CN112884013A - Energy consumption partitioning method based on data mining technology - Google Patents

Abstract

The invention discloses an energy consumption partitioning method based on a data mining technology, belongs to the technical field of building energy conservation, can make up for the blank of energy consumption monitoring system data utilization, and classifies building energy consumption data. Through the steps of data classification, data identification, energy consumption region division and the like, the characteristic value of the energy consumption mode is extracted, and different types of energy consumption data are identified and utilized in a layering mode. For buildings which are not provided with energy consumption monitoring systems, such as buildings which can only obtain the total electricity consumption, the energy consumption composition information of other buildings can be obtained by utilizing the method, and then the classification method is applied to the buildings which are not provided with the energy consumption item monitoring systems according to the characteristics of the same type of buildings, such as similar function subareas, energy consumption characteristics and the like, so that the method is beneficial to knowing the energy consumption composition of the buildings which are not provided with the energy consumption monitoring systems, and the diagnosis efficiency is improved to the accuracy.

Description

Energy consumption partitioning method based on data mining technology

Technical Field

The invention belongs to the technical field of building energy conservation, is mainly used for building energy consumption classification research, and relates to a building energy consumption classification method based on a data mining technology.

Background

Currently, the environment deteriorates sharply and energy is excessively consumed, wherein a building is a large household for energy consumption. According to statistics, global building energy consumption accounts for more than 40% of total energy consumption. This consumption has been exacerbated by the explosive growth and the increasing speed of urbanization of the real estate market in recent years. Under the background, the energy consumption composition of the existing building is effectively known, and effective energy-saving diagnosis is carried out, so that the key for reducing the energy consumption of the building is realized.

At present, energy consumption supervision platforms are established in many public buildings, and a large amount of data are accumulated in the operation process. The method is not only beneficial to the development of energy-saving management work of public buildings, but also can provide data support for scientific research of building energy conservation and government made policies. The utilization of data for the relevant platform can be mainly divided into three directions. The first direction is used for monitoring the operation condition of main energy utilization equipment and revealing the maintenance level problem and the deficiency of extensive management of the building energy utilization equipment in the operation process. The second application direction is to combine with a mathematical algorithm when a monitoring platform is built, so as to realize the automatic reporting and early warning functions of the building. The monitoring platform is combined with the BIM technology, so that the functions of power utilization, water quantity monitoring, equipment information real-time query, fault alarm and maintenance path calculation, energy-saving analysis and the like are realized. The third application direction is to establish an energy consumption prediction model based on data in the monitoring platform, and the energy consumption prediction model is used for researching the overall energy consumption level of the building. Energy consumption itemized monitoring is carried out on 12 teaching buildings, an energy consumption prediction model of the teaching buildings of schools is established, and the main influence factors of energy consumption are found to be the teaching floor area and the number of staff.

Although data support is provided for building energy consumption analysis and energy-saving transformation, a large amount of data bring data disasters, due to the reasons of technology, management and the like, a large amount of data with problems can be generated in the operation process of the energy consumption monitoring platform, managers are difficult to find and process the problem data, and finally the energy consumption monitoring data is far away from the real energy consumption of a building and gradually backs up with the original intention of designing the energy consumption monitoring platform.

Disclosure of Invention

In view of the defects and shortcomings in the prior art, the invention provides an energy consumption distinguishing method based on a data mining technology, which improves the data quality through steps of data classification, data identification, energy consumption area division and the like, improves the data utilization value, and classifies the building energy consumption data to solve the problem that real energy consumption data cannot be obtained due to a large amount of problem data.

In order to solve the technical problems, the invention adopts the following technical scheme: an energy consumption partitioning method based on a data mining technology comprises the following steps:

1) energy consumption data including the total energy consumption of the building, and the individual data of each branch of illumination, air conditioning, power and the like are acquired by a building energy consumption monitoring system;

2) clustering is a process of dividing a data set into a plurality of groups or classes, and enabling data objects in the same group to have higher similarity, while data objects in different groups are dissimilar, and the similar or dissimilar measurement is determined based on the value of the description attribute of the data objects, and is generally described by using the distance between the objectsThe method comprises the following steps of dividing power consumption items such as illumination, air conditioning, power and the like, continuously subdividing the power consumption items downwards according to different functions born in the items, and selecting K initial clustering centers C of energy consumption data by using a K-means initial clustering center selection strategy_k＝{C₁,…,C_k}；

3) If some kind of power consumption data sample x is not known_iDefining ArrayList set aList, calculating x_iEuclidean distance d from all cluster centers_ikD is mixing_ikAscending sort

X is to be_iAdd sequence number i to the aList, according to the data sample x_iR is calculated by a ratio formula of the second minimum distance from the clustering center to the minimum distance_iJudging the category of the data sample;

4) if r is_i>ε, then directly according to x_iIs divided into x by the minimum distance principle_i；

5) If r is_i<E, then x is calculated according to Euclidean distance and the residual data sample_iIn dataset X- { X_pNearest neighbor x in | p ∈ aList }_jAdd sequence number j to the aList if x_jHas been classified into C_kIn the cluster, x is also added_iDivision into C_kIn a cluster, if x_jHas not been divided according to process x_iProcess of (1) calculating x_jDistance from cluster center if r_i>E, then x need not be calculated_jNearest neighbor of (2), directly divide x_iAnd x_jTo x_jIn the cluster to be divided, if r_j<ε, then x is calculated for the sample that also excludes the corresponding aList element_jUntil the sample corresponding to the last element in the set aList can determine the cluster to which the sample belongs, x corresponding to the first element in the set aList_iSamples corresponding to other elements are all assigned to this cluster;

6) after the data samples in the data set are divided one by one according to the steps 3-5, calculating the error square sum E 'of the clustering result, and comparing the error square sum E' with the error square sum E obtained by the last calculation;

7) if E' -E<10^-10Then the algorithm converges and outputs the clustering result; and otherwise, E' is calculated, the center of mass of each cluster is used as a clustering center, and the steps 3-6 are executed in a circulating mode until the algorithm is converged.

As a further development of the invention, X ═ X for the data sets to be clustered_i/x_i∈R^p1, 2, …, n, any two data samples x_iAnd x_jThe Euclidean distance of (A) is:

data sample x_iRatio of second minimum distance to minimum distance from cluster center:

data sample set X data samples remaining after removing some data samples:

X-X’＝{x_i/x_i∈R^pi ≠ 1, 2, …, n and i ≠ p, q, r }

X′＝(x_p，x_q，x_r)

Sum of squares of errors of clustering results:

accuracy of clustering algorithm:

wherein n is₁The number of correctly clustered data samples is n, and the number of the total samples is n;

the clustering result obtained after the algorithm convergence is the final classification condition of the power consumption data, and the composition information of the building power consumption, including illumination, air conditioning, power and the proportion of each tail end, can be obtained from the clustering result.

Compared with the prior art, the invention has the beneficial effects that: the invention provides a method for classifying and processing building energy consumption data. Through the steps of data classification, data identification, energy consumption region division and the like, the characteristic value of the energy consumption mode is extracted, and different types of energy consumption data are identified and utilized in a layering mode. For buildings which are not provided with energy consumption monitoring systems, such as buildings which can only obtain the total electricity consumption, the energy consumption composition information of other buildings can be obtained by utilizing the method, and then the classification method is applied to the buildings which are not provided with the energy consumption item monitoring systems according to the characteristics of the same type of buildings, such as similar function subareas, energy consumption characteristics and the like, so that the method is beneficial to knowing the energy consumption composition of the buildings which are not provided with the energy consumption monitoring systems, and the diagnosis efficiency is improved to the accuracy.

Drawings

The invention will be further described with reference to the following drawings and detailed description:

FIG. 1 is a schematic frame diagram of the present invention;

fig. 2 is a schematic diagram of a result of cluster analysis of power consumption ratios in various areas of a mall in the embodiment.

Detailed Description

For better understanding of the technical solutions and advantages of the present invention, the following detailed description of the present invention is provided in conjunction with the accompanying drawings and specific embodiments, it should be understood that the specific embodiments described herein are only for the understanding of the present invention and are not intended to limit the present invention, and all other embodiments obtained by those of ordinary skill in the art without any inventive work are within the scope of the present invention.

Referring to fig. 1, the present invention provides a technical solution: the energy consumption partitioning method based on the data mining technology comprises the following steps:

1) energy consumption data including the total energy consumption of the building, and the individual data of each branch of illumination, air conditioning, power and the like are acquired by a building energy consumption monitoring system;

2) clustering is a process of dividing a data set into groups or classes, and making data objects in the same group have a high degree of similarity, while data objects in different groups are dissimilar, with the measure of similarity or dissimilarity beingThe method can divide the total electricity of the building into partial items of the electricity for illumination, air conditioning, power and the like, then continuously subdivides the partial items downwards according to different functions born in the partial items, and firstly, an initial clustering center C of K energy consumption data is selected by applying a K-means initial clustering center selection strategy_k＝{C₁,…,C_k}；

3) If some kind of power consumption data sample x is not known_iDefining ArrayList set aList, calculating x_iEuclidean distance d from all cluster centers_ikD is mixing_ikAscending sort

X is to be_iAdd sequence number i to the aList, according to the data sample x_iR is calculated by a ratio formula of the second minimum distance from the clustering center to the minimum distance_iJudging the category of the data sample;

4) if r is_i>ε, then directly according to x_iIs divided into x by the minimum distance principle_i；

5) If r is_i<E, then x is calculated according to Euclidean distance and the residual data sample_iIn dataset X- { X_pNearest neighbor x in | p ∈ aList }_jAdd sequence number j to the aList if x_jHas been classified into C_kIn the cluster, x is also added_iDivision into C_kIn a cluster, if x_jHas not been divided according to process x_iProcess of (1) calculating x_jDistance from cluster center if r_i>E, then x need not be calculated_jNearest neighbor of (2), directly divide x_iAnd x_jTo x_jIn the cluster to be divided, if r_j<ε, then x is calculated for the sample that also excludes the corresponding aList element_jUntil the sample corresponding to the last element in the set aList can determine the cluster to which the sample belongs, x corresponding to the first element in the set aList_iSamples corresponding to other elements are all assigned to this cluster;

6) after the data samples in the data set are divided one by one according to the steps 3-5, calculating the error square sum E 'of the clustering result, and comparing the error square sum E' with the error square sum E obtained by the last calculation;

7) if E' -E<10-¹⁰Then the algorithm converges and outputs the clustering result; and otherwise, E' is calculated, the center of mass of each cluster is used as a clustering center, and the steps 3-6 are executed in a circulating mode until the algorithm is converged.

For a dataset to be clustered, X X_i/x_i∈R^p1, 2, …, n, any two data samples x_iAnd x_jThe Euclidean distance of (A) is:

data sample x_iRatio of second minimum distance to minimum distance from cluster center:

data sample set X data samples remaining after removing some data samples:

X-X’＝{x_i/x_i∈R^pi ≠ 1, 2, …, n and i ≠ p, q, r }

X′＝(x_p，x_q，x_r)

Sum of squares of errors of clustering results:

accuracy of clustering algorithm:

wherein n is₁The number of correctly clustered data samples is n, and the number of the total samples is n;

the clustering result obtained after the algorithm convergence is the final classification condition of the power consumption data, and the composition information of the building power consumption, including illumination, air conditioning, power and the proportion of each tail end, can be obtained from the clustering result.

A market building is selected, an energy consumption item monitoring system is installed in the market, and the monitoring level can reach the equipment level. Data in the energy consumption monitoring system are extracted, energy consumption characteristic analysis of the market is carried out by using the energy consumption clustering analysis method, and the power consumption of each area accounts for the example shown in fig. 2.

As can be seen from fig. 2, the lighting system power usage accounts for about 37% of the total power usage, with commercial lighting accounting for 75% of the total lighting and public lighting accounting for 13% of the total lighting power usage. The electricity consumption of the air conditioning system accounts for about 55% of the total electricity consumption, wherein the electricity consumption of the cold source accounts for about 79% of the total electricity consumption, and the electricity consumption of the tail end accounts for 21%. The electricity consumption of the elevator system accounts for 8% of the total electricity consumption, wherein the passenger elevator accounts for 75% of the total electricity consumption of the power system, and the fire elevator accounts for 25%. On the basis, the data volume is expanded, and the power utilization data of a plurality of buildings of the same type are added, so that the power utilization composition of the buildings of the markets can be more clearly known, and the power utilization health degree judgment of the buildings of the same type is guided.

By taking a market building as an example, the invention can also obtain the proportion of public lighting, commercial lighting and emergency lighting in lighting power, the proportion of a cold source side and a tail end side in an air conditioning system, and the proportion of power equipment such as an elevator, a living water pump and the like in power utilization. The system integrates a plurality of market data, can obtain the power utilization proportion guidance value of each terminal branch of the market building, and can be used for judging the power utilization health degree of unknown markets.

The invention provides a method for classifying and processing building energy consumption data, which is based on a clustering algorithm, wherein the clustering algorithm can ensure that data objects in the same group have higher similarity, and the dissimilarity of the data objects in different groups can be ensured. The method is used for carrying out cluster analysis on the building energy consumption data, separating out the proportion of the power consumption in different areas and different functions, extracting the characteristic value of the power consumption mode, and carrying out layered identification and utilization on the energy consumption data of different types. The method is beneficial to providing guidance for analyzing the energy consumption composition of the building without the energy consumption monitoring system, accurately positioning the equipment or branch with the problem of high energy consumption, and improving the energy consumption diagnosis efficiency.

Claims (2)

1. An energy consumption partitioning method based on a data mining technology is characterized in that: the method comprises the following steps:

1) energy consumption data including the total energy consumption of the building, and the individual data of each branch of illumination, air conditioning, power and the like are acquired by a building energy consumption monitoring system;

2) clustering is a process of dividing a data set into a plurality of groups or classes, and enabling data objects in the same group to have higher similarity, while data objects in different groups are dissimilar, similar or dissimilar measurement is determined based on the value of the description attribute of the data objects, and is generally described by using the distance between the objects_k＝{C₁,…,C_k}；

3) If some kind of power consumption data sample x is not known_iDefining ArrayList set aList, calculating x_iEuclidean distance d from all cluster centers_ikD is mixing_ikAscending sort

X is to be_iAdd sequence number i to the aList, according to the data sample x_iR is calculated by a ratio formula of the second minimum distance from the clustering center to the minimum distance_iJudging the category of the data sample;

4) if r is_i>ε, then directly according to x_iIs divided into x by the minimum distance principle_i；

5) If r is_i<E, then x is calculated according to Euclidean distance and the residual data sample_iIn dataset X- { X_pNearest neighbor x in | p ∈ aList }_jAdd sequence number j to the aList if x_jHas been classified into C_kIn the cluster, x is also added_iDivision into C_kIn a cluster, if x_jHas not been divided according to process x_iProcess of (1) calculating x_jDistance from cluster center if r_i>E, then x need not be calculated_jNearest neighbor of (2), directly divide x_iAnd x_jTo x_jIn the cluster to be divided, if r_j<ε, then x is calculated for the sample that also excludes the corresponding aList element_jUntil the sample corresponding to the last element in the set aList can determine the cluster to which the sample belongs, x corresponding to the first element in the set aList_iSamples corresponding to other elements are all assigned to this cluster;

6) after the data samples in the data set are divided one by one according to the steps 3-5, calculating the error square sum E 'of the clustering result, and comparing the error square sum E' with the error square sum E obtained by the last calculation;

7) if E' -E<10^-10Then the algorithm converges and outputs the clustering result; and otherwise, E' is calculated, the center of mass of each cluster is used as a clustering center, and the steps 3-6 are executed in a circulating mode until the algorithm is converged.

2. The method for partitioning energy consumption based on the data mining technology according to claim 1, wherein: for a dataset to be clustered, X ═ X_i/x_i∈R^p1, 2, …, n, any two data samples x_iAnd x_jThe Euclidean distance of (A) is:

data sample x_iRatio of second minimum distance to minimum distance from cluster center:

data sample set X data samples remaining after removing some data samples:

X-X’＝{x_i/x_i∈R^pi ≠ 1, 2, …, n and i ≠ p, q, r }

X′＝(x_p，x_q，x_r)

Sum of squares of errors of clustering results:

accuracy of clustering algorithm:

wherein n is₁The number of correctly clustered data samples is n, and the number of the total samples is n;

the clustering result obtained after the algorithm convergence is the final classification condition of the power consumption data, and the composition information of the building power consumption, including illumination, air conditioning, power and the proportion of each tail end, can be obtained from the clustering result.

Images