Method and device for estimating coal consumption of generator set, storage medium and computer equipment

Abstract

The invention discloses a method and a device for estimating coal consumption of a generator set, a storage medium and computer equipment, relates to the technical field of data analysis, and can improve the estimation efficiency and the estimation accuracy of the coal consumption of the generator set. Comprising the steps of obtaining real-time operation data of a target generator set, historical operation data of the target generator set in a historical power generation process, coal quality attribute data of fire coal used by the target generator set and flue gas attribute data of flue gas generated by the fire coal in a combustion process in response to a coal consumption estimation signal of the target generator set; based on the historical operation data, performing anomaly detection on the real-time operation data to obtain abnormal data, and removing the abnormal data in the real-time operation data to obtain cleaned real-time operation data; and inputting the cleaned real-time operation data, historical operation data, coal property attribute data and flue gas attribute data into a preset coal consumption prediction model for coal consumption prediction to obtain the unit coal consumption of the target generator set.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular, to a method, device, storage medium, and computer device for predicting the coal consumption of a generator set. Background Art

[0002] Through the analysis of the coal consumption of a generator set (coal-fired generator set), it is possible to guide the operators to adjust the real-time parameters of the generator set, maximize the energy-saving potential of the generator set, and is of great significance for improving the quality and efficiency of the power plant.

[0003] Currently, the coal consumption of a generator set is usually determined by manually measuring the reduction in the amount of raw coal. However, in this way, there will be measurement errors in the reduction in the amount of coal, resulting in a low prediction accuracy of the coal consumption of the generator set. At the same time, manually measuring the reduction in the amount of coal is time-consuming and laborious, resulting in a low prediction efficiency of the coal consumption of the generator set. Summary of the Invention

[0004] The present invention provides a method, device, storage medium, and computer device for predicting the coal consumption of a generator set, mainly capable of improving the prediction efficiency and prediction accuracy of the coal consumption of the unit.

[0005] According to a first aspect of the present invention, there is provided a method for predicting the coal consumption of a generator set, including:

[0006] In response to a coal consumption prediction signal of a target generator set, obtaining real-time operation data of the target generator set, historical operation data of the target generator set during historical power generation, coal quality attribute data of the coal used by the target generator set, and flue gas attribute data of the flue gas generated during the combustion of the coal;

[0007] Based on the historical operation data, performing anomaly detection on the real-time operation data to obtain anomaly data, and removing the anomaly data in the real-time operation data to obtain the cleaned real-time operation data;

[0008] Inputting the cleaned real-time operation data, the historical operation data, the coal quality attribute data, and the flue gas attribute data into a preset coal consumption prediction model for coal consumption prediction to obtain the coal consumption of the target generator set.

[0009] According to a second aspect of the present invention, there is provided a device for predicting the coal consumption of a generator set, including:

[0010] An obtaining unit, configured to obtain real-time operation data of the target generator set, historical operation data of the target generator set during historical power generation, coal quality attribute data of the coal used by the target generator set, and flue gas attribute data of the flue gas generated during the combustion of the coal in response to a coal consumption prediction signal of the target generator set;

[0011] A detection unit, configured to perform anomaly detection on the real-time operation data based on the historical operation data, obtain anomaly data, and remove the anomaly data from the real-time operation data to obtain the cleaned real-time operation data;

[0012] A determination unit, configured to input the cleaned real-time operation data, the historical operation data, the coal quality attribute data, and the flue gas attribute data into a preset coal consumption prediction model to perform coal consumption prediction, and obtain the unit coal consumption of the target generating unit.

[0013] According to a third aspect of the present invention, there is provided a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned method for estimating the coal consumption of a generating unit is implemented.

[0014] According to a fourth aspect of the present invention, there is provided a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the above-mentioned method for estimating the coal consumption of a generating unit is implemented.

[0015] According to a method, device, storage medium, and computer device for estimating the coal consumption of a generating unit provided by the present invention, compared with the current method of determining the coal consumption of a generating unit by manually measuring the reduction of the original coal, the present invention responds to a coal consumption estimation signal of a target generating unit, and obtains the real-time operation data of the target generating unit, the historical operation data of the target generating unit during historical power generation, the coal quality attribute data of the coal used by the target generating unit, and the flue gas attribute data of the flue gas generated during the combustion of the coal; then, based on the historical operation data, perform anomaly detection on the real-time operation data to obtain anomaly data, and remove the anomaly data from the real-time operation data to obtain the cleaned real-time operation data; finally, input the cleaned real-time operation data, the historical operation data, the coal quality attribute data, and the flue gas attribute data into a preset coal consumption prediction model to perform coal consumption prediction, and obtain the unit coal consumption of the target generating unit. Thus, by comprehensively analyzing the real-time operation data, historical operation data, coal quality attribute data, and flue gas attribute data to determine the unit coal consumption, the estimation accuracy of the unit coal consumption can be improved. At the same time, the present invention uses a preset coal consumption prediction model to perform coal consumption prediction, and the whole process does not require manual participation, avoiding the time-consuming and laborious situation of manually measuring the reduction of the original coal and the situation of measurement errors. Therefore, the present invention can improve the estimation efficiency and estimation accuracy of the unit coal consumption. In addition, the present invention uses historical operation data to detect anomaly data in real-time operation data, which can ensure the effectiveness of real-time operation data, thereby further improving the estimation accuracy of the unit coal consumption. Description of the Drawings

[0016] The accompanying drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings:

[0017] Figure 1 A flowchart of a method for predicting the coal consumption of a generator set provided by an embodiment of the present invention is shown;

[0018] Figure 2 A flowchart of another method for predicting the coal consumption of a generator set provided by an embodiment of the present invention is shown;

[0019] Figure 3 A schematic structural diagram of a device for predicting the coal consumption of a generator set provided by an embodiment of the present invention is shown;

[0020] Figure 4 A schematic structural diagram of another device for predicting the coal consumption of a generator set provided by an embodiment of the present invention is shown;

[0021] Figure 5 A schematic physical structure diagram of a computer device provided by an embodiment of the present invention is shown. Detailed implementation manners

[0022] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments. It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other.

[0023] Currently, the method of determining the coal consumption of a generator set by manually measuring the reduction amount of raw coal is time-consuming and laborious, and it is easy to have errors in measuring the reduction amount, resulting in low prediction efficiency and low prediction accuracy of the coal consumption of the unit.

[0024] To solve the above problems, an embodiment of the present invention provides a method for predicting the coal consumption of a generator set, as Figure 1 shown, the method includes:

[0025] 101. In response to the coal consumption prediction signal of the target generator set, obtain the real-time operation data of the target generator set, the historical operation data of the target generator set during the historical power generation process, the coal quality attribute data of the coal used by the target generator set, and the flue gas attribute data of the flue gas generated during the combustion of the coal.

[0026] The target generator set may be a generator set for supplying electricity in a power plant; the real-time operation data include: the real-time unit load of the target generator set, the real-time boiler thermal efficiency, the real-time steam turbine heat rate, the main steam flow, the main steam temperature, the main steam pressure, the real-time power generation, the real-time coal inventory, the real-time heating data, etc.; the historical operation data include: the unit load data of the target generator set over a long period of time in the historical working process, the boiler operation data (the thermal efficiency of the boiler, the combustion parameters of the boiler, the exhaust temperature of the boiler, the exhaust oxygen content, etc.), the steam turbine operation data (the heat rate of the steam turbine, the main steam flow, the main steam temperature, the main steam pressure, etc.); power, exhaust temperature, cylinder efficiency, etc.), power generation, heating data, coal inventory, power generation coal consumption data, historical data of electrical parameters such as power factor and reactive power of generators, number of starts and stops of units and coal consumption records during starts and stops, historical data of plant power consumption rate, historical data of thermal system parameters such as water replenishment rate and feed water temperature, coal quality data, climate data, etc.; coal quality attribute data include: moisture content, ash content, volatile matter, particle size distribution, element content, combustion characteristics, etc. contained in coal; flue gas attribute data include: flue gas flow, flue gas temperature, flue gas composition, flue gas humidity, flue gas density, flue gas pressure, flue gas pressure composition and other information.

[0027] For the embodiment of the present invention, when it is necessary to predict the coal consumption of the target generator set, it is first necessary to measure the real-time operation data of the target generator set through a measuring device such as a sensor, obtain the historical operation data of the target generator set in the historical power generation process in the database, and measure the coal quality attribute data of the coal used by the target generator set and the flue gas attribute data of the flue gas generated during the combustion of the coal through a measuring device, and then comprehensively analyze the real-time operation data, historical operation data, coal quality attribute data, and flue gas attribute data to determine the unit coal consumption of the target generator set. By comprehensively analyzing various factors that affect the unit coal consumption to predict the unit coal consumption, the prediction accuracy of the unit coal consumption can be improved.

[0028] 102. Based on the historical operation data, perform anomaly detection on the real-time operation data to obtain abnormal data, and remove the abnormal data in the real-time operation data to obtain cleaned real-time operation data.

[0029] For the embodiments of the present invention, in order to improve the prediction accuracy of the unit coal consumption, it is first necessary to ensure the quality of the data. Based on this, after obtaining the historical operation data, it is necessary to preprocess the historical operation data. The specific preprocessing methods include: taking any operation data in the historical operation data as a target operation data respectively, and calculating the distance between the target operation data and the remaining historical operation data respectively, where the remaining historical operation data is the historical operation data after removing the target operation data; determining the number of distances between the target operation data and the remaining historical operation data that are greater than a preset distance threshold. If the number of distances is greater than a preset number threshold, then determining the target operation data as an outlier data, and removing the outlier data in the historical operation data to obtain the cleaned historical operation data; determining the data type of the cleaned historical operation data, and determining a regression fitting model corresponding to the data type; using the regression fitting model to smooth the cleaned historical operation data to obtain the denoised historical operation data; dividing the denoised historical operation data into multiple groups of historical operation data according to a preset time period; calculating the data mean corresponding to each group of historical operation data respectively, and using the data mean to replace the maximum historical operation data within the corresponding group to obtain the smoothed historical operation data.

[0030] Among them, the preset distance threshold is set according to actual requirements, and the preset number threshold is set according to actual requirements; the data types include: numerical type (such as continuous variables like temperature, pressure, flow rate, etc.), categorical type (such as discrete variables like equipment status, operation mode, etc.); the preset time period is set according to actual requirements.

[0031] Specifically, for example, if the historical operation data includes: data a, data b, data c, data d, data e, and if data a in the historical operation data is taken as the target operation data, then the remaining historical operation data is: data b, data c, data d, data e; the distance between data a and data b is 10, the distance between data a and data c is 15, the distance between data a and data c is 5, the distance between data a and data d is 8, the distance between data a and data e is 20. If the preset distance threshold is 6 and the preset number threshold is 3, then the number of distances between data a and the remaining historical operation data that are greater than the preset distance threshold is 4. Since the number of distances 4 is greater than the preset number threshold 3, it is determined that data a is an outlier data. Thus, in this way, it is possible to detect whether the remaining historical operation data is an outlier data, and finally remove all the outlier data in the historical operation data to obtain the cleaned historical operation data. By detecting outlier data in the historical operation data, the embodiments of the present invention can identify and remove the noise in the historical operation data, improve the data quality, and further improve the prediction accuracy of the unit coal consumption.

[0032] Further, after detecting outliers in the historical operation data, it is also necessary to smooth the historical operation data. The specific smoothing method is as follows: If the historical operation data is numerical and shows a linear relationship, a linear regression fitting model can be selected and used to smooth the historical operation data. If the historical operation data shows a non-linear relationship, a polynomial regression fitting model can be used to smooth the historical operation data. If the historical operation data is time series data, a time series regression fitting model can be used to smooth the historical operation data. If the relationship of the historical operation data is complex and cannot be represented by a linear or polynomial relationship, a non-linear regression fitting model can be used to smooth the historical operation data. When specifically using the regression fitting model to smooth the historical operation data, first, the regression fitting model needs to be trained using the sample data set (where the sample data set used depends on the model being trained. For example, when training a time series regression fitting model, the sample data set includes the labeled sample time series operation data of the sample generator set, and the label is the actual operation data after smoothing). During the training process, the model will learn the patterns and rules in the data and try to minimize the prediction error. After the training is completed, the model needs to be verified to ensure that its prediction performance meets the expectations. After the model verification passes, the model is used to smooth the historical operation data to obtain the denoised historical operation data. Further, the denoised historical operation data is divided into multiple groups of historical operation data according to a preset time period. For example, the denoised historical operation data is divided at intervals of 1h. If the historical operation data in the past 10h is collected, the denoised historical operation data can be divided into 10 groups, and each group contains the historical operation data at the corresponding collection time point. For each group, the data mean of each historical operation data in the group is calculated respectively, and then the maximum historical operation data in the group is replaced with the data mean to obtain the smoothed historical operation data. By performing regression fitting and removing gross deviations on the historical operation data in the embodiments of the present invention, the data quality can be improved, thereby improving the prediction accuracy of the subsequent unit coal consumption. Finally, based on the smoothed historical operation data, anomaly detection is performed on the real-time operation data to obtain anomaly data. Thus, by preprocessing the historical operation data, the data quality of the historical operation data can be improved, and the anomaly detection accuracy of the real-time operation data can be improved. At the same time, by performing anomaly detection on the real-time operation data in the embodiments of the present invention, the data quality of the real-time operation data can be improved, avoiding the problems of wasted computing resources due to operations on useless data and disturbing the calculation results, thereby further improving the prediction efficiency and prediction accuracy of the subsequent unit coal consumption.

[0033] In another embodiment of the present invention, in the real-time operation data measured by sensors, various noises and interferences are contained in the output signals of the sensors. The weighted recursive average filtering method can be used to filter the output signals. In this way, high-quality real-time operation data can be obtained. Further, the real-time operation data and the historical operation data can be subjected to data steady-state screening and processing. The accurate prediction of the coal consumption of the unit and the reconstruction of the operation state both rely on the authenticity of the actual operation data. At the same time, strong state consistency is also required among the various operation parameters. The so-called state consistency means that under the determined operation boundary conditions, the many operation parameters used to describe the operation state of the unit conform to the determined state correspondence relationship, so as to ensure that the relationship between these parameters has strong regularity. When the unit operates under stable conditions, the parameters in the thermal system fluctuate slightly, and the various parameters have high state consistency. The small fluctuations of the parameters have little impact on the accuracy of the online performance calculation and can meet the calculation accuracy requirements of the monitoring system. However, due to the variability of the actual operation boundary conditions and the existence of the thermal inertia and mass inertia of the thermal equipment, especially critically, the thermal inertia and mass inertia of the system are closely related to the actual operation boundary conditions and equipment characteristics of the actual unit, and even the system connection mode. Therefore, many actual operation data describing the operation conditions of the unit show inertia delay to varying degrees. That is, when the unit transitions from one stable condition to another stable condition, the degree of inertia delay of each parameter changes, causing the parameter fluctuation range of the thermal system to become larger. If the calculation and analysis are carried out based on the instantaneously collected measurement data, under different operation boundary conditions (power generation load, ambient temperature, fuel characteristics, etc.), due to the existence of the changing inertia delay characteristics, the delay degrees of different parameters are different, that is, the asynchronous changes of the various parameters lead to the destruction of the state consistency between the parameters, resulting in large fluctuations in the calculation results of the system, which is obviously inconsistent with the actual operation performance of the unit. Based on the above analysis, the following conclusions can be drawn:

[0034] (1) Only under strict steady-state conditions does there exist strict state parameter state consistency.

[0035] (2) There are no strict steady-state conditions for the actual operating unit, so the state consistency of the actual operation data is poor, and as the speed of the condition change accelerates, its state consistency will deteriorate further, which is particularly obvious for the start-up and shutdown processes of the unit.

[0036] Therefore, if the historical data in the quasi-steady state within the normal operating load range (40% - 100%) is used as the analysis basis, the operation state reconstruction of the unit can be achieved. By analyzing the interaction relationships among the operation parameters, the characteristics of the thermal system and related equipment under different working conditions can be discovered more accurately. The boundaries of the thermal system can be divided into controllable boundary conditions and uncontrollable boundary conditions. Among them, the controllable boundary conditions involve the operation boundary conditions such as main steam pressure, main steam temperature, and reheat temperature that can be adjusted and controlled during operation. The uncontrollable boundary conditions are the boundary conditions such as power generation load, atmospheric environment conditions, and fuel characteristics that are uncontrollable but must be satisfied. For a determined unit, the structural characteristics, material characteristics of its related equipment, and the structure of the system can be regarded as the inherent properties of the system, while the operation state parameters of the system as the output are the responses of the thermal system to the input boundary conditions, and there must be a unique mapping relationship. Therefore, for a determined thermal system, if its boundary conditions do not change, then all the operation state parameters corresponding to its operation are necessarily determined. This is exactly the theoretical basis for the steady-state processing of dynamic data. For real-time dynamic operation data, any operation parameter of the system at time τ can be expressed as x i (τ), and its operation state can be represented by the set X(τ) = {x 1 (τ), …, x i (τ), …, x n (τ)}, where n is the number of necessary state parameters for describing the operation state of the system. The set of system operation boundary conditions is Y(τ) = {y 1 (τ), …, y j (τ), …, y m (τ)}, where m is the number of necessary boundary parameters for determining the operation state of the system. When the boundary condition set Y(τ) is screened and processed through steady state, it can be considered independent of time. Different from the usual method for judging steady-state working conditions in thermal tests with a load change rate of 1% - 5%, only when all the boundary parameters of the thermal system are in a steady state, the operation state parameters of the system can be uniquely determined, and thus the entire thermal system will be in a steady state. For the problem of reconstructing the operation parameters on the steam turbine side, the condition for judging that the unit is in a steady state is that the five parameters of power generation load, circulating water inlet temperature, main steam pressure, main steam temperature, and reheat temperature do not change with time (or the change rate is very small). Usually, the sampling time interval Δτ of the data acquisition system (sensors, etc.) remains unchanged. Therefore, the problem of steady-state screening of the real-time operation data and historical operation data of the unit can be transformed into the problem of determining the time domain where the boundary condition parameter set Y(τ) is approximately unchanged and selecting the allowable fluctuation range. The steady-state discrimination formula for the operation state of the unit is given as follows:

[0037]

[0038] Γ = (j - 1)*Δτ > Γ d

[0039] where y i is the boundary condition, τ 0 is the starting time of data sampling; j is the number of samplings that meet the steady-state discrimination condition, and δ yi is the allowable fluctuation range of the set parameters. Γ is the sampling time threshold for meeting the steady state, and Γ d is the set steady-state time threshold. Considering the dynamic inertia and thermal inertia of the system and each thermal equipment, if Γ d is set too small, the stable time of the unit is too short to meet the requirement of the state consistency of the parameters; conversely, the required stable time is too long, reducing the available records and resulting in insufficient data for the reconstructed model. Γ d can be determined according to the actual effect of screening. Usually, Γ d = 5 min.

[0040] Regarding the boundary parameter y i and the fluctuation range δ yi of the setting, it is usually based on the maximum allowable error and fluctuation range required, and has a large degree of subjectivity. It can be imagined that if δ yi is set too wide, the fluctuation of the operating state of the unit will increase, which will reduce the state consistency of the operating parameters in the operating state set X(τ) = {x 1 (τ), …, x i (τ), … x n (τ)}; conversely, the conditions will be too harsh, resulting in too few screened records and unable to meet the needs of subsequent model analysis.

[0041] In another embodiment of the present invention, the real-time operating data and historical operating data can also be detected according to the threshold screening technology. For example, for the verification of the operating measurement data of a single point, it can be judged whether the measured value is reasonable according to the preset parameter reference value and its threshold. If the measured value exceeds the set threshold range, the measured data is considered incorrect. This method is simple to apply and can exclude some obviously incorrect measured data.

[0042] 103. Input the cleaned real-time operating data, historical operating data, coal quality attribute data, and flue gas attribute data into a preset coal consumption prediction model for coal consumption prediction to obtain the unit coal consumption of the target generating unit.

[0043] For the embodiments of the present invention, in order to reduce the computational amount of historical operation data, it is necessary to streamline the historical operation data, that is, perform reduction processing on the historical operation data. Based on this, the method includes: initializing the centroid vectors corresponding to the centroids of different clusters, and determining the data feature vectors corresponding to the historical operation data; based on the data feature vectors and the centroid vectors, calculating the distances between each of the historical operation data and the centroids corresponding to each of the clusters, and based on the distances, dividing each of the historical operation data into each of the clusters; based on the data feature vectors corresponding to the historical operation data in each of the clusters, determining the updated centroid vectors corresponding to each of the clusters; based on the updated centroid vectors, re-dividing each of the historical operation data into each of the clusters until the updated centroid vectors do not change, and determining the historical operation data finally divided into each of the clusters as the historical operation data under different clustering categories; determining the central operation data and the boundary operation data in each of the clustering categories, and using each of the central operation data and each of the boundary operation data to replace the historical operation data to obtain the reduced historical operation data.

[0044] Specifically, use methods such as word embedding to determine the data feature vectors corresponding to the historical operation data, and select the centroid vectors corresponding to the initial centroids respectively corresponding to K clusters. For the data feature vectors respectively corresponding to the historical operation data, calculate the distances from each data feature vector to the K centroid vectors, and allocate each historical operation data to the cluster corresponding to the centroid vector with the closest distance. Then, for each cluster, recalculate the centroid of each cluster and its corresponding centroid vector, and re-divide the historical operation data into different clusters, and continuously divide the historical operation data in this way until the position of the centroid does not change, that is, the centroid vector does not change. Finally, determine the historical operation data divided into different clusters as the historical operation data under different clustering categories. Then, select the feature point data in the historical operation data in each clustering category, such as the historical operation data with significant features such as the center point and the boundary point, and only retain the historical operation data with significant features such as the center point and the boundary point in each clustering category, and delete the remaining historical operation data without significant features, so as to obtain the streamlined historical operation data, that is, the reduced historical operation data. Since the original historical operation data contains a large amount of data, if these data are directly taken out for analysis, it is time-consuming and laborious. Therefore, by performing reduction processing on the historical operation data, it is possible to reduce the amount of computation for redundant data while retaining the same or similar laws as the original historical operation data, and thus solve the computing resources.

[0045] Furthermore, the cleaned real-time operation data, the reduced historical operation data, the coal quality attribute data, and the flue gas attribute data are input into the coal consumption prediction model to predict the coal consumption and obtain the unit coal consumption. The embodiment of the present invention can improve the data quality by performing pre-processing such as smoothing, denoising, and reducing the historical operation data, avoid wasting computing resources and disturbing the computing results by calculating useless data, and thus improve the subsequent prediction efficiency and accuracy of the unit coal consumption.

[0046] According to a method for estimating coal consumption of a power generation unit provided by the present invention, compared with the current method of determining the coal consumption of a power generation unit by manually measuring the reduction of original coal, the present invention obtains the real-time operation data of the target power generation unit, the historical operation data of the target power generation unit in the historical power generation process, the coal quality attribute data of the coal used by the target power generation unit, and the flue gas attribute data of the flue gas generated during the combustion of the coal by responding to the coal consumption estimation signal of the target power generation unit; then, based on the historical operation data, the real-time operation data is detected for abnormality to obtain abnormal data, and the abnormal data in the real-time operation data is eliminated to obtain the cleaned real-time operation data; finally, the cleaned real-time operation data, the historical operation data, the coal quality attribute data, and the flue gas attribute data are input into a preset coal consumption prediction model for coal consumption prediction to obtain the unit coal consumption of the target power generation unit. Therefore, the coal consumption of the unit is determined by comprehensively analyzing the real-time operation data, historical operation data, coal quality attribute data, and the flue gas attribute data, thereby improving the estimated accuracy of the coal consumption of the unit. At the same time, the present invention predicts coal consumption by using a preset coal consumption prediction model. The entire process does not require human participation, thus avoiding the time-consuming and labor-intensive manual measurement of the reduction in original coal, as well as measurement errors. Therefore, the present invention can improve the estimated efficiency and accuracy of the coal consumption of the unit. In addition, the present invention uses historical operation data to detect abnormal data on real-time operation data, thereby ensuring the validity of real-time operation data, thereby further improving the estimated accuracy of the coal consumption of the unit.

[0047] Further, in order to better illustrate the above process of estimating the coal consumption of a generator set, as a refinement and extension of the above embodiment, the embodiment of the present invention provides another method for estimating the coal consumption of a generator set, such as Figure 2 As shown, the method includes:

[0048] 201. In response to the coal consumption estimation signal of the target power generation unit, obtain the real-time operation data of the target power generation unit, the historical operation data of the target power generation unit in the historical power generation process, the coal quality attribute data of the coal used by the target power generation unit and the flue gas attribute data of the flue gas generated during the combustion of the coal.

[0049] Specifically, when a coal consumption estimation signal of a target power generation unit is received, based on the identification information of the target power generation unit carried in the coal consumption estimation signal, the historical operation data of the target power generation unit in the historical working process is obtained from the database, as well as the real-time operation data of the target power generation unit measured by measuring devices such as sensors, and the coal quality attribute data of the coal used by the target power generation unit and the flue gas attribute data of the flue gas generated during the combustion of the coal are obtained by searching for information and conducting on-site measurements.

[0050] 202. Based on the historical operation data, perform anomaly detection on the real-time operation data to obtain abnormal data, and remove the abnormal data in the real-time operation data to obtain cleaned real-time operation data.

[0051] For the embodiment of the present invention, before using historical operation data to perform anomaly detection on real-time operation data, it is necessary to pre-process the historical operation data. Based on this, the method includes: determining the data collection time point of the historical operation data, and determining the first preset time period where the collection time point is located, the second preset time period before the collection time point, and the third preset time period after the collection time point; determining the first load change value of the target generator group within the first preset time period, the second load change value of the target generator group within the second preset time period, and the third load change value of the target generator group within the third preset time period; based on the first load change value, the second load change value, and the third load change value, determining invalid data in the historical operation data, and eliminating the invalid data in the historical operation data to obtain valid historical operation data; determining the missing data collection time point of the missing data in the valid historical operation data, and determining the missing data Collect each neighbor data within a preset time range corresponding to the time point; determine the second-order interpolation data of the missing data based on each of the neighbor data; determine the operating condition information of the target generator set at the missing data collection time point, obtain the to-be-matched operating condition information at the data collection time point corresponding to the valid historical operating data, and determine the target operating condition information that matches the operating condition information in the to-be-matched operating condition information; determine the matching data collection time point corresponding to the target operating condition information in the data collection time point corresponding to the valid historical operating data, and calculate the data mean and data median of the valid historical operating data corresponding to the matching data collection time point; respectively determine the weight coefficients corresponding to the data mean, the data median, and the second-order interpolation data, and based on the weight coefficients, add the data mean, the data median, and the second-order interpolation data to obtain a supplementary value, and use the supplementary value to supplement the missing data to obtain the supplemented valid historical operating data.

[0052] Among them, the first preset time period, the second preset time period, and the third preset time period are all set according to actual requirements; the preset time range is set according to actual requirements; the operating condition information includes: information such as rotational speed, active power, and reactive power.

[0053] Specifically, taking a certain data a in the historical operation data as an example, if the data collection time point of the data a is 5:40, then the first preset time period is 5:00 - 6:00, the second preset time period is 4:00 - 5:00, and the third preset time period is 6:00 - 7:00. If the load change value of the generator set within the first preset time period of 5:00 - 6:00 is 19 MW, the load change value of the generator set within the second preset time period of 4:00 - 5:00 is 11 MW, and the load change value of the generator set within the third preset time period of 6:00 - 7:00 is 12 MW. If the first load change threshold corresponding to the first preset time period is 18 MW, the first load change threshold corresponding to the second preset time period is 10 MW, and the third load change threshold corresponding to the third preset time period is 10 MW, then it is determined whether the first load change value is greater than the first load change threshold, whether the second load change value is greater than the second load change threshold, and whether the third load change value is greater than the third load change threshold; if the first load change value is greater than the first load change threshold, and the second load change value is greater than the second load change threshold, and the third load change value is greater than the third load change threshold, then it is determined that the target generator set is in a variable operating condition state at this time, and the operation data collected in the variable operating condition state is invalid data, that is, it can be determined that the data a is invalid data. Thus, invalid data can be determined in the remaining historical operation data according to the above method, and finally all invalid data in the historical operation data are removed to obtain valid historical operation data. By removing the invalid historical operation data in the historical operation data, the embodiment of the present invention can improve the data quality, avoid the interference of the invalid historical operation data on the calculation result, and also avoid the computing resources wasted by calculating the invalid historical operation data, thereby improving the subsequent abnormal detection efficiency and detection accuracy of the real-time operation data, and further improving the subsequent prediction efficiency and prediction accuracy of the unit coal consumption.

[0054] Furthermore, due to factors such as missing records, there will also be missing data in the historical operation data. These incomplete data will bring great difficulties to the data mining work. For example, there is only one record within a one-hour interval, and there are even no records at some time points. Therefore, it is necessary to supplement the missing data. The specific supplement method is as follows: First, determine a preset time range around a time period of the collection time point of the missing data. The preset time range can be set according to the sampling frequency of the data, the severity of data changes, and the requirements of interpolation accuracy. Then, determine each historical operation data within the preset time range as the neighboring data of the missing data. After that, substitute the data collection time points of the neighboring data into a pre-constructed quadratic polynomial (this quadratic polynomial is a polynomial of the correlation relationship between the neighboring data and time pre-constructed according to the sample data), solve the quadratic polynomial, and the second-order interpolation data of the missing data can be obtained. Furthermore, determine the operating condition information of the target generating unit at the collection time point of the missing data, and determine the effective historical operation data under the same operating condition as the operating condition information at the collection time point of the missing data in the effective historical operation data. Calculate the data mean and data median of the effective historical operation data under the same operating condition, and set different weight coefficients for the data mean, data median, and second-order interpolation data. Finally, according to the weight coefficients, perform weighted averaging on the data mean, data median, and second-order interpolation data to obtain a supplementary value, and use this supplementary value to supplement the missing data to obtain the supplemented effective historical operation data. Finally, based on the supplemented effective historical operation data, perform anomaly detection on the real-time operation data to obtain anomaly data. By supplementing the missing data in the historical operation data in the embodiments of the present invention, the integrity of the data can be ensured, providing a solid foundation for subsequent data analysis and coal consumption prediction.

[0055] Further, after a series of preprocessing on the historical operation data, it is necessary to perform anomaly detection on the real-time operation data based on the preprocessed historical operation data. Based on this, step 202 specifically includes: determining a reference value and a preset deviation range corresponding to the real-time operation data based on the historical operation data. Among them, the method for determining the reference value and the preset deviation range corresponding to the real-time operation data based on the historical operation data includes: determining the data types of the historical operation data, and dividing each data type into a preset number of data intervals based on the magnitudes of the historical operation data in each data type; taking any two data types in each data type as a first target data type and a second target data type respectively, and taking any data interval in the data interval corresponding to the first target data type as a first target interval respectively, and taking any data interval in the data interval corresponding to the second target data type as a second target interval respectively, and determining a first probability value that the historical operation data in the first target data type falls within the first target interval; determining a second probability value that the historical operation data in the second target data type falls within the second target interval when the historical operation data in the first target data type falls within the first target interval; determining the correlation degree between the historical operation data in the first target data type and the historical operation data in the second target data type based on the first probability value and the second probability value; determining a first reference value corresponding to the first target data type based on the correlation degree of the historical operation data in the first target data type, and determining the same-type real-time operation data of the same type as the first target data type in the real-time operation data, and determining the first reference value as the reference value of the same-type real-time operation data; determining the preset deviation range of the same-type real-time operation data based on the reference value of the same-type real-time operation data; performing anomaly detection on the real-time operation data based on the preset deviation range to obtain the abnormal data in the real-time operation data; and correcting the abnormal data by using the reference value to obtain the corrected real-time operation data.

[0056] Among them, the data types of each historical operation data include: main steam flow type, main steam temperature type, main steam pressure type, etc. Specifically, first, it is necessary to determine the reference value and the preset deviation range of the real-time operation data according to the historical operation data, and judge whether the real-time operation data is within the preset deviation range. If so, it is determined that the real-time operation data is not abnormal data. If the real-time operation data is not within the preset deviation range, it is determined that the real-time operation data is abnormal data. At this time, the abnormal data in the real-time operation data is replaced with the reference value to obtain the corrected real-time operation data. Among them, the specific process of determining the reference value of the real-time operation data is as follows: for example, taking the main steam flow type and the main steam temperature type as an example, the main steam flow type is divided into n intervals (L 1 、L 2 、...L i 、...L n ) according to the main steam flow size, and the main steam temperature type is divided into m intervals (Z 1 、Z 2 、...Z j 、...Z m ) according to the main steam temperature size. Then, it is determined that the first probability b that the main steam flow d 0 falls within the interval L i is b = P(d 0i ), and at the same time, it is determined that when the main steam flow d 0 falls within the interval L i , the second probability s that the main steam temperature t 0 of the main steam temperature type falls within Z j is s = P(d 0i , t 0j ). Then, according to the first probability and the second probability, the third probability C = s / b is calculated. Then, the weight coefficients corresponding to the second probability and the third probability are determined respectively (the weight coefficients are set according to actual needs), and based on the weight coefficients, the second probability and the third probability are added to obtain the correlation degree between the main steam flow d 0 and the main steam temperature t 0 . Thus, the correlation degree between the historical operation data of each data type and the historical operation data of the remaining data types can be determined in the above manner. The maximum correlation degree is determined among the various correlation degrees, and the historical operation data of the data type corresponding to the maximum correlation degree is determined as the reference value corresponding to this data type, and this reference value is determined as the reference value corresponding to the real-time operation data of the same data type.

[0057] Furthermore, the preset deviation range of the real-time operation data of the same type is determined according to the following formula:

[0058]

[0059] Wherein, X is the reference value of real-time operation data of the same type, confidence is the third probability corresponding to the reference value X, support is the second probability corresponding to the reference value X, and x k is the real-time operation data, and σ min is the lower limit value of the preset deviation range, and σ max is the upper limit value of the preset deviation range. Thus, the preset deviation range corresponding to any real-time operation data can be determined in the above manner.

[0060] 203. Determine the real-time feature vector corresponding to the cleaned real-time operation data, the historical feature vector corresponding to the historical operation data, the coal quality feature vector corresponding to the coal quality attribute data, and the flue gas feature vector corresponding to the flue gas attribute data.

[0061] 204. Perform cross-processing on the real-time feature vector, the historical feature vector, the coal quality feature vector, and the flue gas feature vector to obtain a coal consumption cross vector.

[0062] For the embodiments of the present invention, the real-time feature vector corresponding to the cleaned real-time operation data, the historical feature vector corresponding to the historical operation data, the coal quality feature vector corresponding to the coal quality attribute data, and the flue gas feature vector corresponding to the flue gas attribute data are determined by means of word embedding and the like. Then, cross-processing is performed on the above feature vectors. The specific cross-processing methods include: performing feature-level cross-processing on the real-time feature vector, the historical feature vector, the coal quality feature vector, and the flue gas feature vector to obtain a feature cross vector; performing element-level cross-processing on the real-time feature vector, the historical feature vector, the coal quality feature vector, and the flue gas feature vector to obtain an element cross vector; performing low-order cross-processing on the real-time feature vector, the historical feature vector, the coal quality feature vector, and the flue gas feature vector to obtain a low-order cross vector; and performing transformation processing on the feature cross vector, the element cross vector, and the low-order cross vector by using a preset transformation function to obtain a coal consumption cross vector.

[0063] Specifically, in order to make full use of the relationships between data, extract more implicit features, take into account both high-order and low-order processing, make more full use of the data, and make the subsequent prediction results more accurate to meet the requirements of actual application scenarios, it is necessary to perform cross-processing on the real-time feature vector, historical feature vector, coal quality feature vector, and flue gas feature vector. The specific cross-processing method is as follows: If the real-time feature vector is (a1, a2), the historical feature vector is (b1, b2), the coal quality feature vector is (c1, c2), and the flue gas feature vector is (d1, d2), the specific cross-processing method includes: performing cross at the feature level between different feature vectors, that is, after performing the Hadamard product on all elements between the vectors, performing a convolution transformation under a certain weight to obtain a feature cross vector of f(w*(a1*b1*c1*d1, a2*b2*c2*d2, a3*b3*c3*d3)); at the same time, performing cross at the element level on all feature vector data, that is, after performing the Hadamard product on each element between the vectors, assigning different weight values to each product result, and then performing a linear transformation to obtain an element cross vector of f(w1*a1*b1*c1*d1, w2*a2*b2*c2*d2, w3*a3*b3*c3*d3)); in addition, performing low-order cross-processing on all feature vectors, then assigning a weight coefficient to the result of the cross-processing, and then performing a linear transformation to obtain a low-order cross vector of f(w2(a1, a2, b1, b2, c1, c2, d1, d2)); finally, using a preset transformation function (the preset transformation function can be set according to the actual situation, and this embodiment does not limit this) to perform transformation processing on the above feature cross vector, element cross vector, and low-order cross vector, such as horizontal splicing, to obtain a coal consumption cross vector. It should be noted that the above examples are only illustrative and do not limit the embodiments of the present application. Thus, by performing cross-processing on the real-time feature vector, historical feature vector, coal quality feature vector, and flue gas feature vector, different features can be automatically or explicitly combined to generate new feature combinations. These combined features may contain complex non-linear relationships between the original features, enabling the model to capture more refined and rich information in the data, that is, to make full use of the relationships between various data, extract more implicit features, take into account both high-order and low-order processing, make more full use of the data, and make the subsequent unit coal consumption result more accurate to meet the requirements of actual application scenarios.

[0064] 205. Input the coal consumption cross vector into a preset coal consumption prediction model for coal consumption prediction to obtain the unit coal consumption of the target generating unit.

[0065] For the embodiments of the present invention, in order to improve the prediction accuracy of the preset coal consumption prediction model, it is first necessary to train and construct the preset coal consumption prediction model. Based on this, the method includes: constructing an initial coal consumption prediction model and obtaining a sample data set, where the sample data set includes sample data with label information, and the sample data is the real-time operation data, historical operation data, coal quality attribute data, and flue gas characteristic vector corresponding to the sample generator set, and the label information is the actual coal consumption; dividing the sample data set into training data and test data, training the initial coal consumption prediction model with the training data, testing the trained initial coal consumption prediction model with the test data, and finally determining the trained initial coal consumption prediction model that meets the test conditions as the preset coal consumption prediction model. Further, inputting the coal consumption cross vector into the preset coal consumption prediction model for coal consumption prediction, and the unit coal consumption of the target generator set can be output through the preset coal consumption prediction model. By training the model with the sample data set in the embodiments of the present invention, the generalization ability of the preset coal consumption prediction model can be improved, thereby improving the subsequent prediction accuracy of the unit coal consumption.

[0066] Further, in order to obtain more information about the target generator set to adjust the unit operation parameters to achieve the energy conservation and emission reduction effect, it is also necessary to obtain information such as the target controllable operation data of the target generator set. Based on this, the method includes: determining the target controllable operation data of the target generator set during the stable operation process based on the historical operation data; reconstructing the operation state of the target generator set based on the historical operation data and determining the reconstructed operation data of the reconstructed target generator set; determining the energy consumption deviation value of the target generator set based on the cleaned real-time operation data, historical operation data, coal quality attribute data, flue gas attribute data, and reconstructed operation data; and sending the unit coal consumption, the target controllable operation data, and the energy consumption deviation value to the display terminal for display.

[0067] Specifically, based on historical operation data, by optimizing the operating state of the generating unit under system constraints, the optimized target values (target controllable operation data) of the controllable operation parameters of the target generating unit under the current working conditions can be obtained. Among them, the target controllable operation data are the best performance parameters that the generating unit can achieve when operating in an efficient, stable, and environmentally friendly state based on the current equipment and control conditions. The target controllable operation data are determined according to information such as historical operation data and the current actual working conditions of the generating unit. For example, after the target generating unit is put into operation, a large number of operation optimization tests can be carried out to determine the target controllable operation data under different working conditions. The target controllable operation data include: main steam pressure, main steam temperature, reheat steam temperature, feed water temperature, unburned carbon content, and flue gas temperature of the target generating unit, etc. For example, indicators: boiler efficiency, circulating water inlet temperature, flow passage efficiency, main steam flow rate, reheat steam flow rate, regulating pressure, etc. Determine the energy efficiency state of the generating unit: Compare the current value of the indicator with the target value, and the deviation between the two can be used to calculate the impact on the energy efficiency of the generating unit. How to determine the target controllable operation data such as main steam pressure, main steam temperature, reheat steam temperature, number of circulating water pumps, etc., is determined by the system through an online optimization model for historical data. Its technical route is: Consider the influence of uncontrollable parameters such as ambient temperature and load on the power supply coal consumption rate of the generating unit, take the uncontrollable boundary parameters as conditions, and take the power supply coal consumption rate of the generating unit as the objective function to establish an optimal working condition database for the generating unit. And divide the working conditions based on the uncontrollable boundary parameters of this generating unit, compare the real-time operating condition with the working condition under the same conditions in the optimal working condition database. If there is no working condition under this condition in the optimal working condition database, then store this real-time condition as the optimal working condition; if there is a working condition under this condition in the optimal working condition database, compare the power supply coal consumption rates of the two working conditions. If the power supply coal consumption rate of the real-time working condition is large, the optimal working condition database remains unchanged. If the power supply coal consumption rate of the real-time working condition is low, then store this working condition in the optimal working condition database to replace the original optimal working condition under this condition. The optimized operation database established in this way can be used to determine the historical optimal value of the controllable operation parameters, also known as the target value of the controllable operation parameters (target controllable operation data).

[0068] Further, data mining technology is used to extract the operating state characteristics of the unit from historical data, and a state monitoring model of the unit is established according to the operating state characteristics for real-time monitoring of the operating state of the unit. According to the state monitoring model, the current operating state of the unit is reconstructed, and the reconstructed operating data of the target generating unit is determined as the optimal operating state data of the unit under the current working condition. Based on the historical operating data, coal quality attribute data, flue gas attribute data, and optimal operating state data, data mining technology (such as association rule mining, clustering analysis, etc.) or thermodynamic calculation methods are used to determine the energy consumption benchmark value of the unit under different working conditions (the energy consumption benchmark value can also be obtained by inputting the historical operating data, coal quality attribute data, flue gas attribute data, and optimal operating state data into the energy consumption benchmark value model for prediction). According to the real-time operating data, coal quality attribute data, flue gas attribute data, and optimal operating state data, the current real-time energy consumption value of the unit is calculated (for example, by inputting the real-time operating data, coal quality attribute data, flue gas attribute data, and optimal operating state data into the energy consumption model for prediction, the real-time energy consumption value can be obtained). The real-time energy consumption value is compared with the energy consumption benchmark value to calculate the energy consumption deviation value. Finally, the unit coal consumption, target controllable operating data, and energy consumption deviation value of the target generating unit are sent to the display terminal for display and release, which is convenient for relevant personnel to view the information in real time.

[0069] Further, the mining results of the power plant operating data can also be determined according to the historical operating data, including the overall laws of the changes of the superheater outlet pressure and temperature, the high-pressure cylinder exhaust pressure and temperature, and the intermediate-pressure cylinder exhaust pressure and temperature with the load. The results may show that the main steam pressure shows an obvious constant-sliding-constant trend with the load. The temperature at the superheater outlet hardly changes with the load. The high-pressure exhaust pressure and the intermediate-pressure exhaust pressure have a good linear relationship with the load and increase with the increase of the load. In addition, the change of the high-pressure exhaust temperature with the load is more significant than that of the intermediate-pressure exhaust temperature.

[0070] According to another method for predicting the coal consumption of a generator set provided by the present invention, compared with the current method of determining the coal consumption of a generator set by manually measuring the reduction of the original coal, the present invention obtains the real-time operation data of the target generator set, the historical operation data of the target generator set during the historical power generation process, the coal quality attribute data of the coal used by the target generator set, and the flue gas attribute data of the flue gas generated during the combustion of the coal in response to the coal consumption prediction signal of the target generator set; then, based on the historical operation data, anomaly detection is performed on the real-time operation data to obtain anomaly data, and the anomaly data in the real-time operation data is removed to obtain the cleaned real-time operation data; finally, the cleaned real-time operation data, the historical operation data, the coal quality attribute data, and the flue gas attribute data are input into a preset coal consumption prediction model for coal consumption prediction to obtain the unit coal consumption of the target generator set. Thus, by comprehensively analyzing the real-time operation data, historical operation data, coal quality attribute data, and flue gas attribute data to determine the unit coal consumption, the accuracy of predicting the coal consumption of the unit can be improved. At the same time, the present invention uses a preset coal consumption prediction model for coal consumption prediction, and the whole process does not require manual participation, avoiding the time-consuming, laborious, and measurement error situations of manually measuring the reduction of the original coal. Therefore, the present invention can improve the prediction efficiency and accuracy of the unit coal consumption. In addition, the present invention uses historical operation data to detect anomaly data in real-time operation data, which can ensure the effectiveness of the real-time operation data, thereby further improving the accuracy of predicting the coal consumption of the unit.

[0071] Further, as Figure 1 a specific implementation of Figure 3 shown, the embodiment of the present invention provides a device for predicting the coal consumption of a generator set, as

[0072] shown in the figure. The device includes: an acquisition unit 31, a detection unit 32, and a determination unit 33.

[0073] The acquisition unit 31 can be used to obtain the real-time operation data of the target generator set, the historical operation data of the target generator set during the historical power generation process, the coal quality attribute data of the coal used by the target generator set, and the flue gas attribute data of the flue gas generated during the combustion of the coal in response to the coal consumption prediction signal of the target generator set.

[0074] The determination unit 33 can be used to input the cleaned real-time operation data, the historical operation data, the coal quality attribute data, and the flue gas attribute data into a preset coal consumption prediction model to perform coal consumption prediction and obtain the unit coal consumption of the target power generation unit.

[0075] In specific application scenarios, in order to smooth the historical operation data, such as Figure 4 As shown, the device further includes: a smoothing device 34.

[0076] The smoothing device 34 can be used to take any operation data in the historical operation data as a target operation data, and calculate the distance between the target operation data and the remaining historical operation data, wherein the remaining historical operation data is the historical operation data after the target operation data is removed; determine the number of distances between the target operation data and the remaining historical operation data that are greater than a preset distance threshold, and if the distance number is greater than the preset number threshold, determine the target operation data as outlier data, and remove the outlier data in the historical operation data to obtain the cleaned historical operation data; determine the data type of the cleaned historical operation data, and determine the regression fitting model corresponding to the data type; use the regression fitting model to smooth the cleaned historical operation data to obtain the denoised historical operation data; divide the denoised historical operation data into multiple groups of historical operation data according to a preset time period; calculate the data mean corresponding to each group of historical operation data, and use the data mean to replace the maximum historical operation data in the corresponding group to obtain the smoothed historical operation data.

[0077] The detection unit 32 may be specifically configured to perform anomaly detection on the real-time operation data based on the smoothed historical operation data to obtain abnormal data.

[0078] In a specific application scenario, in order to supplement the missing data in the historical operation data, the device further includes: a supplement unit 35.

[0079] The supplement unit 35 can be used to determine the data collection time point of the historical operation data, and determine the first preset time period where the collection time point is located, the second preset time period before the collection time point, and the third preset time period after the collection time point; determine the first load change value of the target generator set within the first preset time period, the second load change value of the target generator set within the second preset time period, and the third load change value of the target generator set within the third preset time period; based on the first load change value, the second load change value, and the third load change value, determine invalid data in the historical operation data, and eliminate the invalid data in the historical operation data to obtain valid historical operation data; determine the missing data collection time point of the missing data in the valid historical operation data, and determine each neighboring data within the preset time range corresponding to the missing data collection time point; based on For each of the neighboring data, determine the second-order interpolation data of the missing data; determine the operating condition information of the target generator set at the time point of the missing data collection, obtain the to-be-matched operating condition information at the data collection time point corresponding to the valid historical operating data, and determine the target operating condition information that matches the operating condition information in the to-be-matched operating condition information; determine the matching data collection time point corresponding to the target operating condition information in the data collection time point corresponding to the valid historical operating data, and calculate the data mean and data median of the valid historical operating data corresponding to the matching data collection time point; respectively determine the weight coefficients corresponding to the data mean, the data median, and the second-order interpolation data, and based on the weight coefficients, add the data mean, the data median, and the second-order interpolation data to obtain a supplementary value, and use the supplementary value to supplement the missing data to obtain the supplemented valid historical operating data.

[0080] The detection unit 32 may be specifically configured to perform anomaly detection on the real-time operation data based on the supplemented valid historical operation data to obtain abnormal data.

[0081] In a specific application scenario, in order to perform anomaly detection on real-time running data, the detection unit 32 includes a first determination module 321 , a detection module 322 , and a correction module 323 .

[0082] The first determination module 321 can be used to determine a reference value and a preset deviation range corresponding to the real-time operation data based on the historical operation data. The method for determining the reference value and the preset deviation range corresponding to the real-time operation data based on the historical operation data includes: determining the data types of the historical operation data, dividing each data type into a preset number of data intervals based on the magnitudes of the historical operation data in each data type; taking any two data types in each data type as a first target data type and a second target data type respectively, and taking any data interval in the data interval corresponding to the first target data type as a first target interval, and taking any data interval in the data interval corresponding to the second target data type as a second target interval, and determining a first probability value that the historical operation data in the first target data type falls within the first target interval; determining a second probability value that the historical operation data in the second target data type falls within the second target interval when the historical operation data in the first target data type falls within the first target interval; determining the correlation degree between the historical operation data in the first target data type and the historical operation data in the second target data type based on the first probability value and the second probability value; determining a first reference value corresponding to the first target data type based on the correlation degree of the historical operation data in the first target data type, determining the same-type real-time operation data of the same type as the first target data type in the real-time operation data, and determining the first reference value as the reference value of the same-type real-time operation data; determining the preset deviation range of the same-type real-time operation data based on the reference value of the same-type real-time operation data.

[0083] The detection module 322 can be used to perform anomaly detection on the real-time operation data based on the preset deviation range to obtain the abnormal data in the real-time operation data.

[0084] The correction module 323 can be used to correct the abnormal data by using the reference value to obtain the corrected real-time operation data.

[0085] In a specific application scenario, in order to perform reduction processing on the historical operation data, the device further includes: a reduction unit 36.

[0086] The specification unit 36 can be used to initialize the centroid vectors corresponding to the centroids of different clusters, and determine the data feature vectors corresponding to the historical operation data; based on the data feature vectors and the centroid vectors, calculate the distances between each of the historical operation data and the centroids corresponding to each of the clusters, and based on the distances, divide each of the historical operation data into each of the clusters; based on the data feature vectors corresponding to the historical operation data in each of the clusters, determine the updated centroid vectors corresponding to each of the clusters; based on the updated centroid vectors, re-divide each of the historical operation data into each of the clusters until the updated centroid vectors do not change, and determine the historical operation data finally divided into each of the clusters as the historical operation data under different clustering categories; determine the central operation data and the boundary operation data in each of the clustering categories, and use each of the central operation data and each of the boundary operation data to replace the historical operation data to obtain the specified historical operation data.

[0087] The determination unit 33 can specifically be used to input the cleaned real-time operation data, the specified historical operation data, the coal quality attribute data, and the flue gas attribute data into a preset coal consumption prediction model to perform coal consumption prediction, and obtain the unit coal consumption of the target generating unit.

[0088] In a specific application scenario, in order to display information on a display terminal, the device further includes: a display unit 37.

[0089] The display unit 37 can be used to determine the target controllable operation data of the target generating unit during stable operation based on the historical operation data; reconstruct the operation state of the target generating unit based on the historical operation data, and determine the reconstructed operation data of the reconstructed target generating unit; determine the energy consumption deviation value of the target generating unit based on the cleaned real-time operation data, the historical operation data, the coal quality attribute data, the flue gas attribute data, and the reconstructed operation data; send the unit coal consumption, the target controllable operation data, and the energy consumption deviation value to the display terminal for display.

[0090] In a specific application scenario, in order to determine the unit coal consumption of the target generating unit, the determination unit 33 includes a second determination module 331, an intersection module 332, and a prediction module 333.

[0091] The second determination module 331 can be used to determine the real-time feature vector corresponding to the cleaned real-time operation data, the historical feature vector corresponding to the historical operation data, the coal quality feature vector corresponding to the coal quality attribute data, and the flue gas feature vector corresponding to the flue gas attribute data.

[0092] The cross module 332 may be used to perform cross processing on the real-time feature vector, the historical feature vector, the coal quality feature vector, and the flue gas feature vector to obtain a coal consumption cross vector.

[0093] The prediction module 333 may be used to input the coal consumption cross vector into a preset coal consumption prediction model to perform coal consumption prediction and obtain the unit coal consumption of the target power generation unit.

[0094] It should be noted that for other corresponding descriptions of the functional modules involved in the device for estimating coal consumption of a generator set provided in an embodiment of the present invention, reference can be made to Figure 1 The corresponding description of the method shown will not be repeated here.

[0095] Based on the above Figure 1 The method shown, accordingly, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the following steps are implemented: in response to a coal consumption estimation signal of a target power generation unit, the real-time operation data of the target power generation unit, the historical operation data of the target power generation unit in the historical power generation process, the coal quality attribute data of the coal used by the target power generation unit, and the flue gas attribute data of the flue gas generated during the combustion of the coal are obtained; based on the historical operation data, anomaly detection is performed on the real-time operation data to obtain abnormal data, and the abnormal data in the real-time operation data is eliminated to obtain the cleaned real-time operation data; the cleaned real-time operation data, the historical operation data, the coal quality attribute data, and the flue gas attribute data are input into a preset coal consumption prediction model to perform coal consumption prediction to obtain the unit coal consumption of the target power generation unit.

[0096] Based on the above Figure 1 The method shown and Figure 3 The embodiment of the device shown in the figure, the embodiment of the present invention also provides a physical structure diagram of a computer device, such as Figure 5As shown in the figure, the computer device includes: a processor 41, a memory 42, and a computer program stored on the memory 42 and executable on the processor. Both the memory 42 and the processor 41 are provided on a bus 43. When the processor 41 executes the program, the following steps are implemented: in response to a coal consumption estimation signal of a target generating set, obtain the real-time operation data of the target generating set, the historical operation data of the target generating set during historical power generation, the coal quality attribute data of the coal used by the target generating set, and the flue gas attribute data of the flue gas generated during the combustion of the coal; based on the historical operation data, perform anomaly detection on the real-time operation data to obtain anomaly data, and remove the anomaly data in the real-time operation data to obtain the cleaned real-time operation data; input the cleaned real-time operation data, the historical operation data, the coal quality attribute data, and the flue gas attribute data into a preset coal consumption prediction model for coal consumption prediction to obtain the unit coal consumption of the target generating set.

[0097] Through the technical solution of the present invention, the present invention obtains the real-time operation data of the target generating set, the historical operation data of the target generating set during historical power generation, the coal quality attribute data of the coal used by the target generating set, and the flue gas attribute data of the flue gas generated during the combustion of the coal in response to a coal consumption estimation signal of the target generating set; then, based on the historical operation data, perform anomaly detection on the real-time operation data to obtain anomaly data, and remove the anomaly data in the real-time operation data to obtain the cleaned real-time operation data; finally, input the cleaned real-time operation data, the historical operation data, the coal quality attribute data, and the flue gas attribute data into a preset coal consumption prediction model for coal consumption prediction to obtain the unit coal consumption of the target generating set. Thus, by comprehensively analyzing the real-time operation data, historical operation data, coal quality attribute data, and flue gas attribute data to determine the unit coal consumption, the estimation accuracy of the unit coal consumption can be improved. At the same time, the present invention uses a preset coal consumption prediction model for coal consumption prediction, and the whole process does not require manual participation, avoiding the time-consuming and laborious situation of manually measuring the original coal reduction amount and the situation of measurement errors. Therefore, the present invention can improve the estimation efficiency and estimation accuracy of the unit coal consumption. In addition, the present invention uses historical operation data to detect anomaly data in real-time operation data, which can ensure the effectiveness of real-time operation data, thereby further improving the estimation accuracy of the unit coal consumption.

[0098] Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. Optionally, they can be implemented by program codes executable by the computing device. Thus, they can be stored in a storage device and executed by the computing device. And in some cases, the steps shown or described can be executed in a sequence different from here, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module to be implemented. In this way, the present invention is not limited to any specific combination of hardware and software.

[0099] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for estimating coal consumption of a power generation unit, characterized in that: include: In response to the coal consumption estimation signal of the target power generation unit, real-time operation data of the target power generation unit, historical operation data of the target power generation unit in the historical power generation process, coal quality attribute data of the coal used by the target power generation unit, and flue gas attribute data of the flue gas generated during the combustion of the coal are acquired; Based on the historical operation data, performing anomaly detection on the real-time operation data to obtain abnormal data, and removing the abnormal data from the real-time operation data to obtain cleaned real-time operation data; The cleaned real-time operation data, the historical operation data, the coal quality attribute data, and the flue gas attribute data are input into a preset coal consumption prediction model to perform coal consumption prediction, and obtain the unit coal consumption of the target power generation unit.

2. The method according to claim 1, characterized in that Before performing anomaly detection on the real-time operation data based on the historical operation data to obtain abnormal data, the method further includes: Taking any operation data in the historical operation data as a target operation data, respectively, and calculating the distance between the target operation data and the remaining historical operation data, wherein the remaining historical operation data is the historical operation data after removing the target operation data; Determine the number of distances between the target operation data and the remaining historical operation data that are greater than a preset distance threshold, and if the distance number is greater than the preset number threshold, determine the target operation data as outlier data, and remove the outlier data from the historical operation data to obtain the cleaned historical operation data; Determine the data type of the cleaned historical operation data, and determine a regression fitting model corresponding to the data type; Using the regression fitting model to smooth the cleaned historical operation data to obtain the denoised historical operation data; Dividing the denoised historical operation data into multiple groups of historical operation data according to a preset time period; Calculate the data mean corresponding to each group of historical operation data respectively, and use the data mean to replace the maximum historical operation data in the corresponding group to obtain the smoothed historical operation data; The performing anomaly detection on the real-time operation data based on the historical operation data to obtain abnormal data includes: Based on the smoothed historical operation data, anomaly detection is performed on the real-time operation data to obtain abnormal data.

3. The method according to claim 1, characterized in that Before performing anomaly detection on the real-time operation data based on the historical operation data to obtain abnormal data, the method further includes: Determine a data collection time point of the historical operation data, and determine a first preset time period within which the collection time point is located, a second preset time period before the collection time point, and a third preset time period after the collection time point; Determine a first load change value of the target generator set within the first preset time period, a second load change value of the target generator set within the second preset time period, and a third load change value of the target generator set within the third preset time period; Based on the first load change value, the second load change value, and the third load change value, determining invalid data in the historical operation data, and removing the invalid data in the historical operation data to obtain valid historical operation data; Determine the missing data collection time point of the missing data in the valid historical operation data, and determine each neighboring data within a preset time range corresponding to the missing data collection time point; Determine the second-order interpolation data of the missing data based on each of the neighboring data; Determine the operating condition information of the target generator set at the missing data collection time point, obtain the to-be-matched operating condition information at the data collection time point corresponding to the valid historical operation data, and determine the target operating condition information that matches the operating condition information in the to-be-matched operating condition information; Determine the matching data collection time point corresponding to the target operating condition information in the data collection time point corresponding to the valid historical operating data, and calculate the data mean and data median of the valid historical operating data corresponding to the matching data collection time point; Determine the weight coefficients corresponding to the data mean, the data median, and the second-order interpolation data respectively, add the data mean, the data median, and the second-order interpolation data based on the weight coefficients to obtain a supplementary value, and use the supplementary value to supplement the missing data to obtain the supplemented valid historical operation data; The performing anomaly detection on the real-time operation data based on the historical operation data to obtain abnormal data includes: Based on the supplemented valid historical operation data, anomaly detection is performed on the real-time operation data to obtain abnormal data.

4. The method according to claim 1, characterized in that: The performing anomaly detection on the real-time operation data based on the historical operation data to obtain abnormal data includes: Based on the historical operation data, determine the benchmark value and preset deviation range corresponding to the real-time operation data, wherein the method for determining the benchmark value and preset deviation range corresponding to the real-time operation data based on the historical operation data includes: determining the data type of each of the historical operation data, and dividing each of the data types into a preset number of data intervals based on the size of each historical operation data in each of the data types; using any two data types in each of the data types as the first target data type and the second target data type, and using any data interval in the data interval corresponding to the first target data type as a first target interval, and using any data interval in the data interval corresponding to the second target data type as a second target interval, and determining that the historical operation data in the first target data type falls within the first target interval a first probability value; determining a second probability value that when the historical operating data in the first target data type falls within the first target interval, the historical operating data in the second target data type falls within the second target interval; determining the degree of association between the historical operating data in the first target data type and the historical operating data in the second target data type based on the first probability value and the second probability value; determining a first reference value corresponding to the first target data type based on the degree of association of the historical operating data in the first target data type, and determining the same type of real-time operating data of the same type as the first target data type in the real-time operating data, and determining the first reference value as the reference value of the same type of real-time operating data; determining a preset deviation range of the same type of real-time operating data based on the reference value of the same type of real-time operating data; Based on the preset deviation range, performing anomaly detection on the real-time operation data to obtain abnormal data in the real-time operation data; After performing anomaly detection on the real-time operation data based on the historical operation data to obtain abnormal data, the method further includes: The abnormal data is corrected using the reference value to obtain the corrected real-time operation data.

5. The method according to claim 1, characterized in that Before inputting the cleaned real-time operation data, the historical operation data, the coal quality attribute data, and the flue gas attribute data into a preset coal consumption prediction model to perform coal consumption prediction and obtain the unit coal consumption of the target power generation unit, the method further includes: Initializing the centroid vectors of the centroids corresponding to different clusters, and determining the data feature vectors corresponding to the historical operation data; Based on the data feature vector and the centroid vector, calculate the distance between each of the historical operation data and the corresponding centroid of each cluster, and divide each of the historical operation data into each of the clusters based on the distance; Determine an updated centroid vector corresponding to each cluster based on a data feature vector corresponding to the historical operation data in each cluster; Based on the updated centroid vector, each of the historical operation data is re-divided into each of the clusters until the updated centroid vector does not change, and the historical operation data finally divided into each of the clusters is determined as historical operation data under different clustering categories; Determine the central operating data and the boundary operating data in each of the cluster categories, and use each of the central operating data and each of the boundary operating data to replace the historical operating data to obtain the reduced historical operating data; The cleaned real-time operation data, the historical operation data, the coal quality attribute data, and the flue gas attribute data are input into a preset coal consumption prediction model to perform coal consumption prediction to obtain the unit coal consumption of the target power generation unit, including: The cleaned real-time operation data, the standardized historical operation data, the coal quality attribute data, and the flue gas attribute data are input into a preset coal consumption prediction model to perform coal consumption prediction, and obtain the unit coal consumption of the target power generation unit.

6. The method according to claim 1, characterized in that The method further comprises: Based on the historical operation data, determining target controllable operation data of the target generator set during stable operation; Based on the historical operation data, reconstructing the operation state of the target generator set, and determining the reconstructed operation data of the target generator set after the reconstruction; Determine the energy consumption deviation value of the target power generation group based on the real-time operation data after the cleaning, the historical operation data, the coal quality attribute data, the flue gas attribute data, and the reconstructed operation data; The unit coal consumption, the target controllable operation data, and the energy consumption deviation value are sent to a display terminal for display.

7. The method according to claim 1, characterized in that The cleaned real-time operation data, the historical operation data, the coal quality attribute data, and the flue gas attribute data are input into a preset coal consumption prediction model to perform coal consumption prediction to obtain the unit coal consumption of the target power generation unit, including: Determine a real-time feature vector corresponding to the real-time operation data after cleaning, a historical feature vector corresponding to the historical operation data, a coal quality feature vector corresponding to the coal quality attribute data, and a flue gas feature vector corresponding to the flue gas attribute data; Cross-processing the real-time feature vector, the historical feature vector, the coal quality feature vector, and the flue gas feature vector to obtain a coal consumption cross-vector; The coal consumption cross vector is input into a preset coal consumption prediction model to perform coal consumption prediction, and the unit coal consumption of the target power generation unit is obtained.

8. A device for estimating coal consumption of a power generation unit, characterized in that: include: an acquisition unit, configured to acquire, in response to a coal consumption estimation signal of a target power generation unit, real-time operation data of the target power generation unit, historical operation data of the target power generation unit in a historical power generation process, coal quality attribute data of the coal used by the target power generation unit, and flue gas attribute data of flue gas generated during the combustion of the coal; a detection unit, configured to perform anomaly detection on the real-time operation data based on the historical operation data to obtain abnormal data, and remove the abnormal data from the real-time operation data to obtain cleaned real-time operation data; A determination unit is used to input the cleaned real-time operation data, the historical operation data, the coal quality attribute data, and the flue gas attribute data into a preset coal consumption prediction model to perform coal consumption prediction and obtain the unit coal consumption of the target power generation unit.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images