Public building energy consumption quota method based on multistage classification and dynamic data correction

Abstract

The invention discloses a public building energy consumption quota method based on multi-level classification and dynamic data correction, and the method comprises the steps: obtaining energy consumption parameters, constructing a multi-feature data set, and carrying out the preprocessing of the multi-feature data set; performing first-level classification based on functionality and energy consumption attributes; performing secondary classification based on a K-means model; based on a quota level and sorting coupling method and a simulation quota method, energy consumption quota and weighted average are carried out respectively, and a comprehensive energy consumption quota is obtained; performing energy consumption characteristic importance analysis based on an SHAP value model; linear regression is carried out based on the annual data of the important features, and an important feature prediction value and a change rate are calculated and further corrected to obtain a comprehensive energy consumption quota of various types of buildings in the future; based on the change rate of the important characteristics of the single building, the energy utilization rationality and the energy-saving enthusiasm are judged, and the comprehensive energy consumption quota is corrected. According to the method, the two energy consumption quotas of the multi-level classification building are corrected based on the dynamic data, and the influence of neglecting different behaviors, properties and characteristics of the building on the energy consumption of the building in a conventional energy consumption quota method is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of computer vision, and in particular relates to a method for public building energy consumption quota based on multi-level classification and dynamic data correction. Background Art

[0002] It is necessary to improve the energy efficiency of public buildings during their operation. The energy consumption quota refers to the upper limit of the total amount of energy that a building needs to consume to maintain its equipment, facilities, and sanitary conditions during the quota period, and is a yardstick for measuring the level of building energy consumption.

[0003] A scientific and reasonable energy consumption quota method is the basis for the evaluation and assessment of the energy management level of public institutions, and is also an important means to achieve the "dual control" indicators of the total energy consumption and intensity of public institutions. However, differences in operation modes and building characteristics between different buildings limit the rationality of the energy consumption benchmark value of a single building, and the benchmark value needs to be revised according to the characteristics of different public buildings. This requires the country to establish a huge database and conduct statistics and evaluation on a large number of historical operation data of different types of public buildings. Although there are many energy consumption quota methods at present, the energy consumption of the same type of buildings in China is affected by factors such as geographical location, climate, and energy users, and there are still many unreasonable aspects in the current energy consumption quota methods. How to most reasonably divide the public building collection and eliminate the impact of different factors on building energy consumption is worth considering. Summary of the invention

[0004] The main purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art, and to provide a method for public building energy consumption quota based on multi-level classification and dynamic data correction, to obtain a comprehensive energy consumption quota value based on a combination of multi-level public building type classification and quota level and sorting coupling method and simulation quota method, and to reasonably correct the comprehensive energy consumption quota value based on dynamic data, so as to provide a more reasonable energy consumption quota method for government building energy consumption quota setters.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] One aspect of the present invention provides a method for public building energy consumption quota based on multi-level classification and dynamic data correction, comprising the following steps:

[0007] Obtain monthly energy consumption parameters of different types of public buildings over several years, including date parameters, load parameters, and energy-saving measures parameters;

[0008] Construct a multi-feature data set based on the acquired energy consumption parameters and perform data verification and cleaning, data feature screening, and data normalization;

[0009] Perform primary logical classification and labeling of multi-feature datasets based on the functionality and energy attributes of public buildings;

[0010] Based on the K-means model, the results of the first-level logic and classification labels are classified into the second level, and the results of the second-level classification are logically interpreted and labeled according to the feature information;

[0011] Based on the quota level and ranking coupling method and the simulation quota method, energy consumption quotas are respectively carried out for public buildings of the same secondary type, and the quota values ​​obtained by the two methods are weighted averaged to obtain the comprehensive energy consumption quota;

[0012] Based on the SHAP value model, the importance of annual energy consumption characteristics of different types of public buildings is analyzed, and the important characteristics and the Shapley value weights of the important characteristics are obtained;

[0013] Perform linear regression based on the annual data average of the important characteristics of each type of public building, and predict the average value and change rate of the important characteristics of each type of public building in the next few years based on the regression curve;

[0014] Based on the change rate of important characteristics and the weighted correction of the Shapley value of important characteristics, the comprehensive energy consumption quota of various types of buildings in the next few years is obtained;

[0015] Based on the monthly data of important characteristics of a single building, the rate of change in the past few years is obtained by linear regression to judge the rationality of energy use and the enthusiasm for energy conservation of a single building, and to correct the comprehensive energy consumption quota of a single building.

[0016] As a preferred technical solution, the different types of public buildings include educational buildings, administrative buildings and medical buildings; the date parameters include date, year, month and quarter; the load parameters include per capita water consumption, building area, number of energy users, number of staff, number of gasoline vehicles, number of diesel vehicles, number of new energy vehicles, electricity consumption, coal consumption, natural gas consumption, oil consumption, other oil consumption, diesel consumption, liquefied petroleum consumption, other energy consumption, solar collector area, solar system installed capacity and geothermal system installed capacity; the energy-saving measures parameters include whether energy-saving renovations have been carried out, the time from the energy-saving renovation to the present, and the energy-saving renovation projects that have been carried out.

[0017] As a preferred technical solution, the data verification and cleaning are specifically as follows: the quartile method is used to perform data verification and cleaning on the acquired energy consumption parameters, specifically: Q3+1.5IQR of the distribution interval is taken as the upper threshold, and Q1-1.5IQR is taken as the lower threshold; wherein Q3 is the third quartile of the distribution interval, that is, the number at the 75% position after the data sequence is arranged from small to large; Q1 is the first quartile of the distribution interval, that is, the number at the 25% position after the data sequence is arranged from small to large; IQR is the interquartile range, IQR=Q3-Q1, and values ​​other than Q1~Q3 are discarded as abnormal values;

[0018] The data feature screening is specifically as follows: using the random forest method to analyze the correlation between the per capita energy consumption and unit area energy consumption of public buildings and each influencing factor in the influencing factor sequence S, screening out the influencing factors with a correlation higher than a preset value, and deleting the influencing factors with a correlation lower than the preset value; wherein the influencing factor sequence S = [year year, month month, quarter quarter, per capita water consumption PWC, building area BA, number of energy users NEU, number of staff H, number of gasoline vehicles NGV, number of diesel vehicles NDV, oil consumption GC, other oil consumption OC, diesel consumption DC, liquefied petroleum consumption CLP, other energy consumption OEC, solar collector area SCA, solar system installed capacity SSIC, geothermal system installed capacity GIC, energy-saving transformation ESR];

[0019] The normalization process of the data is specifically as follows:

[0020]

[0021] Among them, X is the data to be processed, X max is the maximum value in the data set, X min is the minimum value in the data set, X n is the normalized data.

[0022] As a preferred technical solution, the K-means model is used to perform secondary classification on the results of the primary logic and classification labels, and the secondary classification results are logically interpreted and labeled according to the feature information, specifically:

[0023] Based on the K-means model, the multi-feature data sets of different types of public buildings are clustered, and the K value of different types of public building samples is determined according to the "elbow method"; the "elbow method" is specifically: calculating the sum of squares of errors within the cluster, and using the image inflection point to find the optimal number of classifications, and then performing cluster analysis based on the K-means model and input data, and the K value is the number of subcategories after the secondary classification of different types of public buildings; the K-means model inputs energy consumption data directly related to annual energy consumption, including annual energy consumption, annual per capita energy consumption, annual unit area energy consumption, annual electricity consumption, annual oil consumption, and annual natural gas consumption;

[0024] The results of the secondary classification of different types of public buildings are logically explained, specifically: the specific locations of all types of public buildings are marked on the map and analyzed, and the unknown marked clusters of public buildings of the same secondary type are described to obtain the labels of the secondary type of public buildings.

[0025] As a preferred technical solution, the comprehensive energy consumption quota calculation method is as follows:

[0026]

[0027] Among them, EUI C is the comprehensive energy consumption quota value, EUI o The EUI is the original comprehensive energy consumption quota calculated based on the original data using the quota level and ranking coupling method. p It is the simulated comprehensive energy consumption quota calculated by the simulated quota method;

[0028] The specific method of coupling quota level and ranking is as follows:

[0029] Determine whether a certain type of sample conforms to the normal distribution. If it conforms, the quota level method is used to calculate the comprehensive energy consumption quota of the original data, and 20%, 50% and 80% of the quota level are selected as the constraint value, benchmark value and guide value of the energy consumption quota; if it does not conform, sort the data in the sample, and select 20%, 50% and 80% of the sorting ratio as the constraint value, benchmark value and guide value of the energy consumption quota;

[0030] The simulation quota method is as follows:

[0031] Divide the multi-feature dataset into training and test sets in proportion;

[0032] In the training phase, training is performed based on the prediction models, and the hyperparameters of each model are adjusted through optimization algorithms. The prediction models include stepwise linear regression, support vector machine, random forest, gradient boosting method and BP neural network, and the optimization algorithms include nonlinear weight-decreasing particle swarm optimization algorithm, whale algorithm and sparrow algorithm.

[0033] During the testing phase, the annual per capita energy consumption and annual unit area energy consumption of each type of public building are predicted respectively. When the prediction results are evaluated by the preset indicators, the energy consumption prediction model is considered to be trained. Through indicator comparison, the prediction model and optimization algorithm suitable for the current type of public building are selected, and the simulated comprehensive energy consumption quota of each type of public building is predicted.

[0034] As a preferred technical solution, the SHAP value model is used to perform a feature vector correlation analysis on the annual energy consumption of various types of public buildings, specifically:

[0035] Based on the SHAP value model, the characteristic contribution of energy consumption parameters of different types of public buildings is calculated to obtain the contribution order results of all characteristics, specifically: calculate the marginal contribution of a feature when it is added to the model, and consider the different marginal contributions of the feature in all feature sequences, take the average, and obtain the Shapley value weight of the feature;

[0036] Important features are screened out based on the interaction between features and the contribution order of features that are sensitive to the results.

[0037] As a preferred technical solution, the weighted correction based on the change rate of important features and the Shapley value weight of important features is used to obtain the comprehensive energy consumption quota of various types of buildings in the next few years, specifically:

[0038] PCEC next =PCEC now *(a 1 *ω 1 +a 2 *ω 2 +a 3 *ω 3 );

[0039] Among them, PCEC next PCEC is the per capita energy consumption quota of such buildings in a future year. now represents the current per capita energy consumption quota for such buildings, a 1 Represents the proportion of the first eigenvector decreasing after several years, ω 1 Represents the proportion of the Shapley value of the first eigenvector to the total Shapley values ​​of the three eigenvectors, and so on.

[0040] As a preferred technical solution, the monthly data based on the important characteristics of a single building is used to perform linear regression to obtain the change rate in the past few years, judge the rationality of energy use and energy-saving enthusiasm of the single building, and correct the comprehensive energy consumption quota of the single building, specifically:

[0041] If there are abnormal data points in the change curve of the important characteristics of a building in the past few years that exceed 20% of the normal fluctuation range, or the characteristic value change trend is increasing and the increase is greater than 20%, it is judged as unreasonable energy use and the comprehensive energy consumption quota is reduced by 5%; if the change trend of the important characteristics of a building in the past few years is decreasing and the decrease is greater than 20%, it is judged as an active response to the energy conservation and emission reduction policy and the comprehensive energy consumption quota is increased by 5%.

[0042] Another aspect of the present invention further provides a system for public building energy consumption quota based on multi-level classification and dynamic data correction, which is applied to the above-mentioned method for public building energy consumption quota based on multi-level classification and dynamic data correction, including a data acquisition and processing module, a data preprocessing module, a building type primary classification module, a building type secondary classification module, an energy consumption quota module, a feature importance analysis module and an energy consumption quota value correction module;

[0043] The data acquisition and processing module is used to obtain monthly and annual energy consumption parameters of different types of buildings, including date parameters, load parameters and energy-saving measures parameters; and construct a multi-feature data set based on the obtained energy consumption parameters;

[0044] The data preprocessing module is used to perform data verification and cleaning, data feature screening and data normalization on the multi-feature data set;

[0045] The building type primary classification module is used to classify and label the sample set according to the functionality and energy properties of public buildings;

[0046] The building type secondary classification module is used to perform secondary classification on the results of the primary logic and classification labels according to the K-means model, and to logically interpret and label the results of the secondary classification according to the feature information;

[0047] The energy consumption quota module is used to perform energy consumption quotas on public buildings of the same secondary type according to the quota level and ranking coupling method and the simulation quota method, and to perform weighted average of the quota values ​​obtained by the two methods to obtain a comprehensive energy consumption quota;

[0048] The feature importance analysis module is used to perform annual energy consumption feature importance analysis of different types of public buildings based on the SHAP value model according to energy consumption parameters, energy consumption prediction model and energy consumption prediction results, and obtain important features and Shapley value weights of important features;

[0049] The energy consumption quota value correction module is used to perform linear regression based on the annual data average value of the important characteristics of each type of public building, and predict the average value and change rate of the important characteristics of each type of public building in the next few years based on the regression curve;

[0050] Based on the change rate of important characteristics and the weighted correction of the Shapley value of important characteristics, the comprehensive energy consumption quota of various types of buildings in the next few years is obtained;

[0051] Based on the monthly data of important characteristics of a single building, the rate of change in the past few years is obtained by linear regression to judge the rationality of energy use and the enthusiasm for energy conservation of a single building, and to correct the comprehensive energy consumption quota of a single building.

[0052] Another aspect of the present invention provides a storage medium storing a program, which, when executed by a processor, implements the above-mentioned method of public building energy consumption quota based on multi-level classification and dynamic data correction.

[0053] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0054] (1) The present invention realizes energy consumption quotas for unit area energy consumption and per capita energy consumption of different types of public buildings in the field of public building energy consumption quota formulation, considers more energy consumption influencing factors and conducts multi-level reasonable classification, and the quota value is more reasonable; it can predict the energy consumption quota value of future years based on dynamic data, and provide direction for whether different types of public institution buildings need energy-saving transformation, providing support for building energy management and energy conservation; the present invention can automatically complete data verification and filling missing data, complete the construction of a valid data set, and has fast processing speed and convenient operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a flow chart of a method for correcting energy consumption quota of public buildings based on multi-level classification and dynamic data in an embodiment of the present invention;

[0056] Figure 2 It is a flowchart of the primary classification and secondary classification method of public buildings in an embodiment of the present invention;

[0057] Figure 3 It is a schematic diagram of the process of setting energy consumption quotas for various types of buildings in an embodiment of the present invention;

[0058] Figure 4 A schematic diagram of a process for correcting a comprehensive energy consumption quota based on dynamic data according to an embodiment of the present invention;

[0059] Figure 5 It is a structural schematic diagram of a system for correcting public building energy consumption quota based on dynamic data according to an embodiment of the present invention;

[0060] Figure 6 It is a schematic diagram of the structure of the storage medium of an embodiment of the present invention. DETAILED DESCRIPTION

[0061] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present application.

[0062] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0063] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0064] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0065] like Figure 1 As shown, the first embodiment of the present invention discloses a method for correcting the energy consumption quota of public buildings based on multi-level classification and dynamic data, comprising the following steps:

[0066] S1. Obtain annual and monthly energy consumption parameters of different types of public buildings to be energy-consumed over several years, including date parameters, load parameters, and energy-saving measure parameters.

[0067] As one or more implementation modes, the obtained public buildings to be determined include at least educational buildings, administrative buildings, stadium buildings and medical buildings; the date parameters include at least date Date, year year, month month and quarter quarter; the load parameters include at least per capita water consumption PWC, building area BA, number of energy users NEU, number of staff H, number of gasoline vehicles NGV, number of diesel vehicles NDV, number of new energy vehicles NNV, power consumption EC, coal consumption CC, natural gas consumption NGC, oil consumption GC, other oil consumption OC, diesel consumption DC, liquefied petroleum consumption CLP, other energy consumption OEC, solar collector area SCA, solar energy system installed capacity SSIC and geothermal system installed capacity GIC. The energy-saving measure parameters include at least whether energy-saving transformation has been carried out (what kind of energy-saving transformation) ESR, how long it has been since the energy-saving transformation, and the specific energy-saving transformation projects that have been carried out. The power consumption and water consumption information are obtained through the electricity meter and water meter, and other energy consumption information and energy-saving transformation information are obtained through manual reporting.

[0068] S2. Construct a multi-feature data set based on the acquired energy consumption parameters and perform data verification and cleaning, data feature screening, and data normalization.

[0069] Furthermore, in this embodiment, the step of constructing a multi-feature data set according to the acquired energy consumption parameters is as follows:

[0070] S21. Since the acquired related parameters may contain abnormal data and missing data, in this embodiment, the load parameters need to be checked for abnormal data to avoid abnormal values, and the missing data are filled using the advance calculation method; for date data, more dimensional features need to be extracted; for energy-saving measures data, it is necessary to judge and correct it in combination with the changes in energy consumption in previous years. The details are as follows:

[0071] S22. Data verification and cleaning: The quartile method is used to perform data verification and cleaning on the obtained energy consumption load parameters, specifically: Q3+1.5IQR of the distribution interval is taken as the upper threshold, and Q1-1.5IQR is taken as the lower threshold; wherein, Q3 is the third quartile of the distribution interval, that is, the number at the 75% position after the data sequence is arranged from small to large; Q1 is the first quartile of the distribution interval, that is, the number at the 25% position after the data sequence is arranged from small to large; IQR is the interquartile range, IQR=Q3-Q1, and values ​​outside Q1~Q3 are discarded as abnormal values.

[0072] The year, month and quarter are extracted for subsequent data analysis and processing to provide more accurate and detailed energy consumption changes and trends.

[0073] Finally, the format of the verified data is unified to ensure the standardization and consistency of the data. For Boolean data, "0" represents "False" and "1" represents "True". The final sequence S contains the following elements:

[0074] S=[year,month,quarter,PWC,BA,NEU,H,NGV,NDV,GC,OC,DC,CLP,OEC,SCA,SSIC,GIC,ESR].

[0075] In particular, for abnormal situations where historical data is missing, alternative data is generated based on the historical data of the same type of energy-consuming units in the same period or the historical data of the target energy-consuming units in the previous period to fill in the missing reported values; for situations where data to be verified is missing, the missing reported values ​​for the current month are filled in after verification with the target energy-consuming units.

[0076] S23. Data feature screening:

[0077] Since the correlation between the per capita energy consumption and per unit area energy consumption of public buildings and the elements in the sequence S is different, in order to obtain more accurate prediction results, the per capita energy consumption and per unit area energy consumption are predicted separately. Therefore, for the two energy consumption intensities, it is necessary to select data with greater correlation as the input of the model, which can reduce the difficulty of model training and improve the accuracy of the model. Specifically, the following steps are included:

[0078] The random forest model is used to analyze the correlation between the per capita energy consumption and unit area energy consumption of public buildings and the influencing factors in the sequence S, screen out the influencing factors with correlation higher than the preset value, and delete the influencing factors with correlation lower than the preset value.

[0079] S24. Data normalization:

[0080] In order to eliminate the impact of dimensions and magnitudes between different features, make the impact of each feature on the result more balanced, and avoid some features dominating because of too large or too small value range, thereby reducing the accuracy of the model, normalizing all data can make the contribution of each feature more equal and improve the accuracy of the model. The specific formula is as follows:

[0081]

[0082] Among them, X is the data to be processed, X max is the maximum value in the data set, X min is the minimum value in the data set, X n is the normalized data.

[0083] S3. Conduct primary logical classification and labeling of various types of public buildings based on their functionality and energy usage attributes.

[0084] like Figure 2 As shown, as one or more implementation modes, this embodiment performs primary classification and labeling on the sample set based on the functionality and energy consumption attributes of public buildings, such as dividing educational buildings into kindergartens, elementary schools, junior high schools, high schools and adult colleges, dividing medical buildings into primary hospitals, secondary hospitals and tertiary hospitals, and dividing administrative buildings into administrative buildings using split air conditioners, administrative buildings using centralized air conditioners and other administrative buildings.

[0085] S4. Perform secondary classification on the results of primary logic and classification labels based on the K-means model, and logically interpret and label the results of secondary classification based on feature information.

[0086] As one or more implementation modes, this embodiment clusters different types of public building sample sets based on the K-means model. The model inputs energy consumption data directly related to annual energy consumption, such as annual energy consumption, annual per capita energy consumption, annual unit area energy consumption, annual electricity consumption, annual oil consumption, annual natural gas consumption, etc., and determines the K value of different types of public building samples according to the "elbow method", that is, calculating the sum of squared errors (SSE) within the cluster, and using the image inflection point to find the optimal number of classifications, and then performing cluster analysis based on the model and the input data. The K value is the number of subcategories after the secondary classification of different types of public buildings.

[0087] Furthermore, the secondary classification results of public building types are logically interpreted, the specific locations of all types of public buildings are marked and analyzed on the map, and the unknown marked clusters of public buildings of the same secondary type are described to obtain the labels of secondary type public buildings, such as district boundary primary school, district center primary school, etc.

[0088] S5. Energy consumption quotas are set for public buildings of the same secondary type based on the quota level and ranking coupling method.

[0089] Furthermore, the energy consumption quota for the original data of the same secondary type of public buildings is as follows:

[0090] like Figure 3 As shown, the quota level and sorting coupling method is to first determine whether a certain type of sample conforms to the normal distribution or is close to it. If it conforms, the quota level method is used to set the energy consumption quota for the sample, and 20%, 50% and 80% of the quota level are selected as the constraint value, benchmark value and guide value of the energy consumption quota. If it does not conform. The data in the sample are sorted, and the proportions of 20%, 50% and 80% in the sorting are selected as the constraint value, benchmark value and guide value of the energy consumption quota.

[0091] S6. Based on the simulation quota method, energy consumption quotas are set for public buildings of the same secondary type.

[0092] The quota method is simulated to divide the multi-feature data set into a training set and a test set. In one or more preferred embodiments, 70% of the data in the multi-feature data set is used to construct a training set, and 30% of the data in the multi-feature data set is used to construct a test set.

[0093] In the training stage, training is performed based on the prediction models respectively, and the hyperparameters of each model are adjusted respectively by the optimization algorithm; preferably, in this embodiment, the prediction model includes stepwise linear regression, support vector machine, random forest, gradient boosting method and BP neural network, and the optimization algorithm includes nonlinear weight-decreasing particle swarm optimization algorithm, whale algorithm and sparrow algorithm;

[0094] During the testing phase, annual per capita energy consumption and annual unit area energy consumption of each type are predicted respectively. When the prediction results pass the preset index evaluation, the energy consumption prediction model training is considered complete. Through index comparison, the prediction model and optimization algorithm that are most suitable for the current type of public building are selected.

[0095] Predict the energy consumption indicators of various types of public buildings as prediction data based on the prediction model and optimization algorithm that are most suitable for the current type of public buildings;

[0096] S7, repeating the calculation steps of the quota level and ranking coupling method for the forecast data based on the simulated quota method;

[0097] Furthermore, the weighted average of the quota values ​​obtained by the two methods is used to obtain a comprehensive energy consumption quota, and the calculation method of the comprehensive energy consumption quota is as follows:

[0098]

[0099] Among them, EUI C The EUI is the comprehensive energy consumption quota. o EUI is the energy consumption quota value calculated based on the original data. p It is the energy consumption quota value calculated based on simulation data.

[0100] S8. Based on the SHAP value model, the characteristic vector correlation analysis of the annual energy consumption of various types of public buildings is carried out, specifically:

[0101] Based on the SHAP value method, the characteristic contribution of energy consumption parameters of different types of public buildings is calculated to obtain the contribution order results of all characteristics; the SHAP value model is an interpretable contribution analysis method based on model training, which calculates the contribution of each feature to the prediction value calculation, specifically: calculate the marginal contribution of a feature when it is added to the model, and consider the different marginal contributions of the feature in all feature sequences, take the average, and obtain the Shapley value weight of the feature;

[0102] Important features are selected based on the interaction between features and the contribution order of features that are sensitive to the results.

[0103] S9. Analyze the annual and monthly change trends of important characteristics, analyze the annual change trends of the annual energy consumption quota per unit area and the annual per capita energy consumption quota, revise the comprehensive energy consumption quota, and predict the energy consumption quota for future years, specifically:

[0104] like Figure 4 As shown in the figure, a linear regression is performed on the annual data average of the important characteristics of each type of public building, and the average value and change rate of the important characteristics of each type of public building in the next few years are predicted based on the regression curve. Then, the comprehensive energy consumption of different types of public buildings is corrected based on the change rate of the important characteristics and the Shapley value weight. The correction formula is as follows:

[0105] PCEC next =PCEC now *(a 1 *ω 1 +a 2 *ω 2 +a 3 *ω 3 );

[0106] Among them, PCEC nextPCEC is the per capita energy consumption quota of such public buildings in a future year. now represents the current per capita energy consumption quota of such public buildings, a 1 Represents the proportion of the first eigenvector decreasing after several years, ω 1 Represents the proportion of the Shapley value of the first eigenvector to the total Shapley values ​​of the three eigenvectors, and so on.

[0107] Furthermore, based on the monthly data of important characteristics of a single building, the linear regression is used to obtain the change rate in the past few years to judge the rationality of energy use and energy-saving enthusiasm of a single building, and to correct the comprehensive energy consumption quota of a single building, as follows:

[0108] The method for judging the rationality of energy use and energy-saving enthusiasm of a single building is as follows: if there are abnormal data points in the change curve of the important characteristics of a building in the past few years that exceed 20% of the normal fluctuation range, the energy use is unreasonable, and a 5% comprehensive energy consumption quota reduction penalty is imposed; if the change trend of the important characteristics of a building in the past few years is a decreasing trend, and the decline is greater than 20%, it means that the building actively responds to the energy conservation and emission reduction policy, and a 5% increase in the comprehensive energy consumption quota is imposed; on the contrary, if the change trend of the characteristic value in the past few years is an increasing trend, and the increase is greater than 20%, a 5% reduction in the comprehensive energy consumption quota is imposed;

[0109] Embodiment 2

[0110] like Figure 5 As shown, in another embodiment of the present application, a system for public building energy consumption quota based on multi-level classification and dynamic data correction is provided, the system comprising a data acquisition and processing module, a data preprocessing module, a building type primary classification module, a building type secondary classification module, an energy consumption quota module, a feature importance analysis module and an energy consumption quota value correction module;

[0111] The data acquisition and processing module is used to obtain monthly and annual energy consumption parameters of different types of public buildings, including date parameters, load parameters and energy-saving measure parameters; and construct a multi-feature data set based on the obtained energy consumption parameters;

[0112] The data preprocessing module is used to perform data verification and cleaning, data feature screening and data normalization on the multi-feature data set;

[0113] The building type primary classification module is used to classify and label the sample set according to the functionality and energy properties of public buildings;

[0114] The building type secondary classification module is used to perform secondary classification on the results of the primary logic and classification labels according to the K-means model, and to logically interpret and label the results of the secondary classification according to the feature information;

[0115] The energy consumption quota module is used to perform energy consumption quotas on public buildings of the same secondary type according to the quota level and ranking coupling method and the simulation quota method, and to perform weighted average of the quota values ​​obtained by the two methods to obtain a comprehensive energy consumption quota;

[0116] The feature importance analysis module is used to perform annual energy consumption feature importance analysis of different types of public buildings based on the SHAP value model according to energy consumption parameters, energy consumption prediction model and energy consumption prediction results, and obtain important features and Shapley value weights of important features;

[0117] The energy consumption quota value correction module is used to perform linear regression based on the annual data average value of the important characteristics of each type of public building, and predict the average value and change rate of the important characteristics of each type of public building in the next few years based on the regression curve;

[0118] Based on the change rate of important characteristics and the weighted correction of the Shapley value of important characteristics, the comprehensive energy consumption quota of various types of public buildings in the next few years is obtained;

[0119] Based on the monthly data of important characteristics of a single building, the rate of change in the past few years is obtained by linear regression to judge the rationality of energy use and the enthusiasm for energy conservation of a single building, and to correct the comprehensive energy consumption quota of a single building.

[0120] It should be noted here that the system provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure is divided into different functional modules to complete all or part of the functions described above. The system is a method of energy consumption quota for public buildings based on multi-level classification and dynamic data correction applied to the above embodiment.

[0121] Embodiment 3

[0122] like Figure 6 As shown, in another embodiment of the present application, a computer-readable storage medium and an electronic device are further provided, wherein the computer-readable storage medium stores a program, and when the program is executed by the processor of the electronic device, a method for correcting the energy consumption quota of a public building based on multi-level classification and dynamic data of the above embodiment is implemented, specifically:

[0123] Obtain monthly energy consumption parameters of different types of public buildings over several years, including date parameters, load parameters, and energy-saving measures parameters;

[0124] Construct a multi-feature data set based on the acquired energy consumption parameters and perform data verification and cleaning, data feature screening, and data normalization;

[0125] Perform primary logical classification and labeling of multi-feature datasets based on the functionality and energy attributes of public buildings;

[0126] Based on the K-means model, the results of the first-level logic and classification labels are classified into the second level, and the results of the second-level classification are logically interpreted and labeled according to the feature information;

[0127] Based on the quota level and ranking coupling method and the simulation quota method, energy consumption quotas are respectively carried out for public buildings of the same secondary type, and the quota values ​​obtained by the two methods are weighted averaged to obtain the comprehensive energy consumption quota;

[0128] Based on the SHAP value model, the importance of annual energy consumption characteristics of different types of public buildings is analyzed, and the important characteristics and the Shapley value weights of the important characteristics are obtained;

[0129] Perform linear regression based on the annual data average of the important characteristics of each type of public building, and predict the average value and change rate of the important characteristics of each type of public building in the next few years based on the regression curve;

[0130] Based on the change rate of important characteristics and the weighted correction of the Shapley value of important characteristics, the comprehensive energy consumption quota of various types of buildings in the next few years is obtained;

[0131] Based on the monthly data of important characteristics of a single building, the rate of change in the past few years is obtained by linear regression to judge the rationality of energy use and the enthusiasm for energy conservation of a single building, and to correct the comprehensive energy consumption quota of a single building.

[0132] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0133] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A method for public building energy consumption quota based on multi-level classification and dynamic data correction, characterized in that: The steps include: Obtain monthly energy consumption parameters of different types of public buildings over several years, including date parameters, load parameters, and energy-saving measures parameters; Construct a multi-feature data set based on the acquired energy consumption parameters and perform data verification and cleaning, data feature screening, and data normalization; Perform primary logical classification and labeling of multi-feature datasets based on the functionality and energy attributes of public buildings; Based on the K-means model, the results of the first-level logic and classification labels are classified into the second level, and the results of the second-level classification are logically interpreted and labeled according to the feature information; Based on the quota level and ranking coupling method and the simulation quota method, energy consumption quotas are respectively carried out for public buildings of the same secondary type, and the quota values ​​obtained by the two methods are weighted averaged to obtain the comprehensive energy consumption quota; Based on the SHAP value model, the importance of annual energy consumption characteristics of different types of public buildings is analyzed, and the important characteristics and the Shapley value weights of the important characteristics are obtained; Perform linear regression based on the annual data average of the important characteristics of each type of public building, and predict the average value and change rate of the important characteristics of each type of public building in the next few years based on the regression curve; Based on the change rate of important characteristics and the weighted correction of the Shapley value of important characteristics, the comprehensive energy consumption quota of various types of buildings in the next few years is obtained; Based on the monthly data of important characteristics of a single building, the rate of change in the past few years is obtained by linear regression to judge the rationality of energy use and the enthusiasm for energy conservation of a single building, and to correct the comprehensive energy consumption quota of a single building.

2. The method for public building energy consumption quota based on multi-level classification and dynamic data correction according to claim 1 is characterized in that: The different types of public buildings include educational buildings, administrative buildings and medical buildings; the date parameters include date, year, month and quarter; the load parameters include per capita water consumption, building area, number of energy users, number of staff, number of gasoline vehicles, number of diesel vehicles, number of new energy vehicles, electricity consumption, coal consumption, natural gas consumption, oil consumption, other oil consumption, diesel consumption, liquefied petroleum consumption, other energy consumption, solar collector area, solar system installed capacity and geothermal system installed capacity; the energy-saving measures parameters include whether energy-saving renovations have been carried out, the time from the energy-saving renovation to the present, and the energy-saving renovation projects that have been carried out.

3. The method for public building energy consumption quota based on multi-level classification and dynamic data correction according to claim 1 is characterized in that: The data verification and cleaning are specifically as follows: the quartile method is used to perform data verification and cleaning on the acquired energy consumption parameters, specifically: Q3+1.5IQR of the distribution interval is taken as the upper threshold, and Q1-1.5IQR is taken as the lower threshold; wherein Q3 is the third quartile of the distribution interval, that is, the number at the 75% position after the data sequence is arranged from small to large; Q1 is the first quartile of the distribution interval, that is, the number at the 25% position after the data sequence is arranged from small to large; IQR is the interquartile range, IQR=Q3-Q1, and values ​​other than Q1 to Q3 are discarded as abnormal values; The data feature screening is specifically as follows: using the random forest method to analyze the correlation between the per capita energy consumption and unit area energy consumption of public buildings and each influencing factor in the influencing factor sequence S, screening out the influencing factors with a correlation higher than a preset value, and deleting the influencing factors with a correlation lower than the preset value; wherein the influencing factor sequence S = [year year, month month, quarter quarter, per capita water consumption PWC, building area BA, number of energy users NEU, number of staff H, number of gasoline vehicles NGV, number of diesel vehicles NDV, oil consumption GC, other oil consumption OC, diesel consumption DC, liquefied petroleum consumption CLP, other energy consumption OEC, solar collector area SCA, solar system installed capacity SSIC, geothermal system installed capacity GIC, energy-saving transformation ESR]; The normalization process of the data is specifically as follows: Among them, X is the data to be processed, X max is the maximum value in the data set, X min is the minimum value in the data set, X n is the normalized data.

4. The method for public building energy consumption quota based on multi-level classification and dynamic data correction according to claim 1 is characterized in that: The K-means model is used to perform secondary classification on the results of the primary logic and classification labels, and the secondary classification results are logically interpreted and labeled according to the feature information, specifically: Based on the K-means model, the multi-feature data sets of different types of public buildings are clustered, and the K value of different types of public building samples is determined according to the "elbow method"; the "elbow method" is specifically: calculating the sum of squares of errors within the cluster, and using the image inflection point to find the optimal number of classifications, and then performing cluster analysis based on the K-means model and input data. The K value is the number of subcategories after the secondary classification of different types of public buildings; the K-means model inputs energy consumption data directly related to annual energy consumption, including annual energy consumption, annual per capita energy consumption, annual unit area energy consumption, annual electricity consumption, annual oil consumption, and annual natural gas consumption; The results of the secondary classification of different types of public buildings are logically explained, specifically: the specific locations of all types of public buildings are marked on the map and analyzed, and the unknown marked clusters of public buildings of the same secondary type are described to obtain the labels of the secondary type of public buildings.

5. The method for public building energy consumption quota based on multi-level classification and dynamic data correction according to claim 1 is characterized in that: The calculation method of the comprehensive energy consumption quota is as follows: Among them, EUI C is the comprehensive energy consumption quota value, EUI o The EUI is the original comprehensive energy consumption quota calculated based on the original data using the quota level and ranking coupling method. p It is the simulated comprehensive energy consumption quota calculated by the simulated quota method; The specific method of coupling quota level and ranking is as follows: Determine whether a certain type of sample conforms to the normal distribution. If it conforms, the quota level method is used to calculate the comprehensive energy consumption quota of the original data, and 20%, 50% and 80% of the quota level are selected as the constraint value, benchmark value and guide value of the energy consumption quota; if it does not conform, sort the data in the sample, and select 20%, 50% and 80% of the sorting ratio as the constraint value, benchmark value and guide value of the energy consumption quota; The simulation quota method is as follows: Divide the multi-feature dataset into training and test sets in proportion; In the training phase, training is performed based on the prediction models, and the hyperparameters of each model are adjusted through optimization algorithms. The prediction models include stepwise linear regression, support vector machine, random forest, gradient boosting method and BP neural network, and the optimization algorithms include nonlinear weight-decreasing particle swarm optimization algorithm, whale algorithm and sparrow algorithm. In the testing phase, the annual per capita energy consumption and annual unit area energy consumption of each type of public building are predicted respectively. When the prediction results pass the preset indicator evaluation, the energy consumption prediction model training is considered complete; Through indicator comparison, the prediction model and optimization algorithm suitable for the current type of public buildings are selected, and the simulated comprehensive energy consumption quota of various types of public buildings is predicted.

6. The method for public building energy consumption quota based on multi-level classification and dynamic data correction according to claim 1 is characterized in that: The SHAP value model is used to perform a eigenvector correlation analysis on the annual energy consumption of various types of public buildings, specifically: Based on the SHAP value model, the characteristic contribution of energy consumption parameters of different types of public buildings is calculated to obtain the contribution order results of all characteristics, specifically: calculate the marginal contribution of a feature when it is added to the model, and consider the different marginal contributions of the feature in all feature sequences, take the average, and obtain the Shapley value weight of the feature; Important features are screened out based on the interaction between features and the contribution order of features that are sensitive to the results.

7. The method for public building energy consumption quota based on multi-level classification and dynamic data correction according to claim 1 is characterized in that: The weighted correction based on the change rate of important characteristics and the Shapley value weight of important characteristics is used to obtain the comprehensive energy consumption quota of various types of buildings in the next few years, specifically: <h2 style=";text-align:left;direction:ltr">PCEC<h2 style=";text-align:left;direction:ltr"> next <h2 style=";text-align:left;direction:ltr"> =PCEC<h2 style=";text-align:left;direction:ltr"> now <h2 style=";text-align:left;direction:ltr"> *(a1*ω1+a2*ω2+a3*ω3); Among them, PCEC next PCEC is the per capita energy consumption quota of such buildings in a future year. now represents the current per capita energy consumption quota for such buildings, a1 represents the proportion of the first eigenvector's decrease after several years, ω1 represents the proportion of the first eigenvector's Shapley value in the total Shapley value of the three eigenvectors, and so on.

8. The method for public building energy consumption quota based on multi-level classification and dynamic data correction according to claim 1 is characterized in that: The monthly data based on the important characteristics of a single building are used to perform linear regression to obtain the change rate in the past few years, to judge the rationality of energy use and the enthusiasm for energy saving of a single building, and to correct the comprehensive energy consumption quota of a single building, specifically: If there are abnormal data points in the change curve of the important characteristics of a building in the past few years that exceed 20% of the normal fluctuation range, or the characteristic value change trend is increasing and the increase is greater than 20%, it is judged as unreasonable energy use and the comprehensive energy consumption quota is reduced by 5%; if the change trend of the important characteristics of a building in the past few years is decreasing and the decrease is greater than 20%, it is judged as an active response to the energy conservation and emission reduction policy and the comprehensive energy consumption quota is increased by 5%.

9. A system for public building energy consumption quota based on multi-level classification and dynamic data correction, characterized in that: A method for public building energy consumption quota based on multi-level classification and dynamic data correction applied to any one of claims 1-8, comprising a data acquisition and processing module, a data preprocessing module, a building type primary classification module, a building type secondary classification module, an energy consumption quota module, a feature importance analysis module and an energy consumption quota value correction module; The data acquisition and processing module is used to obtain monthly and annual energy consumption parameters of different types of buildings, including date parameters, load parameters and energy-saving measures parameters; and construct a multi-feature data set based on the obtained energy consumption parameters; The data preprocessing module is used to perform data verification and cleaning, data feature screening and data normalization on the multi-feature data set; The building type primary classification module is used to classify and label the sample set according to the functionality and energy properties of public buildings; The building type secondary classification module is used to perform secondary classification on the results of the primary logic and classification labels according to the K-means model, and to logically interpret and label the results of the secondary classification according to the feature information; The energy consumption quota module is used to perform energy consumption quotas on public buildings of the same secondary type according to the quota level and ranking coupling method and the simulation quota method, and to perform weighted average of the quota values ​​obtained by the two methods to obtain a comprehensive energy consumption quota; The feature importance analysis module is used to perform annual energy consumption feature importance analysis of different types of public buildings based on the SHAP value model according to energy consumption parameters, energy consumption prediction model and energy consumption prediction results, and obtain important features and Shapley value weights of important features; The energy consumption quota value correction module is used to perform linear regression based on the annual data average value of the important characteristics of each type of public building, and predict the average value and change rate of the important characteristics of each type of public building in the next few years based on the regression curve; Based on the change rate of important characteristics and the weighted correction of the Shapley value of important characteristics, the comprehensive energy consumption quota of various types of buildings in the next few years is obtained; Based on the monthly data of important characteristics of a single building, the rate of change in the past few years is obtained by linear regression to judge the rationality of energy use and the enthusiasm for energy conservation of a single building, and to correct the comprehensive energy consumption quota of a single building.

10. A storage medium storing a program, characterized in that: When the program is executed by a processor, the method for public building energy consumption quota based on multi-level classification and dynamic data correction as described in any one of claims 1 to 8 is implemented.

Images