fitting of time-period load and energy consumption, constructing overall or partial load-energy consumption models to estimate system energy efficiency levels. However, the response of various components in thermal power units (such as boilers, turbines, and auxiliary equipment) to load changes exhibits time lag, and energy consumption fluctuations in different components do not occur simultaneously. Existing methods typically process load changes and energy consumption data synchronously, ignoring this asynchronous response characteristic. This leads to a mismatch between regulation events and energy consumption fluctuations, severely impacting the accuracy of the fitting and the interpretability of the results. Furthermore, in actual operation, energy consumption fluctuations not only originate from load regulation but also include factors such as the inherent regulation flexibility of the equipment and control oscillations. Existing methods often include all energy consumption fluctuations in regulation costs, making it difficult to isolate the additional energy consumption truly caused by regulation. This can easily lead to overestimation of regulation costs and inaccurate problem identification. Summary of the Invention

[0004] This invention provides a method for evaluating the output performance of thermal power units based on time-sharing energy consumption fitting. By phase alignment and elastic stripping, the steady-state and regulation parts of the time-sharing energy consumption of thermal power units are separated, and a hyperbola evaluation system is constructed to achieve high-precision output performance evaluation and abnormal response identification.

[0005] The method for evaluating the output performance of thermal power units based on time-of-use energy consumption fitting includes the following steps: S1, Construct cross-component time-sharing energy consumption response trajectory: Divide the operating cycle of the thermal power unit into multiple time periods, collect independent energy consumption data of the boiler system, turbine system and main auxiliary system in each time period, and collect the actual load at the same time, calculate the response time delay between the energy consumption change of each component and the load change, and form a cross-component time-sharing energy consumption response trajectory to describe the asynchronous energy consumption evolution law caused by load adjustment in the time period; S2, Establish an energy consumption response phase alignment and response elastic window decoupling fitting model: Based on the cross-component time-sharing energy consumption response trajectory, perform phase alignment on the energy consumption data of each component, and redistribute the lagging energy consumption to the actual adjustment period; after phase alignment, construct the component response elastic window according to the local sensitivity of each component's energy consumption to load changes, elastically strip away the energy consumption fluctuations that can be absorbed by the component itself, retain only the real dynamic adjustment energy consumption that cannot be absorbed, and establish a time-sharing energy consumption decoupling fitting framework including phase adjustment function and response elastic window, outputting quasi-steady-state energy consumption and purified dynamic additional energy consumption; S3, Perform output performance evaluation under phase purification: Fit the quasi-steady-state energy consumption and dynamic additional energy consumption to the time-segmented output data respectively to generate the output performance curve after phase purification, and identify the abnormal response stage with high adjustment cost based on the proportion of the two types of energy consumption and the performance deviation characteristics, and output a comprehensive evaluation result including performance level, response burden ratio and energy consumption deviation source.

[0006] Optionally, the time-segmentation is based on the variable duration triggered by the event, including dividing the operating cycle into multiple time segments with the significant change point of the load command received by the thermal power unit as the boundary; wherein, the significant change point is the moment when the change amplitude of the load command exceeds the load change amplitude threshold or the change rate exceeds the load change rate threshold, so that each time segment represents a complete load adjustment event.

[0007] Optionally, key energy consumption data may be collected synchronously within each of the aforementioned time periods: Coal consumption, air volume, and flue gas temperature of the boiler system; Internal efficiency of the steam turbine system and extraction parameters at each stage; In addition, the power consumption of the main auxiliary systems, including the feed water pump, condensate pump, and forced draft fan; As independent energy consumption data for each system; Simultaneously, the actual load generated by the unit is collected.

[0008] The response time delay is calculated based on cross-correlation analysis. Specifically, for each load adjustment event, the actual load data is used as the baseline sequence, and the key energy consumption data of each component is collected as the comparison sequence. The time shift when the cross-correlation function reaches its maximum value is calculated by the sliding window cross-correlation analysis method. The time shift is determined as the response time delay of the corresponding component in the event of energy consumption change relative to load change.

[0009] Optionally, S1 further includes integrating the start time, end time, average load, independent energy consumption data sequence of each component, and calculated response time delay of each time period into a cross-component time-sharing energy consumption response trajectory, which is used to describe the asynchronous energy consumption evolution law of each component being out of sync in time due to load adjustment within the time period.

[0010] Optionally, the phase alignment specifically includes: acquiring the cross-component time-sharing energy consumption response trajectory formed by S1; for each load adjustment event in the response trajectory, reading the response delay of each component calculated for the load adjustment event; based on the response delay, constructing a phase adjustment function, shifting the energy consumption data sequence of the current component in the time period forward on the time axis with the response delay as a single unit, generating a phase-aligned energy consumption data sequence, and redistributing the energy consumption caused by the response delay in time to the load adjustment period that actually caused the energy consumption.

[0011] Optionally, the construction of the response elastic window includes: based on the phase-aligned data, calculating the local sensitivity of the energy consumption of each component to load changes, wherein the local sensitivity is the ratio of the component's energy consumption change to the load change within a preset short-term load fluctuation range; based on historical operating data, statistically analyzing the normal distribution range of the energy consumption fluctuation described by the local sensitivity of each component at the same load point, and defining the normal distribution range as the response elastic window of the component.

[0012] Optionally, the elastic stripping includes: for the phase-aligned energy consumption data sequence, comparing its energy consumption value at each moment with the quasi-steady-state expected energy consumption fitted from historical data based on the load at that moment, calculating the instantaneous energy consumption deviation; if the instantaneous energy consumption deviation falls within the response elastic window, then the energy consumption fluctuation is determined to be a normal fluctuation that can be absorbed by the component itself, and it is stripped from the deviation; if the instantaneous energy consumption deviation exceeds the response elastic window, then the excess energy consumption is retained as the purified, real dynamic additional energy consumption that cannot be absorbed by the component.

[0013] Optionally, S2 further includes integrating the phase adjustment function and the response elastic window to establish a time-sharing energy consumption decoupling fitting framework. The input to the time-sharing energy consumption decoupling fitting framework is the original cross-component time-sharing energy consumption response trajectory. After processing, it finally outputs two results for each component within each time period, including: The quasi-steady-state expected energy consumption is taken as the quasi-steady-state energy consumption; The dynamic additional energy consumption sequence obtained after purification by the flexible window is used as the purified dynamic additional energy consumption.

[0014] Optionally, S3 includes generating the output performance curve after phase purification through bilinear fitting, specifically including: Quasi-steady-state performance curve fitting: The quasi-steady-state energy consumption of each time period output by S2 is fitted with the average output data of the corresponding time period using the least squares method for linear or nonlinear regression fitting, generating a benchmark performance curve characterizing the energy consumption-output relationship of the unit under ideal steady-state conditions.

[0015] Dynamic additional energy consumption-output relationship fitting: The dynamic additional energy consumption of each time period output by S2 is fitted with the average output data in the corresponding time period by scatter distribution to generate a dynamic additional energy consumption envelope that represents the dynamic adjustment cost under different output levels. The baseline performance curve and the dynamic additional energy consumption envelope together constitute the output performance curve after phase purification. The identification of the abnormal response phase includes: Calculate the dynamic additional energy consumption percentage for each time period; Calculate the performance deviation rate for each time period; When the dynamic additional energy consumption ratio and performance deviation rate of a certain time period both exceed the dual thresholds set according to the unit design and historical operation data, the current stage is determined to be an abnormal response stage.

[0016] Optionally, the generation of the comprehensive evaluation results includes: Performance rating: Based on the average value and dispersion of the performance deviation rate, the overall performance of the unit throughout the entire operating cycle is classified into excellent, qualified, warning, and poor levels; Response load ratio: defined as the ratio of total dynamic additional energy consumption to total quasi-steady-state energy consumption over the entire assessment period, used to quantify the overall regulation cost of the unit; Sources of energy consumption deviation: By analyzing the contribution ratio of the dynamic additional energy consumption of the three major systems of boiler, steam turbine and main auxiliary equipment to the total dynamic additional energy consumption during the abnormal response phase, the main source system leading to performance degradation is identified.

[0017] The beneficial effects of this invention are: 1. This invention proposes a strategy based on cross-correlation analysis to calculate response delay and perform phase alignment, which solves the problems of asynchronous component responses and misalignment of energy consumption and load changes in traditional energy consumption assessment. By shifting the energy consumption sequence of each component forward according to the response delay, it aligns the component with the actual adjustment behavior in time, constructs a cross-component time-sharing energy consumption response trajectory, and improves the correlation between energy consumption behavior and system state.

[0018] 2. Based on traditional energy consumption fitting, this invention introduces a component response elastic window and sets upper and lower floating thresholds based on the local sensitivity distribution of historical operating data statistics to determine whether the energy consumption deviation falls within the system's self-absorption range. Through this mechanism, energy consumption fluctuations that can be buffered by the components themselves are stripped away, leaving only non-absorbable dynamic additional energy consumption. This achieves the purification and fine decomposition of energy consumption signals, solves the pain point of confusing normal fluctuations with abnormal consumption in traditional assessments, improves the accuracy of dynamic energy efficiency assessments, and provides a solid foundation for identifying high-adjustment-cost operating behaviors.

[0019] 3. This invention generates a phase-purified output performance curve composed of a performance benchmark curve and a dynamic energy consumption envelope by fitting the steady-state energy consumption and dynamic additional energy consumption respectively. It constructs a dual reference system of ideal state and regulation cost. At the same time, it proposes a dual threshold judgment method combining the proportion of dynamic additional energy consumption and the performance deviation rate to identify abnormal response stages with high regulation costs and low operating efficiency. Through system attribution analysis, it quantifies the specific contributions of boiler, turbine and auxiliary equipment to performance degradation, forming an evaluation process from assessment, identification and location. This provides a highly implementable and quantifiable support path for power plant operation optimization and regulation strategy adjustment. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the evaluation method flow according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the construction of cross-component time-sharing energy consumption response trajectory in an embodiment of the present invention. Detailed Implementation

[0022] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. For some well-known technologies, those skilled in the art may also use other alternative methods to implement the invention. Moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0023] like Figures 1-2 As shown, the method for evaluating the output performance of thermal power units based on time-of-use energy consumption fitting includes the following steps: S1. Construct cross-component time-sharing energy consumption response trajectory: Divide the operating cycle of the thermal power unit into multiple time periods, collect independent energy consumption data of the boiler system, turbine system and main auxiliary equipment system in each time period, and collect the actual generated load simultaneously. Calculate the response time delay between the energy consumption change and load change of each component to form a cross-component time-sharing energy consumption response trajectory, which is used to describe the asynchronous energy consumption evolution law caused by load adjustment in the time period.

[0024] S1 specifically includes: S11, Event-triggered variable-duration time-segmentation: Based on the load command sequence received during the operation of the thermal power unit, significant inflection points of load changes are identified, and the operating cycle is divided into multiple continuous, non-overlapping time segments; each significant change point is a moment that meets one of the following conditions: The change range of the load command satisfies: ; Or: The rate of change satisfies: ;in, Indicates time The load command value (MW), Indicates the sampling time interval (s). Indicates the threshold for the magnitude of load change. ; For the unit's rated output, a deviation of less than 3% may lead to misinterpreting minor fluctuations as events and causing over-segmentation; a deviation of more than 8% may cause some moderate but energy-intensive regulation behaviors to be missed. Therefore, a value of [missing value] is recommended. Rated power, that is, 30MW for a 600MW unit. The load change rate threshold, with a range of values. The data is derived from power plant peak-shaving capacity or load variation statistics. Values ​​less than 2 MW / min are too sensitive and may misinterpret normal minor adjustments as regulation events. Values ​​greater than 6 MW / min can only capture large and rapid adjustments, losing some analytical value. A value of [value missing] is recommended. MW / min, suitable for most medium-sized thermal power units.

[0025] This section addresses the issue of how to scientifically divide time periods in the performance evaluation of thermal power units. The core idea is to adopt an event-triggered, variable-duration division method, which differs from the traditional method of dividing by fixed time intervals and better reflects the actual load regulation behavior during the operation of thermal power units. In actual operation, regulation behavior often does not occur uniformly but is driven by certain load abrupt events triggered by dispatch commands. This method proposes to abandon the fixed time window division and instead consider each significant load change as the start or end of a new regulation event.

[0026] The identification of significant change points used two judgment criteria: First, the load amplitude change is greater than the threshold: that is, the load value suddenly increases or decreases by a sufficiently large amount, indicating that the system has undergone adjustment behavior worthy of analysis; Second, the load change rate is too fast: that is, the load changes very quickly per unit time, indicating that the system is in a state of drastic adjustment.

[0027] If any of the above conditions are met, it is determined to be an event boundary, and each time period with practical regulatory significance is divided accordingly.

[0028] This method ensures that each time period corresponds to a complete adjustment event, which facilitates subsequent dynamic energy consumption separation. The time period length is adaptive and variable, which can more accurately capture the dynamic energy consumption pattern in the unsteady process of the unit and solve the problem that traditional fixed window analysis methods are prone to interrupting events or mixing multiple events.

[0029] S12, Synchronous Acquisition of Multi-Source Heterogeneous Data: Within each divided time period, the following multi-source data acquisition is performed: Boiler system data collection: coal consumption (kg / s), primary air volume and secondary air volume (Nm3 / s), flue gas temperature (°C); Steam turbine system data collection: internal efficiency (%), extraction steam pressure and temperature at each stage (MPa, °C); Data collected from major auxiliary systems: active power (kW) of feedwater pumps, condensate pumps, forced draft fans, and induced draft fans; Actual load data: Actual load (MW), with a sampling frequency of not less than 1Hz, to ensure that load and energy consumption data have high temporal resolution.

[0030] Actual load (i.e. power generation output) is the core reference variable and the most important operating target of thermal power units. It determines the power output level per unit time. When calculating the response lag of energy consumption of various components such as boilers, turbines, and auxiliary equipment to load changes, actual load is used as the benchmark sequence. Actual load is also a direct observation variable for judging whether a regulation event has occurred.

[0031] S13, Response time delay calculation based on cross-correlation analysis: For load adjustment events within each time period, based on actual load... As a baseline sequence, the key energy consumption sequences of each component are used respectively. Using parameters such as coal combustion flow and active power as a comparison sequence, the sliding window cross-correlation analysis method is employed to calculate the response time delay of energy consumption changes relative to load changes. Specifically: This formula indicates that this is for components. Perform a sliding cross-correlation analysis between the energy consumption series and the actual load series to find the time offset with the largest cross-correlation function value. This offset This refers to the response time lag of the component's energy consumption to load changes, specifically how many seconds it takes for it to begin to show a significant response to load variations. Indicates load and the first The cross-correlation function of the energy consumption of each component is defined as: This formula indicates that the load and the first... Energy consumption of individual components over time The cross-correlation function below indicates when the energy consumption curve shifts forward / backward. How strong is the overall overlap (correlation) with the load curve in seconds? In other words, it measures the degree of overlap with the load curve during offset. Under these conditions, the degree of time synchronization between energy consumption and load. Among them, This represents the optimal time shift of energy consumption change relative to load change, i.e., response time delay.

[0032] The above addresses the question of how to quantify the energy consumption response lag of various components, specifically how to determine whether the energy consumption changes of the boiler, turbine, or auxiliary equipment occur before, simultaneously with, or after the load change during a load adjustment event, thus revealing the time-series response relationship of energy consumption changes to load adjustment. In thermal power units, load changes do not immediately reflect in the energy consumption of all components simultaneously. For example, if the main engine load has just increased, the boiler's coal consumption may increase immediately, while the condensate pump's energy consumption may not show a significant change until several seconds or tens of seconds later. To capture this response lag, cross-correlation analysis is used as described above. Specifically, it includes the following: 1. Based on actual load As a reference curve (input), because it represents the actual adjustment target of the entire system; 2. Based on the energy consumption data of a certain component As a response curve (output), such as the amount of coal burned in a boiler or the power of an auxiliary machine; 3. Within a time window, perform cross-correlation calculations on the two curves, that is, slide one curve forward / backward and multiply it point by point with the other curve to find the time difference with the highest similarity; 4. Obtain the time lag corresponding to the maximum cross-correlation value; this is the response time lag of this component. .

[0033] The reason for adopting the above scheme is that if the alignment of two sequences is highest after multiplication at a certain time offset, then this offset can be considered the optimal time shift for the response. In this case, it can be said that the energy consumption response of the component in this event is later than that of the load. Second.

[0034] S14, Trajectory Construction and Integration: During each unit adjustment process, a representative energy consumption behavior record is generated. This record includes multiple time series and key features. Organizing this data through a unified structure facilitates subsequent analysis, such as energy consumption model fitting, anomaly identification, and performance evaluation. Within each time segment identified through event triggering, the following tasks have been completed: The start and end times of this time slot have been clearly defined; Key energy consumption parameters (time series) of various systems such as boilers, steam turbines, and auxiliary equipment were collected. The actual load was recorded synchronously; For each component, its time lag in response to the load (i.e., response time delay) was calculated.

[0035] The above data is packaged into a single object in a unified format for subsequent processing. This can be understood as encapsulating it into a data trajectory, with each trajectory representing an adjustment process. In other words, the data within each time period is organized into a structured trajectory object, containing: Time boundaries: When does this adjustment process begin and end; that is, the start and end times of each time period. ; Average load The average output level throughout the entire process; Energy consumption sequence of each component For example, the curve of boiler coal flow rate changing over time; Response time delay of each component The time by which the energy consumption of this component lags behind the load change.

[0036] The above data is integrated to construct a cross-component time-sharing energy consumption response trajectory. Formally, it can be represented as: This trajectory is used to describe the asynchronous energy consumption evolution of various components due to load regulation during this time period, and serves as the basic input for subsequent response phase alignment and performance evaluation analysis.

[0037] S2, Establish an energy consumption response phase alignment and response elastic window decoupling fitting model: Based on the cross-component time-sharing energy consumption response trajectory, perform phase alignment on the energy consumption data of each component, and redistribute the lagging energy consumption to the actual adjustment period; after phase alignment, construct the component response elastic window according to the local sensitivity of each component's energy consumption to load changes, elastically strip away the energy consumption fluctuations that can be absorbed by the component itself, retain only the real dynamic adjustment energy consumption that cannot be absorbed, and establish a time-sharing energy consumption decoupling fitting framework including the phase adjustment function and the response elastic window, and output the quasi-steady-state energy consumption and the purified dynamic additional energy consumption.

[0038] S21, Phase Alignment and Energy Redistribution: In thermal power units, different components respond to load adjustments at different speeds, meaning their energy consumption changes do not occur simultaneously with load changes. For example, boiler coal consumption may increase immediately at the start of load increases, while the forced draft fan or extraction system may only show a significant energy consumption response after several seconds or tens of seconds. This asynchronous response creates a time lag. If left unaddressed, this will lead to a misalignment between energy consumption changes and actual adjustment behavior, distorting the analyzed energy consumption patterns and failing to accurately reflect which time period's load changes caused which energy consumption. Therefore, it is necessary to align the lagging energy consumption to the actual adjustment period. Specifically, this includes: for the cross-component time-sharing energy consumption response trajectory constructed in S1, for the load adjustment event corresponding to each trajectory, obtaining the response time lag of each component. Based on this, a phase adjustment function for component energy consumption data is constructed to shift the lagging energy consumption sequence forward on the time axis. Each unit corresponds to the adjustment period that actually causes changes in energy consumption. The adjusted energy consumption data is expressed as follows: ;in, To adjust the previous number Each component at the current moment Energy consumption value, For the adjusted energy consumption data sequence, For components The response time delay enables the realignment of the energy consumption phase, ensuring that subsequent analysis focuses on the actual energy consumption response corresponding to load changes, rather than the delayed behavior.

[0039] S22, Constructing a Local Sensitivity-Driven Response Elasticity Window: This part aims to further identify and isolate normal energy consumption fluctuations that can be absorbed by the components themselves, based on the phase alignment already completed, thus leaving only the abnormal additional energy consumption truly caused by regulation. A key concept is proposed: the response elasticity window, which is the acceptable range of energy consumption fluctuation for each component during normal operation. Specifically, this includes the energy consumption data sequence after phase alignment. Based on this, the local sensitivity of component energy consumption to load changes is calculated. This sensitivity is used to characterize the energy consumption response of a component to short-term load disturbances, and is defined as: Local sensitivity is essentially a slope index, indicating how much the energy consumption of a component changes due to load changes within a short time range. This ratio shows how much the energy consumption of the component changes (e.g., how many kJ or kW) for every unit change in load (e.g., 1 MW). For components Energy consumption variation within a short time window This represents the load variation within the corresponding time period.

[0040] We know that minor fluctuations will always occur in the components of the generating unit during normal operation. These do not necessarily indicate a malfunction, nor are they necessarily additional energy consumption caused by adjustments. For example, inaccurate adjustments, fluctuations in the system control itself, and slight changes in coal quality may cause these fluctuations. To avoid misinterpreting these normal fluctuations as abnormal energy consumption, a flexible range, or response flexibility window, is set for each component. This includes adjustments based on historical operating data for each component at typical load points. The resulting energy consumption fluctuations are statistically analyzed to obtain the normal range of energy consumption fluctuations, which is defined as the response elasticity window: ;in, This represents the maximum absorbable fluctuation threshold obtained by fitting the historical sensitivity distribution of the component, and indicates the acceptable range of energy consumption fluctuations for the component without external intervention.

[0041] It is an empirical statistical threshold, automatically determined based on the actual operating data of the component, using the following method: 1. Aggregate historical data: Collect data on this component over multiple operating cycles, and extract local sensitivity over a short period of time each time the load is at a certain level (300MW). .

[0042] 2. Statistical analysis of each load point Distribution: 500 points were calculated at a load point of 300MW. These values ​​will exhibit a certain range of fluctuation, with most concentrated in... nearby.

[0043] 3. Fit the fluctuation range and set a threshold: For the above... The data is fitted to a normal or quantile distribution, and the middle 90% or 95% range (e.g., ±1.96 times the standard deviation) is usually taken as the elasticity range, multiplied by the current load change. , and you can get : ;in, The standard deviation of component sensitivity (or the maximum value based on the percentile), This represents the load change magnitude within the current adjustment cycle.

[0044] Local sensitivity It refers to the component's sensitivity to load fluctuations, and its response elasticity window. It refers to the range of energy consumption fluctuations that it can absorb under healthy conditions. It is the maximum tolerable fluctuation range calculated based on historical fluctuations, used to determine what constitutes true abnormal energy consumption.

[0045] S23, Elasticity Stripping and Dynamic Additional Energy Consumption Extraction: During load regulation, energy consumption changes in certain components may not be directly caused by the regulation behavior, but rather by factors such as small oscillations in the control strategy itself, system feedback lag or regulation residuals, environmental disturbances, or fuel fluctuations. Although these energy consumption fluctuations manifest as changes, they are actually absorbed by the system's inherent elasticity and should not all be counted as regulation costs or additional energy consumption. Directly including all fluctuations would overestimate the true dynamic energy consumption cost. Therefore, part S23, based on phase alignment and elastic window construction, strips away the normal fluctuation portion from the actually observed energy consumption data to identify the true abnormal additional energy consumption. This is the most critical decoupling step in the entire method. Specifically, this includes the phase-aligned energy consumption data sequence... Quasi-steady-state energy consumption curve fitted from historical data For reference, calculate the instantaneous deviation at each moment: This step determines whether the component's energy consumption is too high or too low at the current moment, and then determines whether it falls within the response elastic window. Determine whether energy consumption fluctuations can be absorbed autonomously by the components: like If the fluctuation is determined to be absorbable energy consumption fluctuation, it means that this is a normal fluctuation that the component can absorb on its own, and therefore it will no longer be included in the additional energy consumption statistics and will be removed. like The excess portion is extracted as the actual dynamic additional energy consumption and calculated as follows: ;in, This represents the first value obtained by fitting under the current load. Expected quasi-steady-state energy consumption of each component This represents the non-absorbable dynamic additional energy consumption retained after purification. In other words, it's the difference between the allowable fluctuation range and the absolute deviation; this is the final dynamic additional energy consumption remaining after purification.

[0046] Quasi-steady-state energy consumption curve fitted from historical data This serves as the fundamental baseline for the entire energy consumption stripping and performance evaluation process. It measures the expected energy consumption level of a component under ideal conditions without adjustment or abnormal disturbances. To fit this curve, targeted modeling based on quasi-steady-state operating conditions from historical data is required. Specifically: I. The core objective is to fit an ideal energy consumption-load relationship function, hoping to achieve this for each component. A function was found: ;in, For a moment Time Quasi-steady-state energy consumption of each component For the unit's output (actual load), Other operating parameters that affect energy consumption include main steam temperature, main steam pressure, and ambient temperature. This is the fitted multivariate relationship model.

[0047] II. Fitting the quasi-steady-state energy consumption curve includes: 1. Select data segments that meet the following conditions from the long-term collected unit operation database: The load remains stable: the rate of load change is low within the selected time window; Small energy consumption fluctuations: The energy consumption curve is stable, with no obvious sudden increases or decreases; Main steam parameters are stable: key parameters such as main steam pressure and temperature do not fluctuate drastically; No adjustment event flag: The scheduling system has not issued frequent load increase / decrease commands.

[0048] These segments can be considered as quasi-steady-state segments, representing samples that are closest to ideal operating conditions in actual operation.

[0049] 2. Construct the fitted dataset: For each quasi-steady-state segment, extract: Average load during this period ; Average component energy consumption during this period ; The average values ​​of operating condition variables such as main steam temperature and ambient temperature during this period are compiled into structured sample points: This forms a training set that can be used for regression modeling.

[0050] 3. Based on the data characteristics, multinomial regression is used to approximate nonlinear trends. The modeling objective is to minimize the residual between the predicted and actual energy consumption values. ; 4. Finally, a quasi-steady-state energy consumption prediction model for each component is formed. In practical analysis, given the output and operating condition data at a certain moment, the model can be called to output the expected energy consumption value. This value is used as a reference baseline under ideal conditions, and will be compared with the actual value to identify whether there is any abnormal additional energy consumption.

[0051] S24 integrates the phase adjustment function and response elastic window mechanism described above to construct the final time-sharing energy consumption decoupling fitting framework. The input to this framework is the cross-component time-sharing energy consumption response trajectory constructed in S1. After processing in S21 to S23, the following two key results are output: 1. Quasi-steady-state energy consumption sequence This indicates the expected energy consumption of a component under ideal steady-state operation at the current load level. 2. Dynamic Additional Energy Consumption Sequence This represents the actual additional energy expenditure caused by load regulation that cannot be buffered by the component itself.

[0052] When thermal power units adjust their load, their energy consumption changes are affected by multiple factors, including not only the energy consumption that should be consumed under normal operation (quasi-steady-state energy consumption) but also the additional expenses caused by the adjustment (dynamic additional energy consumption). If these two parts are analyzed together without distinction, it is easy to misjudge the energy efficiency level and make it difficult to identify abnormal segments with high adjustment costs. Therefore, after completing the data cleaning and decomposition preparation in the previous steps, S24 integrates these steps into a complete framework system and formally outputs the two types of energy consumption data.

[0053] S3, Perform output performance evaluation under phase purification: Fit the quasi-steady-state energy consumption and dynamic additional energy consumption to the time-segmented output data respectively to generate the output performance curve after phase purification, and identify the abnormal response stage with high adjustment cost based on the proportion of the two types of energy consumption and the performance deviation characteristics, and output a comprehensive evaluation result including performance level, response burden ratio and energy consumption deviation source.

[0054] S31, generating the output performance curve after phase purification using dual-line fitting: S311, Quasi-steady-state performance curve fitting: The total quasi-steady-state energy consumption output by S2 for each time period. Compared with the average actual load of the corresponding period The dataset was used to fit the data, and linear or nonlinear regression was performed using the least squares method to fit the following performance benchmark curve: ;in, For quasi-steady-state energy consumption, This is the quasi-steady-state energy consumption function corresponding to the output. This is the fitting residual term.

[0055] This section constructs a benchmark curve describing the ideal operating performance of thermal power units, also known as the quasi-steady-state performance curve, based on the quasi-steady-state energy consumption under different output levels. This curve clearly reflects how much energy the unit should consume to generate 1MW of electricity under conditions of no regulation interference and normal energy consumption fluctuations. The goal here is to fit a functional relationship. This curve describes the regular relationship between average load and quasi-steady-state energy consumption. It represents the baseline energy efficiency level of thermal power units and serves as a reference standard for subsequently identifying adjustment anomalies and performance deviations. To establish this curve, we first need to construct a sample set, where each sample point contains two types of information: First, the average actual load : indicates the first The average output of the unit within each time period comes from the cross-component time-sharing energy consumption response trajectory in S2.

[0056] Second, quasi-steady-state energy consumption : This represents the ideal energy consumption value that should exist under this output, predicted based on the fitting method within the same time period. It is also one of the output results of S2.

[0057] In this way, we obtained several data points, which constitute a two-dimensional dataset of load and energy consumption.

[0058] The least squares fitting model is as follows: ;in, The slope represents the change in energy consumption caused by a unit change in load. The intercept reflects fixed losses (standby and auxiliary equipment consumption). For the first The fitting error for each sample point is calculated by minimizing the sum of squared errors. , The optimal value: This is a typical least squares linear regression, which is a linear fitting form.

[0059] If the energy consumption-load relationship exhibits a curved trend (e.g., rising initially and then slowing down), a polynomial nonlinear fitting method, such as quadratic regression, can be used. The fitting objective remains to minimize the sum of squared residuals.

[0060] After fitting, the resulting function It is used to predict ideal energy consumption at any load point and serves as a benchmark curve to compare with actual energy consumption and determine whether there is any additional consumption.

[0061] S312, Dynamic Additional Energy Consumption Envelope Fitting: This involves fitting the total dynamic additional energy consumption for each time period. With the corresponding average actual load Group By constructing a scatter set and fitting it with regression or piecewise upper envelope, a dynamic additional energy consumption envelope is generated: ;in, To dynamically adjust the envelope function of the cost as a function of output, This is the fitting residual.

[0062] Based on the dynamic additional energy consumption obtained for each time period, S312 further analyzes the relationship between these dynamic energy consumptions and output, and constructs a representative curve, the dynamic additional energy consumption envelope, through fitting. This curve, together with the quasi-steady-state performance benchmark curve, constitutes a complete output performance curve system after phase purification. Dynamic additional energy consumption refers to the extra energy consumed by components in response to load changes during each adjustment process when they cannot be absorbed by their own elasticity. It reflects the system's adjustment cost or operating efficiency loss. However, the actual measured dynamic additional energy consumption will vary at different output levels, appearing as a discrete scattering. In order to extract the overall pattern, we hope to establish a trend curve through fitting to characterize the upper limit or typical value of dynamic additional energy consumption at a certain output level. Traditional regression fitting seeks a mean trend line, but the distribution of dynamic additional energy consumption often deviates from symmetry and exhibits an upward surge, making it unsuitable for simple mean fitting. Therefore, envelope fitting is more appropriate. Its goal is to find a smooth curve that passes through the upper edge of the data points. This curve represents the upper limit cost of typical maximum regulation energy consumption or regulation response under different output levels, making it easier to identify high-risk and high-burden areas.

[0063] Constructing the fitted dataset: Horizontal axis: Average actual load for each time period ; Vertical axis: Total dynamic additional energy consumption for this time period ; After summarizing all the data from different time periods, a two-dimensional scatter plot is generated, which serves as the fitted dataset.

[0064] The fitting process aims to obtain a function. , so that: ;in, The trend curve of dynamic additional energy consumption as a function of power output. The residual represents the fitting error, ideally under the following conditions. That is, the scatter points are not lower than the fitted line (upper envelope). This process can be achieved by extracting local extrema and fitting a smooth curve.

[0065] This curve reflects the worst-case energy consumption level of the unit under different output conditions, and can be used to identify sensitive operating areas with high output and susceptibility to anomalies.

[0066] S313, the two curves mentioned above together constitute the output performance curve system after phase purification, which respectively characterize the ideal operating baseline and the actual regulation response burden.

[0067] S32, Identifying Abnormal Response Stages Based on Energy Consumption Ratio and Deviation: The aim is to identify periods of abnormally high regulation costs during the operation of thermal power units, i.e., the so-called abnormal response stages. To ensure the identification process has a quantitative basis and objectivity, the scheme proposes to use two specific evaluation indicators: dynamic additional energy consumption ratio and performance deviation rate. These two indicators reflect the degree of system regulation burden from different dimensions, and are ultimately identified through a dual threshold judgment method. Each of these indicators has its own emphasis: the dynamic additional energy consumption ratio emphasizes structural bias, i.e., whether the additional component accounts for an abnormally high proportion of energy consumption during this period; the performance deviation rate emphasizes absolute numerical deviation, i.e., whether the energy consumption during this period is far higher than normal consumption. Only when both exceed the standard is it considered a true abnormal response stage, indicating not only a high additional cost ratio but also a large total energy consumption deviation, thus avoiding misjudgment.

[0068] S321, Calculation of Dynamic Additional Energy Consumption Ratio: This indicator represents the proportion of energy consumption in a given time period that is additional consumption resulting from the dynamic adjustment process. The numerator represents the dynamic additional energy consumption for that time period; the denominator represents the total energy consumption for that time period (quasi-steady-state energy consumption + dynamic additional energy consumption). The larger this ratio, the higher the adjustment cost and the further the operation deviates from the ideal operating condition. This represents the total dynamic additional energy consumption for a specific time period. This represents the total quasi-steady-state energy consumption during that time period. This indicates the proportion of dynamic additional energy consumption in total energy consumption.

[0069] S322, Performance Deviation Rate Calculation: This indicator shows how much the actual total energy consumption during a given period is higher than the ideal energy consumption value. ;in, This represents the actual total energy consumption during a specific time period, i.e. , The performance deviation rate is used to quantify the degree to which the actual energy consumption of the unit deviates from the ideal value. D>0 indicates that there is additional consumption, and the larger D is, the more serious the deviation.

[0070] S323, Abnormal Response Phase Identification Criteria: Setting Dual Judgment Thresholds , When the following conditions are met: If the current time period is determined to be an abnormal response phase, it indicates that the unit's regulation costs have increased abnormally or its operating efficiency has deteriorated significantly during that period.

[0071] The value range is [0.10, 0.25], with a default value of 0.15. This means that the additional energy consumption accounts for more than 15% of the total energy consumption. For most stably operating units, the dynamic additional energy consumption generally only accounts for 5% to 10% of the total energy consumption. A value exceeding 15% usually indicates frequent adjustments, poor control system efficiency, or difficulty in absorbing fluctuations. Range of values ,default value This means that the actual energy consumption is more than 8% higher than the benchmark energy consumption. According to the statistics of historical steady-state data, the energy consumption deviation of most qualified operating sections does not exceed 5%, and a deviation of more than 8% can be regarded as a relatively serious performance deviation.

[0072] S33. After completing the calculation of the output performance evaluation data (performance deviation rate), quantitative classification and attribution analysis are performed on the evaluation results of the entire operation cycle. In the previous analysis, the performance deviation rate D of each time period has been calculated, that is, the degree to which the actual energy consumption of each operation period is higher than the quasi-steady-state expected energy consumption.

[0073] Based on the above fitting and recognition results, the following three performance evaluation metrics are output: S331, Performance Level: Based on performance deviation rate across all time periods throughout the entire cycle. Mean performance deviation rate Define the rules for classifying levels: excellent: ; qualified: ; Warning: ; bad: .

[0074] in, The preset grading threshold reflects the deviation rate. The tolerance level has the following values: ; ; ; Based on the operating characteristics of typical thermal power units, when the load is stable, the performance deviation rate mostly remains between 3% and 7%. A deviation rate exceeding 10% reflects adjustment mismatch, equipment response lag, or inappropriate control strategies. A deviation rate exceeding 15% indicates a significant abnormal operating condition, which may be related to frequent start-stop operations, drastic load fluctuations, etc. Therefore, setting the level thresholds to 5%, 10%, and 15% can effectively classify operating quality, reasonably reflect normal fluctuations, and provide timely warnings of potential problems.

[0075] S332, Response Burst Ratio: This value reflects the proportion of dynamic energy consumption over the entire cycle, quantifying the overall system regulation cost burden. It represents the proportion of dynamic additional energy consumption caused by regulation to the total quasi-steady-state energy consumption over the entire cycle. A high value indicates frequent regulation and significant additional losses during unit operation; it can be used for horizontal comparison of the regulation burden of different units or different operating stages.

[0076] S333, Energy Consumption Deviation Source Analysis: In all time periods identified as abnormal response phases, the dynamic additional energy consumption collected during the abnormal response phase is classified and statistically analyzed according to three systems: boiler system, turbine system, and auxiliary equipment system. The contribution rate of each system is calculated, that is, the contribution ratio of the three major systems to the total dynamic additional energy consumption is calculated. ;in, This indicates the various systems, including the boiler system, turbine system, and auxiliary systems, used to pinpoint the main sources of performance degradation. It allows for a clear identification of which system is the primary source of performance degradation, aiding in locating the problematic system and developing targeted optimization strategies.

[0077] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0078] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for evaluating the output performance of thermal power units based on time-of-use energy consumption fitting, characterized in that, Includes the following steps: S1 divides the operating cycle of the thermal power unit into multiple time periods, collects independent energy consumption data of the boiler system, turbine system and main auxiliary system in each time period, and simultaneously collects the actual load, calculates the response time delay between the energy consumption change and load change of each component, and forms a cross-component time-sharing energy consumption response trajectory. S2, based on the cross-component time-sharing energy consumption response trajectory, performs phase alignment on the energy consumption data of each component and redistributes the lagging energy consumption to the actual adjustment period; after phase alignment, constructs a component response elastic window according to the local sensitivity of each component's energy consumption to load changes, elastically strips the energy consumption fluctuations that can be absorbed by the component itself, retains only the real dynamic adjustment energy consumption that cannot be absorbed, and establishes a time-sharing energy consumption decoupling fitting framework including the phase adjustment function and the response elastic window, outputting the quasi-steady-state energy consumption and the purified dynamic additional energy consumption; The construction of the response elastic window includes: based on the phase-aligned data, calculating the local sensitivity of each component's energy consumption to load changes, wherein the local sensitivity is the ratio of the component's energy consumption change to the load change within a preset short-term load fluctuation range; Based on historical operating data, the normal distribution range of energy consumption fluctuations described by local sensitivity of each component under the same load point is statistically analyzed, and the normal distribution range is defined as the response elasticity window of the component. S3, the quasi-steady-state energy consumption and dynamic additional energy consumption are fitted with the time-segmented output data respectively to generate the output performance curve after phase purification, and the abnormal response stage with high adjustment cost is identified based on the proportion of the two types of energy consumption and the performance deviation characteristics, and a comprehensive evaluation result including performance level, response burden ratio and energy consumption deviation source is output.

2. The method for evaluating the output performance of thermal power units based on time-sharing energy consumption fitting according to claim 1, characterized in that, The time-segmentation is based on event-triggered variable duration, including dividing the operating cycle into multiple time segments with significant change points in the load commands received by the thermal power unit as boundaries; wherein, the significant change point is the moment when the change amplitude of the load command exceeds the load change amplitude threshold or the change rate exceeds the load change rate threshold, so that each time segment represents a complete load adjustment event.

3. The method for evaluating the output performance of thermal power units based on time-sharing energy consumption fitting according to claim 1, characterized in that, Within each of the aforementioned time periods, key energy consumption data are collected synchronously: Coal consumption, air volume, and flue gas temperature of the boiler system; Internal efficiency of the steam turbine system and extraction parameters at each stage; In addition, the power consumption of the main auxiliary systems, including the feed water pump, condensate pump, and forced draft fan; As independent energy consumption data for each system; Simultaneously collect the actual load generated by the unit; The response time delay is calculated based on cross-correlation analysis. Specifically, for each load adjustment event, the actual load data is used as the baseline sequence, and the key energy consumption data of each component is collected as the comparison sequence. The time shift when the cross-correlation function reaches its maximum value is calculated by the sliding window cross-correlation analysis method. The time shift is determined as the response time delay of the corresponding component in the event of energy consumption change relative to load change.

4. The method for evaluating the output performance of thermal power units based on time-sharing energy consumption fitting according to claim 3, characterized in that, S1 further includes integrating the start time, end time, average load, independent energy consumption data sequence of each component, and calculated response delay of each time period into a cross-component time-sharing energy consumption response trajectory, which is used to describe the asynchronous energy consumption evolution law of each component being out of sync in time due to load adjustment within the time period.

5. The method for evaluating the output performance of thermal power units based on time-sharing energy consumption fitting according to claim 1, characterized in that, The phase alignment specifically includes: acquiring the cross-component time-sharing energy consumption response trajectory formed by S1; for each load adjustment event in the response trajectory, reading the response delay of each component calculated for the load adjustment event; based on the response delay, constructing a phase adjustment function, shifting the energy consumption data sequence of the current component in the time period forward on the time axis with the response delay as a single unit, generating a phase-aligned energy consumption data sequence, and redistributing the energy consumption caused by the response delay in time to the load adjustment period that actually caused the energy consumption.

6. The method for evaluating the output performance of thermal power units based on time-sharing energy consumption fitting according to claim 1, characterized in that, The elastic stripping includes: for the phase-aligned energy consumption data sequence, comparing its energy consumption value at each moment with the quasi-steady-state expected energy consumption fitted from historical data based on the load at that moment, calculating the instantaneous energy consumption deviation; if the instantaneous energy consumption deviation falls within the response elastic window, then the energy consumption fluctuation is determined to be a normal fluctuation that can be absorbed by the component itself, and it is stripped from the deviation; if the instantaneous energy consumption deviation exceeds the response elastic window, then the excess energy consumption is retained as the purified, real dynamic additional energy consumption that cannot be absorbed by the component.

7. The method for evaluating the output performance of thermal power units based on time-sharing energy consumption fitting according to claim 1, characterized in that, S2 further includes integrating the phase adjustment function and the response elastic window to establish a time-sharing energy consumption decoupling fitting framework. The input of the time-sharing energy consumption decoupling fitting framework is the original cross-component time-sharing energy consumption response trajectory. After processing, it finally outputs two results for each component within each time period, including: Quasi-steady-state expected energy consumption, as quasi-steady-state energy consumption; The dynamic additional energy consumption sequence obtained after purification by the flexible window is used as the purified dynamic additional energy consumption.

8. The method for evaluating the output performance of thermal power units based on time-sharing energy consumption fitting according to claim 1, characterized in that, S3 includes generating the output performance curve after phase purification through bilinear fitting, specifically including: Quasi-steady-state performance curve fitting: The quasi-steady-state energy consumption of each time period output by S2 is fitted with the average output data of the corresponding time period using the least squares method for linear or nonlinear regression fitting, generating a benchmark performance curve characterizing the energy consumption-output relationship of the unit under ideal steady-state conditions. Dynamic additional energy consumption-output relationship fitting: The dynamic additional energy consumption of each time period output by S2 is fitted with the average output data in the corresponding time period by scatter distribution to generate a dynamic additional energy consumption envelope that represents the dynamic adjustment cost under different output levels. The baseline performance curve and the dynamic additional energy consumption envelope together constitute the output performance curve after phase purification. The identification of the abnormal response phase includes: Calculate the dynamic additional energy consumption percentage for each time period; Calculate the performance deviation rate for each time period; When the dynamic additional energy consumption ratio and performance deviation rate of a certain time period both exceed the dual thresholds set according to the unit design and historical operation data, the current stage is determined to be an abnormal response stage.

9. The method for evaluating the output performance of thermal power units based on time-sharing energy consumption fitting according to claim 8, characterized in that, The generation of the comprehensive evaluation results includes: Performance rating: Based on the average value and dispersion of the performance deviation rate, the overall performance of the unit throughout the entire operating cycle is classified into excellent, qualified, warning, and poor levels; Response load ratio: defined as the ratio of total dynamic additional energy consumption to total quasi-steady-state energy consumption over the entire assessment period, used to quantify the overall regulation cost of the unit; Sources of energy consumption deviation: By analyzing the contribution ratio of the dynamic additional energy consumption of the three major systems of boiler, steam turbine and main auxiliary equipment to the total dynamic additional energy consumption during the abnormal response phase, the main source system leading to performance degradation is identified.

Images