Coal power unit dynamic energy efficiency evaluation method based on edge cloud collaborative size model

Abstract

The application discloses a coal power unit dynamic energy efficiency evaluation method based on an edge cloud cooperative size model, and relates to the technical field of energy efficiency evaluation. The method realizes intelligent optimization of load power through the cooperative operation of an edge side small model and a terminal large model. The small model collects historical operation parameters of the coal power unit, constructs load power, power generation and coal supply sequences, generates a coal conversion index curve and evaluates the steady-state coal conversion index under different loads, and then locks the optimal load power interval. The optimal interval of each unit is synchronized to the terminal large model in real time, the large model allocates the load according to the energy efficiency priority principle or adjusts it according to the margin, triggers manual confirmation if necessary, and finally distributes the regulation and control results to each small model for execution, so that the dynamic evaluation and cooperative optimization of the energy efficiency of the coal power unit are realized.

Description

Technical Field

[0001] This invention belongs to the field of energy efficiency assessment technology, specifically, it relates to a dynamic energy efficiency assessment method for coal-fired power units based on an edge cloud collaborative size model. Background Technology

[0002] The integration of edge computing and artificial intelligence technologies has provided a new paradigm for real-time data processing in industrial settings. The collaborative architecture of small and large models can leverage the advantages of low latency and high response of small models on the edge side, while also enabling global optimization decisions with the help of large models in the cloud.

[0003] Traditional methods typically rely on offline performance tests or periodic thermal tests to obtain the coal consumption characteristics of generating units. This static evaluation approach cannot capture the dynamic efficiency drift caused by factors such as equipment aging, coal quality fluctuations, and environmental changes during actual operation. It is also relatively crude in handling the timing matching of coal supply and power generation, often ignoring coal delivery delays or using simple averaging, resulting in a distortion of the coal-power conversion relationship. In addition, traditional evaluation methods have difficulty distinguishing between steady-state and non-steady-state operating conditions of the units, and are prone to misjudging transient efficiency anomalies during load fluctuations as performance degradation. Furthermore, traditional load allocation strategies of coal-fired power plants are often based on fixed unit efficiency ranking or proportional allocation principles, ignoring the differences in the real-time optimal power range of each unit, and lacking the rapid response capability of edge and cloud collaboration. They are highly dependent on human experience and lack clear incremental adjustment logic when the total demand power exceeds the unit's optimal capacity.

[0004] To address the aforementioned issues, this invention proposes a dynamic energy efficiency evaluation method for coal-fired power units based on an edge cloud collaborative size model. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a dynamic energy efficiency assessment method for coal-fired power units based on an edge cloud collaborative size model, solving the problems of lagging energy efficiency assessment and coarse load allocation in existing technologies.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A dynamic energy efficiency evaluation method for coal-fired power units based on an edge cloud collaborative size model, the method comprising:

[0008] Step 1: Using small models deployed at the edge, the historical operating parameters collected during any historical normal operation of the coal-fired power unit are discretized and sorted to form the load power sequence of the coal-fired power unit and its aligned associated power generation sequence, and the coal supply scheduling list is locked synchronously.

[0009] Step 2: Perform sampling time averaging operation on the coal feed scheduling list, lock the coal feed sequence that is aligned with the sampling time of the power generation sequence, and generate the coal-power conversion index curve associated with the coal-fired power unit based on the numerical relationship between the coal feed and power generation at the same moment.

[0010] Step 3: Evaluate the steady-state coal-power conversion index associated with the coal-fired power unit under the current load power sequence state based on the fluctuation trend and numerical performance of the coal-fired power conversion index curve.

[0011] Further determine the steady-state coal-power conversion index value of coal-fired power units under different load power demands, reverse evaluate the optimal load power range of coal-fired power units, and synchronize it to the terminal large model;

[0012] Step four involves extracting the optimal load power range for each coal-fired power unit in real time from each small model and transmitting it synchronously to the terminal large model. The load power of each coal-fired power unit is then optimized and controlled in conjunction with the total power demand of the coal-fired power plant.

[0013] As a further aspect of the present invention, in step one, the specific method for constructing the load power sequence of the coal-fired power unit and its aligned and associated power generation sequence is as follows:

[0014] The coal-fired power plant is identified and denoted as Q;

[0015] Determine the total number of coal-fired power units in the coal-fired power plant, denoted as j, and randomly arrange them to obtain the sequence of coal-fired power units M1, M2, ..., Mj;

[0016] Based on the small model deployed at the edge of any coal-fired power unit Mi, the historical operating parameters of coal-fired power unit Mi during any historical normal operation are collected, including the power generation and load power at each moment during the historical normal operation.

[0017] The historical normal operation process is defined as the start time t_str of the coal-fired power unit Mi starting to generate electricity, and the end time t_end of the historical normal operation process is the time when the power generation ends or the load power is adjusted.

[0018] Count the total number of times from t_str to t_end, and denote it as n;

[0019] Sort the power generation corresponding to n time points in chronological order to obtain the power generation sequence D1, D2, ..., Dn, where D1=0, D2 is the total power generation from the start time t_str to the next time point, and so on.

[0020] The load power sequence D1, D2, ..., Dn is determined synchronously in the manner described above.

[0021] As a further aspect of the present invention, the specific method for synchronously locking the coal feed scheduling list in step one is as follows:

[0022] Extract the coal quantity scheduling list that was pre-set by the operator during the historical normal operation of the coal-fired power unit Mi;

[0023] The coal quantity scheduling list includes the coal quantity input to coal-fired power unit Mi and its corresponding time for n time periods from t_str to t_end. If no coal quantity is input to coal-fired power unit Mi at any time, the coal quantity at the corresponding time is recorded as 0.

[0024] As a further aspect of the present invention, in step two, the sampling time averaging operation is performed on the coal feed scheduling list, and the specific method for locking the coal feed sequence that is aligned with the sampling time of the power generation sequence is as follows:

[0025] Extract the n time points corresponding to the power generation sequence D1, D2, ..., Dn, and discretize them into a time sequence t1, t2, ..., tn, where t1 = t_str and tn = t_end;

[0026] Construct a time axis corresponding to the time sequence t1, t2, ..., tn, and mark the coal quantity and corresponding time of all input coal-fired power units Mi in the coal supply scheduling list on the time axis. Determine the coal quantity G1 of the first input coal-fired power unit Mi and the coal quantity G2 of the second input coal-fired power unit, as well as the occurrence times t1', t2' of the two input coal-fired power unit coal quantity events and the time interval T(t1', t2') from t1' to t2'. Divide the coal quantity G1 by the time interval T(t1', t2') and record the calculation result as the coal supply quantity of coal-fired power unit Mi at each time within the time interval T(t1', t2').

[0027] Similarly, by determining the coal feed rate of coal-fired power unit Mi at n times from t_str to t_end, we obtain the coal feed rate sequence H1, H2, ..., Hn.

[0028] As a further aspect of the present invention, in step two, the specific method for generating the coal-fired power conversion index curve associated with the coal-fired power unit based on the numerical relationship between the coal feed rate and the power generation at the same moment is as follows:

[0029] The coal feed sequence H1, H2, ..., Hn and the power generation sequence D1, D2, ..., D are standardized in grams and kilowatt-hours, respectively.

[0030] The coal-fired power conversion index R1 is calculated using H1 / D1=R1. The smaller the value of R1, the higher the energy conversion efficiency of the coal-fired power unit Mi, and vice versa.

[0031] Similarly, determine the coal-to-electricity conversion index R1 at n time points, and combine them in chronological order to form the coal-to-electricity conversion index sequence R1, R2, ..., Rn;

[0032] In the two-dimensional coordinate system XY1, curve fitting is performed based on the coal-to-electricity conversion index sequence R1, R2, ..., Rn to obtain the coal-to-electricity conversion index curve Smd, where the horizontal axis of XY1 is time and the vertical axis is the coal-to-electricity conversion index.

[0033] As a further aspect of the present invention, the specific method for evaluating the steady-state coal-fired power conversion index associated with the coal-fired power unit under the current load power sequence state in step three is as follows:

[0034] Extract the preset sliding window length L, where 1≤L≤n;

[0035] The coal-fired power conversion index curve Smd is divided into windows by combining the sliding window length L and the preset step size s, where 1≤s≤L;

[0036] The total number of statistical windows is denoted as o;

[0037] Determine the standard deviation σk and mean μk of the coal-to-electricity conversion index within any window k;

[0038] If σk < ε, the running state of window k is determined to be steady state, and ε is a preset stability threshold; otherwise, the running state of window k is determined to be unsteady state, and the processing of window k is skipped.

[0039] For a steady-state window k, its mean μk is marked as a candidate steady-state coal-fired power conversion index, and the average load power D_k_avg within the window k is locked based on the load power sequence D1, D2, ..., Dn;

[0040] The average load power D_k_avg is combined with the candidate steady-state coal-fired power conversion index to form a candidate steady-state point pair;

[0041] Traverse o windows and extract all candidate steady-state point pairs within each window, denoted as the candidate steady-state point pair set;

[0042] Extract the target load power D_tar allocated by the operator to the coal-fired power unit Mi within the period from t_str to t_end;

[0043] For all candidate steady-state point pairs, the arithmetic mean of the candidate steady-state coal-fired power conversion index is taken as the steady-state coal-fired power conversion index of coal-fired power unit Mi under the target load power D_tar, and is denoted as Ri_D_tar.

[0044] As a further aspect of the present invention, in step three, the specific method for further locking the steady-state coal-fired power conversion index value of the coal-fired power unit under different load power demands, and for reverse evaluation of the preferred load power of the coal-fired power unit, and synchronizing it to the terminal large model, is as follows:

[0045] The steady-state coal-fired power conversion index associated with coal-fired power unit Mi under different target load power is determined based on the historical operating parameters collected during the historical normal operation of coal-fired power unit Mi.

[0046] The target load power is sorted from smallest to largest as the target load power sequence D_tar1, D_tar2, ..., D_tarK, where K is the total number of target load power. Similarly, the steady-state coal-fired power conversion index sequence Ri_D_tar1, Ri_D_tar2, ..., Ri_D_tarK is obtained.

[0047] A two-dimensional coordinate system XY2 is constructed with average load power as the horizontal axis and steady-state coal-fired power conversion index as the vertical axis;

[0048] Based on the target load power sequence and the steady-state coal-fired power conversion index sequence, the function curve EW of the steady-state coal-fired power conversion index as a function of the target load power was plotted.

[0049] The extreme value search of the steady-state coal-to-electricity conversion index is performed on the function curve EW within the rated load power range of the coal-fired power unit Mi. The minimum steady-state coal-to-electricity conversion index is marked, and the target load power corresponding to the minimum steady-state coal-to-electricity conversion index is locked and marked as the preferred load power of the coal-fired power unit Mi.

[0050] Then, using the minimum steady-state coal-fired power conversion index as a benchmark, extract the v target load powers that are closest to the minimum steady-state coal-fired power conversion index value, and combine them with the preferred load power D_gd to form the preferred load power range Mi for coal-fired power units, which is then synchronized to the terminal large model.

[0051] As a further aspect of the present invention, in step three, the preferred load power and preferred load power range of the coal-fired power unit Mi are updated once every time the update cycle preset by the operator, wherein the update operation is performed based on the latest historical operating parameters of the coal-fired power unit Mi.

[0052] As a further aspect of the present invention, the specific method for optimizing and controlling the load power of each coal-fired power unit in step four, based on the current total power demand of the coal-fired power plant, is as follows:

[0053] Based on the j small models equipped with j coal-fired power units, the current preferred load power range of the j coal-fired power units is transmitted to the terminal large model.

[0054] The total power demand Pall of coal-fired power plants is extracted in real time from the terminal large model;

[0055] Based on the preferred load power range of each of the j coal-fired power units, the preferred load power of the j coal-fired power units is determined, and the j coal-fired power units are arranged from largest to smallest according to their preferred load power to obtain the sequence of units to be allocated;

[0056] If the sum of the preferred load power of j coal-fired power units is greater than or equal to the total demand power Pall, the total demand power Pall is allocated to the j coal-fired power units in the order of the sequence of units to be allocated.

[0057] Conversely, extract the sequence of units to be allocated, and sort the sequence of units to be allocated from largest to smallest according to the difference between the preferred load power of each coal-fired power unit and the maximum load power in the preferred load power range, to obtain the secondary sequence of units to be allocated;

[0058] The load power of the coal-fired power units in the secondary allocation sequence is adjusted sequentially to the maximum load power in their respective preferred load power ranges until the total demand power Pall is absorbed.

[0059] If the total demand power Pall still cannot be absorbed, a manual confirmation instruction is generated. Based on the above adjustment operations, the operator adaptively adjusts the load power of each coal-fired power unit to absorb the total demand power Pall. The adjustment results for all coal-fired power units are then distributed to the small model through the large model, and the small model optimizes and controls each coal-fired power unit.

[0060] The beneficial effects of this invention are:

[0061] (1) This invention discretizes and aligns the historical operating parameters of coal-fired power units through small models deployed at the edge, generates coal-fired power conversion index curves, dynamically evaluates the steady-state coal-fired power conversion efficiency of the units under different loads, and then identifies the optimal load power range in reverse, reducing the dependence of traditional evaluation methods on expert experience and improving the objectivity and real-time performance of energy efficiency evaluation; the terminal large model gathers the optimal ranges of each unit and performs global optimization and control in combination with the total power demand of the power plant, realizing intelligent allocation of coal-fired power unit load, significantly reducing unnecessary coal consumption and improving the overall power generation economy; the edge and cloud collaborative architecture not only ensures low latency and localization of data processing, but also gives full play to the ability of global optimization, which helps to reduce carbon emissions and enhance the power plant's adaptive adjustment capability to load fluctuations;

[0062] (2) This invention achieves precise synchronization and fusion of multi-source heterogeneous data by constructing a time series that strictly aligns the load power, power generation and coal feed of coal-fired power units. It transforms the originally discrete and discontinuous coal feed scheduling events into a continuous coal feed sequence that is point-by-point aligned with the power generation sequence through the sampling time averaging operation, eliminating data deviation caused by inconsistent sampling frequencies. At the same time, the small model based on edge deployment automatically collects historical normal operation parameters, ensuring the authenticity of the data source and the degree of automation. It provides a data foundation for subsequent coal consumption characteristic analysis, load response modeling and coal feed strategy optimization.

[0063] (3) By constructing a coal-power conversion index curve and introducing a sliding window steady-state determination, this invention improves the accuracy and anti-interference capability of coal-fired power unit energy efficiency assessment. It can automatically eliminate abnormal data under load fluctuations or non-steady-state conditions, ensuring the reliability of candidate steady-state point pairs, thus providing a solid data foundation for subsequent load optimization. On this basis, by mapping the function curve of steady-state coal-power conversion index with load change and accurately searching for extreme values, the optimal load power and its efficient operating range that minimizes unit coal consumption can be quickly locked, directly reducing coal consumption per unit of power generation and achieving significant energy-saving and cost-reduction benefits. At the same time, the optimization results are synchronized to the terminal large model and updated regularly, so that the recommended strategy can adapt to equipment aging or changes in operating conditions, forming a closed-loop optimization, improving the economic operation level and intelligent decision-making capability of coal-fired power units.

[0064] (4) This invention achieves intelligent optimization allocation of load power in coal-fired power plants through the collaboration of large and small terminal models. It prioritizes the allocation of load power from large to small, which can quickly match the total demand power, reduce redundant calculations, and improve scheduling efficiency. When the total power is insufficient, it sorts the load power in a second time and prioritizes adjusting the units to the maximum load, which can fully tap the potential of each unit and maximize the total output. If the demand still cannot be met, a manual confirmation instruction is introduced to ensure the reliability in extreme cases. The manual adjustment results are fed back to the small model for learning, forming a closed-loop optimization and continuously improving the model's adaptability. The whole process reduces repeated manual trial and error, reduces the burden on operators, and ensures the accuracy and economy of power consumption, effectively balancing the flexibility of automation and manual intervention. Attached Figure Description

[0065] The invention will now be further described with reference to the accompanying drawings.

[0066] Figure 1 This is a flowchart illustrating the method described in this invention. Detailed Implementation

[0067] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0068] like Figure 1 As shown, this application provides a dynamic energy efficiency evaluation method for coal-fired power units based on an edge cloud collaborative large-scale model. Its overall architecture consists of edge computing devices deployed on each coal-fired power unit side, a built-in small model, and a terminal large model located in the power plant control center or group cloud.

[0069] Edge computing devices can adopt industrial smart gateways based on ARM architecture, on which lightweight machine learning models run to complete local data preprocessing, feature extraction and energy efficiency index calculation; the terminal large model is deployed on a private cloud server equipped with a GPU cluster, running a time-series decision model based on the Transformer architecture, responsible for global load optimization and control.

[0070] The edge device and the terminal model achieve bidirectional data synchronization through industrial Ethernet or 5G network modules, and the transmission protocol can be MQTT or OPCUA.

[0071] As an embodiment 1 of this application, it specifically includes:

[0072] Step 1: Using small models deployed at the edge, the historical operating parameters collected during any historical normal operation of the coal-fired power unit are discretized and sorted to form the load power sequence of the coal-fired power unit and its aligned associated power generation sequence, and the coal supply scheduling list is locked synchronously.

[0073] Step 2: Perform sampling time averaging operation on the coal feed scheduling list, lock the coal feed sequence that is aligned with the sampling time of the power generation sequence, and generate the coal-power conversion index curve associated with the coal-fired power unit based on the numerical relationship between the coal feed and power generation at the same moment.

[0074] Step 3: Evaluate the steady-state coal-power conversion index associated with the coal-fired power unit under the current load power sequence state based on the fluctuation trend and numerical performance of the coal-fired power conversion index curve.

[0075] Further determine the steady-state coal-power conversion index value of coal-fired power units under different load power demands, reverse evaluate the optimal load power range of coal-fired power units, and synchronize it to the terminal large model;

[0076] Step four involves extracting the optimal load power range for each coal-fired power unit in real time from each small model and transmitting it synchronously to the terminal large model. The load power of each coal-fired power unit is then optimized and controlled in conjunction with the total power demand of the coal-fired power plant.

[0077] Example 2

[0078] This embodiment is based on Embodiment 1, and details the specific implementation process of Step 1 and Step 2, namely, how to construct the load power sequence and power generation sequence of the coal-fired power unit, lock the coal supply scheduling list, and generate the coal-power conversion index curve. The specific content is as follows:

[0079] Taking a coal-fired power plant Q as an example, there are j=4 coal-fired power units in the plant, which are randomly arranged and denoted as unit sequence M1, M2, M3, and M4. Each unit is equipped with the aforementioned edge computing device, which is connected to the unit's distributed control system (DCS) via an RS485 bus to collect operating parameters in real time or near real time.

[0080] In step one, for any coal-fired power unit Mi (i=1,2,3,4), the small model on the edge device extracts all operating parameters of Mi during a complete historical normal operation process.

[0081] The start time t_str of the historical normal operation process is defined as the moment when the unit begins to output electrical energy to the grid, and the end time t_end is defined as the moment when the unit is normally shut down or when the operator actively changes the load power setting value.

[0082] In this embodiment, the sampling period for a coal-fired power unit Mi during a continuous 48-hour operation is 1 second, and the total number of time points n=172800.

[0083] The small model sorts the power generation data at n time points in chronological order to obtain the power generation sequence E1, E2, ..., En, where E1=0, indicating that the cumulative power generation at the beginning time is zero, E2 is the power generation increment from t_str to the second sampling time, and so on.

[0084] Synchronously, according to the exact same time correspondence, the instantaneous load power values ​​collected at n times are arranged into a load power sequence D1, D2, ..., Dn, with the unit being megawatts.

[0085] The method for locking the coal feeding schedule list is as follows: the small model reads the coal feeding plan record pre-prepared by the operators during the historical process through the DCS interface. The record is stored in the form of a table, and each record contains the coal feeding time (accurate to the second) and the amount and quality of coal fed (in kilograms).

[0086] If no coal feeding action occurs at a certain sampling time, the coal quantity value at the corresponding time is recorded as 0. For example, 8000 kg of coal is fed in the 5th second after t_str, 5000 kg of coal is fed in the 7200th second, and so on, forming a complete coal feeding scheduling list.

[0087] Next, in step two, the sampling time averaging operation is performed to extract the n time points corresponding to the power generation sequence, forming the time sequence t1, t2, ..., tn, where t1 = t_str and tn = t_end.

[0088] Using the timeline as a reference, mark all coal feeding events and coal quantities in the coal feeding schedule list. Determine the coal quantity of the first coal feeding event as G1=8000kg, with the occurrence time t1'=t_str+5s; and the coal quantity of the second coal feeding event as G2=5000kg, with the occurrence time t2'=t_str+7200s.

[0089] The time interval T(t1',t2') between the two coal feeding events is calculated to be 7195 seconds. The coal quantity G1 is divided by this time interval to obtain the equivalent coal feeding rate during this period: 8000kg÷7195s≈1.1119kg / s, which is converted to 1111.9g / s. This value is used as the characteristic value of the coal feeding quantity for all sampling times within the time interval (i.e., every time from the 5th second to the 7200th second).

[0090] Similarly, each pair of adjacent coal feeding events is processed sequentially until all n time points are covered, finally obtaining a coal feeding sequence H1, H2, ..., Hn aligned with the time sequence, with the unit being grams per second.

[0091] Then, a coal-to-electricity conversion index curve is generated, and the units of the coal feed sequence and the power generation sequence are unified. The coal feed is kept in grams per second, and the power generation is unified in kilowatt-hours.

[0092] Define the coal-fired power conversion index Ri = Hi / Ei (Ri is empty when Ei = 0).

[0093] The physical meaning of Ri is the equivalent coal feeding rate corresponding to the generation of 1 kilowatt-hour of electricity. The smaller the value, the higher the energy conversion efficiency. By calculating Ri at each effective time point in sequence, the coal-to-electricity conversion index sequence R2, R3, ..., Rn is obtained.

[0094] A two-dimensional coordinate system XY1 is constructed with time on the horizontal axis and the coal-to-power conversion index on the vertical axis. Cubic spline interpolation is then performed in XY1 using the coal-to-power conversion index sequence R2, R3, ..., Rn to obtain a smooth coal-to-power conversion index curve Smd, which fully reflects the dynamic changes in energy conversion efficiency of coal-fired power units Mi throughout the entire historical operating cycle.

[0095] This embodiment completes data discretization, time alignment, and coal feed averaging in-situ using a small edge model, eliminating energy efficiency calculation deviations caused by inconsistencies between coal feed time and power generation sampling time. It directly generates the coal-to-power conversion index curve using the point-by-point ratio of the coal feed sequence and the power generation sequence, without relying on complex thermodynamic modeling. This allows for an intuitive and continuous depiction of the unit's real-time energy efficiency fluctuations, laying a precise data foundation for subsequent steady-state analysis and optimized scheduling.

[0096] Example 3

[0097] Based on the coal-to-power conversion index curve Smd obtained in Example 2, this embodiment focuses on implementing the specific method for evaluating the steady-state coal-to-power conversion index associated with the coal-fired power unit under the current load power sequence state in step three of Example 1, as follows:

[0098] It should be noted that the edge device mini-model has pre-set configurable parameters, including a sliding window length L=30 (satisfying 1≤L≤n), a sliding step size s=15, a stability threshold ε=0.018, and a load tolerance δ=4MW. Operators can modify these parameters through the edge device's management interface.

[0099] For the first moment on the coal-electricity conversion index curve Smd, the small model slides along the time axis with a window length L=30 and a step size s=15, dividing the time into a total of o windows.

[0100] For each window k, extract 30 coal-electricity conversion index values ​​within the window and calculate their arithmetic mean μk and standard deviation σk.

[0101] The standard deviation calculation formula adopts the form of the overall standard deviation. If σk < ε, it is determined that the unit is in steady-state operation within the time period corresponding to window k. The average value μk of this window is marked as the candidate steady-state coal-fired power conversion index.

[0102] Meanwhile, load power values ​​corresponding to 30 time points covered by window k are extracted from the load power sequence D1, D2, ..., Dn, and their arithmetic mean is calculated and denoted as the average load power P_k_avg. (P_k_avg, μk) are combined into a candidate steady-state point pair. If σk≥ε, it is determined to be an unsteady window and is not processed.

[0103] After traversing all o windows, all candidate steady-state point pairs are collected to form a candidate steady-state point pair set. The candidate steady-state point pair set represents the coal-fired power conversion index characteristics of the corresponding coal-fired power unit under different load power levels during stable operation.

[0104] Extract the target load power D_tar set by the operator for any coal-fired power unit Mi within the range of t_str to t_end. In this embodiment, D_tar = 350MW.

[0105] From the candidate steady-state point pair set, those point pairs whose absolute value of the difference between the average load power P_k_avg and D_tar is less than the load tolerance δ=4MW are selected. The arithmetic mean of the candidate steady-state coal-fired power conversion index μk of all selected point pairs is the steady-state coal-fired power conversion index Mi of the coal-fired power unit at the target load power of 350MW, denoted as Ri_D_tar. The small model reports the calculation result to the terminal large model through the communication module for further analysis.

[0106] This embodiment automatically distinguishes between steady-state and non-steady-state operating conditions based on the standard deviation analysis of a sliding window, eliminating the interference of transient processes with varying loads on energy efficiency assessment. This makes the extraction of the steady-state coal-fired power conversion index entirely data-driven, without the need for manual labeling. By aggregating steady-state points of adjacent load conditions through a load tolerance mechanism, the statistical robustness of the index estimation is enhanced, and reliable energy efficiency characterization values ​​can still be obtained even when there is a small amount of noise in the historical data.

[0107] Example 4

[0108] This embodiment is implemented based on the above embodiment, and explains step three in embodiment 1. It discloses a method for further locking the steady-state coal-fired power conversion index of coal-fired power units under different load power demands, reversely evaluating and optimizing the load power range, and synchronizing it to the terminal large model. Specifically, it includes the following:

[0109] First, the process of Example 3 is repeated using the edge small model to calculate the corresponding steady-state coal-fired power conversion index for coal-fired power unit Mi under multiple different target load powers D_tar1, D_tar2, ..., D_tarK.

[0110] In this embodiment, the small model selects a range from the minimum technical output of the unit of 180MW to the rated power of 600MW, with a step size of 10MW and a total of K=43 target load power points, and obtains the steady-state coal-fired power conversion index sequence Ri_D_tar1,Ri_D_tar2,...,Ri_D_tar43 in sequence.

[0111] The target load power is sorted from smallest to largest and kept in correspondence with the exponential sequence.

[0112] A two-dimensional coordinate system XY2 is constructed with the average load power (i.e., the target load power, in MW) as the horizontal axis and the steady-state coal-fired power conversion index (in g / (kWh·s)) as the vertical axis. Using point plotting and piecewise linear interpolation methods, the function curve EW of the steady-state coal-fired power conversion index (the steady-state coal-fired power conversion index sequence Ri_D_tar1, Ri_D_tar2,..., Ri_D_tar43) as a function of the target load power is plotted.

[0113] Within the rated load power range [180MW, 600MW], a minimum value search is performed on the function curve EW with a search step size of 1MW. The vertical coordinate values ​​are compared point by point to find the globally minimum steady-state coal-fired power conversion index value.

[0114] The load power corresponding to the global minimum steady-state coal-fired power conversion index value is marked as the preferred load power D_gd of the coal-fired power unit Mi. For example, the search found that when the load power is 420MW, the steady-state coal-fired power conversion index reaches the minimum value of 285g / (kWh·s), then D_gd=420MW.

[0115] Subsequently, the small model uses the minimum steady-state coal-fired power conversion index as a benchmark, and extracts three points (excluding 420MW itself) with the smallest difference between the steady-state coal-fired power conversion index and 285 from the existing 43 target load power points. It is assumed that their corresponding load powers are 400MW, 440MW and 450MW respectively.

[0116] These load powers are combined with D_gd to form the preferred load power range [400MW, 450MW] for coal-fired power unit Mi. The small model immediately encapsulates the preferred load power range and D_gd value into a JSON message and synchronizes it to the terminal large model via the MQTT protocol.

[0117] It should be noted that operators can preset the update cycle for each unit through the large model interactive interface, for example, setting the update cycle to 720 hours.

[0118] When the timer inside the edge small model expires, it automatically extracts the most recent complete historical operation data from the DCS historical database, re-executes all the logic of steps one to three, generates a new preferred load power and preferred load power range, overwrites the original parameters, and resynchronizes to the large model. If the coal-fired power unit has undergone major repairs or the coal quality has changed significantly during this period, the operator can also manually trigger an instant update.

[0119] This embodiment constructs a load steady-state energy efficiency function and automatically seeks the extreme value to accurately lock the load operating point and recommended operating range with optimal energy efficiency for each unit, replacing the fixed load allocation mode based on experience; periodic model updates enable energy efficiency assessment to track the time-varying characteristics of unit equipment status and coal quality in real time, maintaining the accuracy and economy of load scheduling in the long term.

[0120] Example 5

[0121] This embodiment is implemented based on the above embodiment. Specifically, it explains the complete decision-making process in step four, where the terminal large model combines the current total power demand of the coal-fired power plant to optimize and control the load power of each coal-fired power unit, as follows:

[0122] It has been confirmed that all four generating units (j=4) within the current coal-fired power plant Q have completed the generation and synchronization of their optimal load power ranges, and the current range data for each unit is as follows:

[0123] M1 has a preferred load power range of [280MW, 330MW], with a preferred load power of 300MW;

[0124] M2 has a preferred load power range of [400MW, 460MW], with a preferred load power of 430MW;

[0125] M3 has a preferred load power range of [350MW, 400MW], with a preferred load power of 370MW;

[0126] M4 has a preferred load power range of [200MW, 250MW], with a preferred load power of 220MW.

[0127] The terminal large model obtains the current total power demand of coal-fired power plants in real time, Pall=1100MW, through the power dispatch data network.

[0128] The large model first extracts the preferred load power of each unit and sorts them from largest to smallest to form the sequence of units to be allocated: M2 (430MW), M3 (370MW), M1 (300MW), M4 (220MW). Then, the sum of the preferred load power is calculated: 430+370+300+220=1320MW, which is greater than Pall.

[0129] Therefore, the following allocations are made in sequence: M2 is allocated 430MW, with 670MW remaining; M3 is allocated 370MW, with 300MW remaining; M1 is allocated 300MW, with 0MW remaining; M4 is allocated 0MW, meaning M4 is idle, thus completing the regulation.

[0130] In another scenario, the total current power demand of the coal-fired power plant is Pall = 950MW, while the sum of the preferred load power is 940MW, which is insufficient to meet the demand.

[0131] The large model first sorts the units to be allocated in descending order of preferred load power, specifically: M2 (350), M3 (240), M1 (200), M4 (150).

[0132] Next, the difference between the preferred load power of each coal-fired power unit and the maximum load power of its respective range is calculated:

[0133] The difference in M2 is: 400-350=50MW;

[0134] The difference in M3 is: 280-240=40MW;

[0135] The difference in M1 is: 250-200=50MW;

[0136] The difference in M4 is: 180-150=30MW;

[0137] The units to be allocated are sorted in descending order of the calculated differences to form a secondary sequence: M2(50), M1(50), M3(40), M4(30). If the differences are the same, they can be further sorted by capacity.

[0138] The large model sequentially adjusts the coal-fired power units in the secondary allocation sequence to their maximum load power within their respective ranges. After each adjustment, it checks whether the total power reaches Pall. First, M2 is adjusted to 400MW. The cumulative allocated power is 400+240+200+150=990MW, which exceeds 950MW.

[0139] At this point, a pullback is needed within the M2 range: 990-950=40MW of excess capacity. M2 is reduced by 40MW from 400MW to 360MW, resulting in the final allocation: M2=360, M3=240, M1=200, M4=150, with a total power of 950MW. All values ​​are within their respective ranges, thus completing the regulation.

[0140] In the above scenario, if the total demand power Pall = 1100MW, after adjusting all units to the maximum value of the range, the total power is 400 + 280 + 250 + 180 = 1110MW, which is still greater than the demand. It can be adjusted back to meet the demand. If Pall = 1150MW, then the total power of all units operating at maximum power is 1110MW < 1150MW, which cannot be fully absorbed. At this time, the terminal model generates a manual confirmation instruction, prompting the operator on the interface: "The current total maximum output of all units is 1110MW, which cannot meet the total demand of 1150MW. Please manually increase the load of some units or purchase electricity from outside."

[0141] Based on this instruction, operators can temporarily increase the power of certain units to exceed the upper limit of the recommended range, such as overload operation, according to the actual safety boundary of the units, to complete the power absorption.

[0142] After the adjustments are completed, the large model will send the final load power setpoints of each unit to the corresponding edge small models via MQTT. The small models will then directly act on the DCS load control loop through the analog output module to achieve closed-loop control.

[0143] This embodiment utilizes a load allocation strategy that coordinates large and small models to automatically complete the economical allocation of plant-level loads while prioritizing the operation of each unit within its high-efficiency range, thereby maximizing overall power generation efficiency. When abnormal operating conditions such as power deficit occur, human decision-making is introduced in a timely manner. This approach leverages the advantages of automation while retaining the final say of operators in extreme situations, thus balancing economy and safety.

[0144] All data in the formulas described above have been calculated with dimensions removed. Furthermore, any content not described in detail in this specification is existing technology known to those skilled in the art.

[0145] The above description is merely an example and illustration of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

[0146] It should be stated that all user data collected in this application was collected with the user's consent and authorization. Furthermore, the uses of user data are legal and compliant, and the use and processing of user data comply with the relevant laws, regulations, and standards of the relevant regions.

Claims

1. A dynamic energy efficiency evaluation method for coal-fired power units based on an edge cloud collaborative size model, characterized in that, The method includes: Step 1: Using small models deployed at the edge, the historical operating parameters collected during any historical normal operation of the coal-fired power unit are discretized and sorted to form the load power sequence of the coal-fired power unit and its aligned associated power generation sequence, and the coal supply scheduling list is locked synchronously. Step 2: Perform sampling time averaging operation on the coal feed scheduling list, lock the coal feed sequence that is aligned with the sampling time of the power generation sequence, and generate the coal-power conversion index curve associated with the coal-fired power unit based on the numerical relationship between the coal feed and power generation at the same moment. Step 3: Evaluate the steady-state coal-power conversion index associated with the coal-fired power unit under the current load power sequence state based on the fluctuation trend and numerical performance of the coal-fired power conversion index curve. Further determine the steady-state coal-power conversion index value of coal-fired power units under different load power demands, reverse evaluate the optimal load power range of coal-fired power units, and synchronize it to the terminal large model; Step four involves extracting the optimal load power range for each coal-fired power unit in real time from each small model and transmitting it synchronously to the terminal large model. The load power of each coal-fired power unit is then optimized and controlled in conjunction with the total power demand of the coal-fired power plant.

2. The method according to claim 1, characterized in that, In step one, the specific method for constructing the load power sequence of the coal-fired power unit and its aligned power generation sequence is as follows: The coal-fired power plant is identified and denoted as Q; Determine the total number of coal-fired power units in the coal-fired power plant, denoted as j, and randomly arrange them to obtain the sequence of coal-fired power units M1, M2, ..., Mj; Based on the small model deployed at the edge of any coal-fired power unit Mi, the historical operating parameters of coal-fired power unit Mi during any historical normal operation are collected, including the power generation and load power at each moment during the historical normal operation. The historical normal operation process is defined as the start time t_str of the coal-fired power unit Mi starting to generate electricity, and the end time t_end of the historical normal operation process is the time when the power generation ends or the load power is adjusted. Count the total number of times from t_str to t_end, and denote it as n; Sort the power generation corresponding to n time points in chronological order to obtain the power generation sequence D1, D2, ..., Dn, where D1=0, D2 is the total power generation from the start time t_str to the next time point, and so on. The load power sequence D1, D2, ..., Dn is determined synchronously in the manner described above.

3. The method according to claim 2, characterized in that, In step one, the specific method for synchronously locking the coal feed scheduling list is as follows: Extract the coal quantity scheduling list that was pre-set by the operator during the historical normal operation of the coal-fired power unit Mi; The coal quantity scheduling list includes the coal quantity input to coal-fired power unit Mi and its corresponding time for n time periods from t_str to t_end. If no coal quantity is input to coal-fired power unit Mi at any time, the coal quantity at the corresponding time is recorded as 0.

4. The method according to claim 3, characterized in that, In step two, the sampling time averaging operation is performed on the coal supply scheduling list. The specific method for locking the coal supply sequence that is aligned with the sampling time of the power generation sequence is as follows: Extract the n time points corresponding to the power generation sequence D1, D2, ..., Dn, and discretize them into a time sequence t1, t2, ..., tn, where t1 = t_str and tn = t_end; Construct a time axis corresponding to the time sequence t1, t2, ..., tn, and mark the coal quantity and corresponding time of all input coal-fired power units Mi in the coal supply scheduling list on the time axis. Determine the coal quantity G1 of the first input coal-fired power unit Mi and the coal quantity G2 of the second input coal-fired power unit, as well as the occurrence times t1', t2' of the two input coal-fired power unit coal quantity events and the time interval T(t1', t2') from t1' to t2'. Divide the coal quantity G1 by the time interval T(t1', t2') and record the calculation result as the coal supply quantity of coal-fired power unit Mi at each time within the time interval T(t1', t2'). Similarly, by determining the coal feed rate of coal-fired power unit Mi at n times from t_str to t_end, we obtain the coal feed rate sequence H1, H2, ..., Hn.

5. The method according to claim 4, characterized in that, In step two, the specific method for generating the coal-to-electricity conversion index curve associated with the coal-fired power unit based on the numerical relationship between the coal feed rate and the power generation at the same moment is as follows: The coal feed sequence H1, H2, ..., Hn and the power generation sequence D1, D2, ..., D are standardized in grams and kilowatt-hours, respectively. The coal-fired power conversion index R1 is calculated using H1 / D1=R1. The smaller the value of R1, the higher the energy conversion efficiency of the coal-fired power unit Mi, and vice versa. Similarly, determine the coal-to-electricity conversion index R1 at n time points, and combine them in chronological order to form the coal-to-electricity conversion index sequence R1, R2, ..., Rn; In the two-dimensional coordinate system XY1, curve fitting is performed based on the coal-to-electricity conversion index sequence R1, R2, ..., Rn to obtain the coal-to-electricity conversion index curve Smd, where the horizontal axis of XY1 is time and the vertical axis is the coal-to-electricity conversion index.

6. The method according to claim 5, characterized in that, In step three, the specific method for evaluating the steady-state coal-fired power conversion index associated with the coal-fired power unit under the current load power sequence is as follows: Extract the preset sliding window length L, where 1≤L≤n; The coal-electricity conversion index curve Smd is divided into windows by combining the sliding window length L and the preset step size s, where 1≤s≤L; The total number of statistical windows is denoted as o; Determine the standard deviation σk and mean μk of the coal-to-electricity conversion index within any window k; If σk < ε, the running state of window k is determined to be steady state, and ε is a preset stability threshold; otherwise, the running state of window k is determined to be unsteady state, and the processing of window k is skipped. For a steady-state window k, its mean μk is marked as a candidate steady-state coal-fired power conversion index, and the average load power D_k_avg within window k is locked based on the load power sequence D1, D2, ..., Dn; The average load power D_k_avg is combined with the candidate steady-state coal-fired power conversion index to form a candidate steady-state point pair; Traverse the o windows and extract all candidate steady-state point pairs within each window, denoted as the candidate steady-state point pair set; Extract the target load power D_tar allocated by the operator to the coal-fired power unit Mi within the period from t_str to t_end; For all candidate steady-state point pairs, the arithmetic mean of the candidate steady-state coal-fired power conversion index is taken as the steady-state coal-fired power conversion index of coal-fired power unit Mi under the target load power D_tar, and is denoted as Ri_D_tar.

7. The method according to claim 6, characterized in that, In step three, the steady-state coal-fired power conversion index of the coal-fired power units under different load power demands is further locked, and the optimal load power of the coal-fired power units is evaluated in reverse. The specific method for synchronizing this to the terminal large model is as follows: The steady-state coal-fired power conversion index associated with coal-fired power unit Mi under different target load power is determined based on the historical operating parameters collected during the historical normal operation of coal-fired power unit Mi. The target load power is sorted from smallest to largest as the target load power sequence D_tar1, D_tar2, ..., D_tarK, where K is the total number of target load power. Similarly, the steady-state coal-fired power conversion index sequence Ri_D_tar1, Ri_D_tar2, ..., Ri_D_tarK is obtained. A two-dimensional coordinate system XY2 is constructed with average load power as the horizontal axis and steady-state coal-fired power conversion index as the vertical axis; Based on the target load power sequence and the steady-state coal-fired power conversion index sequence, the function curve EW of the steady-state coal-fired power conversion index as a function of the target load power was plotted. The extreme value search of the steady-state coal-to-electricity conversion index is performed on the function curve EW within the rated load power range of the coal-fired power unit Mi. The minimum steady-state coal-to-electricity conversion index is marked, and the target load power corresponding to the minimum steady-state coal-to-electricity conversion index is locked and marked as the preferred load power of the coal-fired power unit Mi. Then, using the minimum steady-state coal-fired power conversion index as a benchmark, extract the v target load powers that are closest to the minimum steady-state coal-fired power conversion index value, and combine them with the preferred load power D_gd to form the preferred load power range Mi for coal-fired power units, which is then synchronized to the terminal large model.

8. The method according to claim 7, characterized in that, In step three, the preferred load power and preferred load power range of coal-fired power unit Mi are updated once every time the operator presets an update cycle. The update operation is based on the latest historical operating parameters of coal-fired power unit Mi.

9. The method according to claim 8, characterized in that, In step four, the specific method for optimizing and controlling the load power of each coal-fired power unit based on the current total power demand of the coal-fired power plant is as follows: Based on the j small models equipped with j coal-fired power units, the current preferred load power range of the j coal-fired power units is transmitted to the terminal large model. The total power demand Pall of coal-fired power plants is extracted in real time from the terminal large model; Based on the preferred load power range of each of the j coal-fired power units, the preferred load power of the j coal-fired power units is determined, and the j coal-fired power units are arranged from largest to smallest according to their preferred load power to obtain the sequence of units to be allocated; If the sum of the preferred load power of j coal-fired power units is greater than or equal to the total demand power Pall, the total demand power Pall is allocated to the j coal-fired power units in the order of the sequence of units to be allocated. Conversely, extract the sequence of units to be allocated, and sort the sequence of units to be allocated from largest to smallest according to the difference between the preferred load power of each coal-fired power unit and the maximum load power in the preferred load power range, to obtain the secondary sequence of units to be allocated; The load power of the coal-fired power units in the secondary allocation sequence is adjusted sequentially to the maximum load power in their respective preferred load power ranges until the total demand power Pall is absorbed. If the total demand power Pall still cannot be absorbed, a manual confirmation instruction is generated. Based on the above adjustment operations, the operator adaptively adjusts the load power of each coal-fired power unit to absorb the total demand power Pall. The adjustment results for all coal-fired power units are then distributed to the small model through the large model, and the small model optimizes and controls each coal-fired power unit.

Images