• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    An improved data processing method for analyzing the energy consumption of a site from measurement data comprising: - selection of time series of data, - segmentation of time series into sections, - projection of sections, - display of data projections of data sections, - selection of at least a first group of sections, - establishment of a numerical classification model characterizing the first group of data sections selected from calendar metadata - establishment of a numerical classification model characterizing the first group of data sections selected from the section profiles
    公开号:FR3032786A1
    申请号:FR1551302
    申请日:2015-02-17
    公开日:2016-08-19
    发明作者:Sylvain Marie；Frederic Suard
    申请人:Schneider Electric Industries SAS；
    IPC主号:
    专利说明:

    [0001] TECHNICAL FIELD AND PRIOR ART The present invention relates to the field of computer processing of measurement data and the modeling of these data to enable the processing of data. analysis of the energy consumption of sites such as residential buildings, commercial premises, factories, data centers. To carry out an energy consumption audit of a site, in addition to the measurement readings of electricity meters, platforms have been developed equipped with sensors that measure many parameters at different positions of the site, including temperature, humidity, temperature, humidity and humidity. differential pressure, with high sampling rates. On the other hand the platforms used in planning, control and supervision systems of industrial processes are also connected, providing many parameters among which the hours worked, the quantities produced. This set is likely to produce huge amounts of data to analyze for example for energy experts. There is a need for a new computer tool to rapidly understand the mode of operation of a site in terms of energy consumption, to analyze its behavior over a period of time and in particular to check whether this site has followed a method of expected operation, and to model the functioning of this site for a future follow-up of its performances. SUMMARY OF THE INVENTION An embodiment of the present invention provides a data processing method for analyzing the energy consumption of at least one site from measurement data, including measurement data of 3032786. electrical consumption as well as measurement data of one or more physical parameters, derived from at least one measuring device disposed on this site, the method comprising the steps of, via a computer processing system: a) selecting, in a selected time span of analysis, time series of data from the measuring device, the time series of data being associated with explanatory metadata of said time series of data, b) segmenting the time series of data by a plurality of data sections of the same duration, c) performing a projection of the selected data sections and situate within this time span of analysis in a space of at least two dimensions according to their degree of similarity, d) displaying in a software graphical interface via a display device connected to the computer processing system, a graphical representation of said space including projections of the data sections, some projections being grouped as one or more groups of data sections. The measuring device is provided with one or more measuring elements including at least one power consumption meter and / or one or more sensors for measuring physical quantities. The creation of the groups of sections can be established automatically by means of an unsupervised classification algorithm, for example by partitioning, with methods of the type of the mobile centers (k-means), partitioning around the medoids (PAM, k- medoids), CLARA (acronym for "Clustering LARge Application"), dynamic cloud method, fuzzy classification method (FANNY), or DBSCAN (acronym for "Density-Based Spatial Clustering of Applications with Noise"); or for example by hierarchical classification, with hierarchical ascending classification methods (AGNES, acronym for "agglomerative nesting"), or hierarchical descending classification (DIANA, acronym for "divisive analyzing").
    [0002] Such a graphical representation facilitates data selection and outlier data discrimination. Thereafter, the processing may comprise steps of: e) selecting at least a first group of data sections, f) establishing one or more digital models characterizing the first group of data sections selected. Such models can allow in a later stage after acquisition of new data, to monitor the performance of the site and detect any consumption anomalies on a site.
    [0003] Among the models established in step f) may be an individual numerical model characterizing the membership of the data sections to a first group, constructed from an analysis of temporal measurement profiles contained in the data sections of this group. first group. This model can be linear, piecewise linear or nonlinear, and for example based on detection thresholds, a class support vector machine or core component analysis. Among the models established in step f) may also include a calendar individual numerical model giving a condition of belonging of new data to a group of data from temporal metadata associated with said new data.
    [0004] After step f), it is possible to carry out the following steps: - to acquire new data, - to use the characteristic individual numerical model to check the membership of the new data in a group of data. Among the models established in step f) may be a regression model capable of providing an estimate of the measurement data of at least one selected element of the measurement device from data from one or more other elements of the measurement system. measure (s) of the measuring device. For example, this model can be linear, piecewise linear or nonlinear, based on a least squares method or a carrier vector machine or a relevance vector machine.
    [0005] According to one possible embodiment, the method may comprise the following steps: - establishing for the first group a standard profile of magnitude (s) measured by the measuring device for a duration corresponding to that of the 5 sections, then - display this typical profile through the GUI. The method may include the steps of: - selecting, through the software graphical interface, a particular group of data sections, - associating that particular group with an identifier. The identifier may be generated using a naming algorithm implementing a decision tree and using metadata associated with the data stretches of the particular group. The selection of time series of data in step a) may include a selection of at least one type of metadata, in particular: a site area descriptor for selecting time series of measurement data from one or more particular areas of the site among a set of areas of the site; and / or a type of measurement device descriptor for selecting time series data from one or more types of particular measuring means. among a set of measurement means of the measuring device, and / or a site measurement parameter type descriptor for selecting time series of data relating to one or more types of physical parameters from a set of physical parameters that the measuring device is capable of measuring. According to an implementation possibility, the method may furthermore comprise: a selection, through the software graphical interface and from among said projections of sections of one or more particular projections which does not belong to any group.
    [0006] The graphical representation obtained in step d) can be located in a first window of the software graphical interface and in which the sections are also represented in another form in a second window of the graphical interface, the selection the first group in said first window resulting in a selection and highlighting in said second window sections of the first group. The selection of the time series of data carried out in step a) or the selection of the first group carried out in step e) can be performed by means of this other graphical representation.
    [0007] The degree of similarity in step c) can be established by selecting a metric. This metric can be selected through the software GUI from a list of several different metrics. A selection of a projection algorithm may accompany the selection of this metric. This selection can also be made through the software graphical interface from a list of several different projection algorithms. A selection of an unsupervised classification algorithm can accompany the selection of this metric and this projection algorithm.
    [0008] This selection can also be made through the software graphical interface from a list of several different unsupervised classification algorithms. The data processing method may further comprise the steps of: - receiving new data, - detecting using a calendar template, from the temporal metadata associated with these new data, if these new data belong to to a particular group among said groups, - checking the membership of these new data to the particular group 30 using a characteristic individual model associated with the particular group, - using a regression model to obtain a value nominal energy consumption, - compare nominal consumption value with new data at.
    [0009] An embodiment of the present invention provides a computer program comprising program code instructions to enable the computer processing system to perform one or more of the steps of the method as defined above. An embodiment of the present invention also provides a digital data carrier usable by a computer processing system, including code instructions of a computer program as referred to above. BRIEF DESCRIPTION OF THE DRAWINGS The present invention will be better understood on reading the description of exemplary embodiments given purely by way of indication and in no way limiting, with reference to the appended drawings in which: FIG. 1 schematically represents a computer system comprising a an expert platform equipped with a data analysis and modeling tool which is the subject of an embodiment of the present invention and which makes it possible to analyze the energy consumption of sites from measurement data made on these sites ; FIG. 2 represents an exemplary flow chart showing a first sequence of processing steps that can be implemented by the data analysis and modeling tool; FIG. 3 represents an exemplary flow chart showing a second sequence of processing steps that can be implemented by the data analysis and modeling tool; FIGS. 4A-4H illustrate a dashboard of an exemplary graphical interface of the data analysis and modeling tool; FIGS. 5A-5B illustrate groups of projected data and their association with an identifier using a decision tree by the data analysis and modeling tool; FIG. 6 represents a typical profile of electrical energy consumed as a function of the temperature that can be obtained and displayed using the data analysis and modeling tool; FIG. 7 represents an example of processing carried out by a configured performance monitoring system using one or more reference models provided by the data analysis and modeling tool according to the invention; Identical, similar or equivalent parts of the various figures described below bear the same numerical references so as to facilitate the passage from one figure to another. The different parts shown in the figures are not necessarily in a uniform scale, to make the figures more readable.
    [0010] The different possibilities (variants and embodiments) must be understood as not being exclusive of one another and can be combined with one another. DETAILED DESCRIPTION OF PARTICULAR EMBODIMENTS An example of a data processing system for analyzing the energetic behavior of a site over a given time range is illustrated in FIG. 1. The studied site may be for example a building or a building. factory or a commercial office or data center for which an energy audit is desired. This site is equipped with at least one measuring device 10 producing data to be processed. The measuring device 10 may be provided with one or more measuring elements, in particular at least one power consumption counter 11, as well as one or more measurement sensors 13, in particular sensors for measuring quantities. such as, for example, temperature, humidity, differential pressure, air velocity. The measuring device 10 is capable of providing a large amount of data to be analyzed. Some sensors can indeed provide data at a frequency of the order of at least one measurement per minute.
    [0011] Among the data 21 to be processed from the measuring device 10 are time series of data 21. These time series can be sequences of time / value pairs of measurements. The time series are accompanied by metadata 22, that is to say explanatory data of the time series. Among these metadata 22 it is possible to distinguish metadata of a first type called "series" and which group first descriptors of the time series. An example of a series metadata: [Building = Site_38TEC, Zone = Office, Type = Energy, Sensor = PM200] indicates a particular zone of the building being studied, as well as the type of measurement data studied and the type of sensor from which it is derived. this measurement data. Metadata of a second type called "temporal" also accompany the time series of data. Temporal metadata are descriptors of a time. For example, a date "18/02/14 00:00" can be described with the following time metadata: [year = 2014, month = February, day = 18, day of the week = Tuesday, quarter = 1, semester = 1].
    [0012] The analysis of the measurement data is carried out by means of an application 31, which is the subject of the present invention, more particularly an analysis and modeling software which can be installed on a computer platform called the "expert platform". ". The expert platform can be centralized or alternatively distributed on a plurality of computer equipment.
    [0013] The application 31 may be optionally accessible by a computer system of the "cloud computing" type. The analysis and modeling application 31 includes in particular a visual analysis tool allowing a visual and interactive exploration as well as a graphical representation of the imported measurement data.
    [0014] The analysis and modeling application 31 makes it possible to search for groups of data having similar temporal profiles, even for a large volume of data. The analysis and modeling application 31 can be controlled by a user and / or at least one computer script. In both cases, the data processing performed by the application 31 allows the establishment of one or more numerical models of energy behavior of the studied site in a given time range. The models produced by the analysis application 31 can be integrated into computer files that can be used by office applications. These files can be for example in a first form called "report" or a second form called "project" and can facilitate understanding, profiling the energy behavior of the site and performing comparative analysis.
    [0015] An example of a digital model gives the electric consumption as a function of the average temperature in the site for a given period, for example a daily period. Graphical representations of such models can be visualized by means of a graphical interface of the application 31. By way of example, FIG. 6 gives a graphical representation C1 obtained by means of the application 31 of the model of FIG. the average energy consumed by a site as a function of temperature. The application 31 makes it possible to better discriminate the measurement points Pi by creating and naming groups, and to isolate aberrant measurement points more easily: here only the group "non-aberrant working days" has been kept which enabled the construction of a powerful digital model. Thus, in addition to the implementation of numerical models, the application 31 allows a user to perform a visual analysis of the measurement data via the graphical interface. The application 31 also makes it possible to extract and / or identify groups of data relating to similar situations and then possibly to associate these groups to an identifier generated using a naming algorithm, and to produce a numerical model of group behavior. For example, in the case where the study site is a commercial space, a group of measurement data made on non-working days and holidays can be created and studied in order to compare it with another group of measured data. working days. The application 31 may comprise, in addition to an analysis tool, a querying tool for searching, extracting, sorting and formatting data. Application 31 may also include user interface libraries, graphics libraries, management and data transformation libraries in memory, data generation libraries from templates, a storage and management tool a storage and model management tool, a storage and project management tool. Models 40 and data produced by the expert platform 30 can then be exploited by a site performance monitoring computer system. Such a system 50 is also capable of receiving data from the measurement device 10. The performance monitoring system 50 can be configured to make a comparison between a model or set of reference models provided by the application 31 of the expert platform 30 and the real behavior of the site given in real time and continuously through data from the measuring device 10. This system 50 performance monitoring can be centralized or alternatively distributed on several equipment 52 on which applications developed by the applicant such as "Energy Operation First" or "Facility Insight", "StruxureWare Plant Operation Ampla", "StruxureWare Resource 25 Advisor", "Remote Energy Management" can also be deployed and allow to as models produced by the platform 30 to make a prediction, for example that of a neighbor Peak consumption and its cost, or to detect a malfunction, for example abnormal electrical consumption, or to make a comparison with other sites to define levers 30 of energy performance improvements of the site studied.
    [0016] FIGS. 2 and 3 illustrate data processing phases comprising steps that are also called "actions" that can be performed by the analysis and modeling application 31. In these figures, each action (represented by a solid line frame) of configuration or algorithm execution produces a result (represented by an arrow from the dashed frame), which can be used to create or one or more interactive graphic views displayed (the display action being represented by a dashed frame) on a screen, in particular a computer equipment of the platform 30. The implementation of certain actions, the creation or the update of the 10 interactive graphical views may in certain cases be conditioned by the prior receipt of results (represented by arrows arranged above the frames) processing or input by the user. In the case where a human user controls the application 31, the interactive graphic views help him in his analysis of the data from the site and 15 allow him to choose a next action or to establish a diagnosis on the energy consumption of the site . Prior to a data processing phase allowing the establishment of models, a so-called "upstream" phase illustrated on the sequential flow chart of FIG. 2 is implemented by the application 31.
    [0017] During this upstream phase, selection, preprocessing and configuration actions under the control of the user U or a computer script are performed. An import (step E000 in FIG. 2) of the time series and their metadata to the expert platform 30 can be performed before this upstream phase.
    [0018] The imported data and metadata can come for example from: - one or more remote management computer equipment, such as those of the Supervisory Control and Data Acquisition (SCADA) type and / or continuous control form connected to the measuring device 10 of the studied site and / or - a dedicated system of data collection and metadata.
    [0019] The import step is optional when the data and metadata to be processed are in the form of one or more pre-recorded files in the platform 30. Next, the selection of data series to be analyzed is carried out. This selection can be done directly (step E011) or through a selection of particular metadata (step E010) and can then be displayed (steps E020 and E021). This selection is also accompanied by a selection of a time range of analysis (step E012) on which it is desired to concentrate the analysis. This time range can be continuous, for example over a year or a semester, or 10 over a month or a week. Alternatively, this time range may have discontinuities. For example, you can select all the working days of a year or all the working hours of a day. Another example of discontinuous range selection includes site data taken from Monday to Thursday for all weeks in a year.
    [0020] Furthermore, a count of a missing data rate over a selected time range, for example on a day, can be determined by the application 31 and displayed. This can help the user to make his choice of the study time range. After this selection, a pre-treatment step (step E012) is performed. This pre-treatment may include a cleaning sub-step, in particular to remove outliers. For example, measurement data located outside a working range in which the sensors of the measuring device 10 are supposed to operate can be suppressed during such a step. The preprocessing may also include synchronization or re-synchronization of the time series over a fixed sampling interval chosen by the user. This synchronization can be performed by interpolation. The fixed sampling interval may be, for example, 1 minute, or 10 minutes, or 1 hour, or 1 day. This step can be useful especially when the sampling of measurement data is irregular.
    [0021] The pretreatment may also comprise a substep of normalization of the time series during which a change of scale may be particularly effected. The result of the preprocessing step can then be displayed (step E022).
    [0022] The "upstream" phase further comprises an additional step (step E014) during which a temporal metadata is selected, the selected time series then being segmented into sections of the same duration. For example, it is possible to segment the time series of data into sections of one month or one week, or preferably one day. This segmentation step makes it possible to define situations to be compared. The segmentation may also include a selection of a particular serial metadata. For example, a site area descriptor may be selected to select only data relating to one or more particular areas of the site from the set of areas covered by the metering device 10. A comparison of energy consumption of areas of the site may be thus be implemented. For example, it is also possible to select a sensor type descriptor for selecting only data from one or more particular sensors of the measuring device 10. Such a selection makes it possible to compare data from a first sensor. group of sensors with respect to data from a second group of sensors. According to another example, a measurement parameter type descriptor can also be selected to select only data relating to one or more particular types of physical parameters from the set of physical parameters measured by the measuring device 10. After upstream phase, a so-called "analysis" processing phase leading to a projection of the data sections and a "group" processing phase making it possible to form groups of projected data sections having similarities is carried out.
    [0023] In FIG. 3, an exemplary flowchart of steps performed during such an analysis phase is given. In order to be able to perform a similarity measurement between data sections, a distance matrix is first established (step E100 on the sequential flowchart of FIG. 3). For this, prior to this step, a metric selection from a list of pre-recorded metrics can be performed. A metric is a function that defines the distance between elements in a set. The selected metric may for example be of the Euclidean distance type, or for example of the DTW type (for "Dynamic Time Warping" type) or of the CID type (for "distance complexity invariance" type) or of the Manhattan distance type or of the DcorT type type. and as presented, for example, in the document by Ahlmane Douzal-Choukria and Cécile Amblard: "Crees for Lime series classification", Pattern Recognition Journal, vol. 2, No. 45, p. 1076-1091, 2011.
    [0024] A metric configured or imported by the user U himself can also be selected and used. Thus, the application 31 may also include a tool allowing the user U to define himself a metric different from those which are pre-recorded. Once the distance matrix is established by the application 31, a projection algorithm is executed (Step E102) to project the time series sections in a space of at least two dimensions according to their degree of similarity. The projection algorithm may also be selected beforehand by the user U from among a plurality of projection algorithms. The projection algorithm may for example be an ISOMAP or hierarchical clustering type of algorithm or of the Principal Component Analysis type (also referred to as ACP). The result of this projection can then be displayed (step E122) through the software graphical interface of the application 31. The metric selection, the metric import, the projection algorithm selection can be performed. by the user U using for example drop-down menus of a control window of the graphical interface. This control window may optionally be the one from which a data import and a selection of study time range can be made. Next, a classification of the data (step E104) also called "clustering" is performed. This step may include selecting an unsupervised classification algorithm, and driving this algorithm to the formation or demonstration of similar projected data sets, and highlighting or discriminating outlier data. . To carry out this classification, a mobile-type method (k-means) partitioning algorithm, partitioning around medoids (PAM, k-medoids), CLARA (acronym for "Clustering LARge Application"), dynamic cloud method, fuzzy classification method (FANNY), or DBSCAN (acronym for "Density-Based Spatial Clustering of Applications with Noise"); or a hierarchical hierarchical hierarchical classification algorithm (AGNES, acronym for "agglomerative nesting"), or hierarchical descendant classification (DIANA, acronym for "divisive analyzing") may be employed. This classification can be made from the result of the projection step or alternatively (not shown in Figure 3) can be coupled to the projection step.
    [0025] In addition to this "automatic" classification in data group, a "manual" classification implemented by the user can also be performed, for example from a view on which the projected data are displayed. A display (step E124) of the groups and their associated characteristics for example of a typical profile, associated metadata, can be realized using different interactive views. Examples of such interactive views are given in Figures 4A-4H. Each group of data can then be associated with an identifier. A naming of the groups can be done by the user in a "manual" manner, from his understanding of the group provided for example by the different interactive views on which these groups and their associated characteristics are represented. An automatic group naming, or manual naming aid, can also be provided (step E106) using an algorithm using a decision tree executed on the metadata associated with these groups. Alternatively, the association of a group of data to an identifier is performed semi-automatically. In this case, the application 31 proposes an identifier for a group of data established according to the detected similarities. In FIG. 5B, an example of a decision tree that has been used to name three groups 101, 102, 103 of identified data is given. The corresponding groups are shown in an illustrated view in FIG. 5A. In this example, the time series have been segmented into sections of duration equal to one day. The naming algorithm using the decision tree 15 identifies a first group 101 of data having similarities on several criteria since most of the data of the first group correspond to data taken during one of the first 5 days of the week and data on a measurement day of a parameter greater than or equal to an average of a given threshold. A second group 102 of data is also identified.
    [0026] In this example, most of the data in the second group corresponds to data taken on the 6th or 7th day of the week. A third group 103 of data is also identified. The third data group 103 includes a subgroup 1031 of data taken during one of the first 5 days of the week and corresponding to a measurement of a parameter lower than a given threshold. The third data group 103 includes another subgroup 1032 of data taken on the 6th or 7th day of the week. The algorithm names for example the first data group "day [1-5] and aver> = 75". In addition to helping to naming projected data groups, the naming algorithm and its associated decision tree thus make it possible to highlight the membership criteria for a group of data for the user. The user can then re-name a group based on his or her understanding of it.
    [0027] A step (step E126) of displaying a graphical summary summarizing the characteristics of all or a subset of the data groups can then be performed. The view on which the projections of data sections are shown may be a dashboard window as shown in Figure 4A.
    [0028] On this window, four clusters corresponding to four groups 401, 402, 403, 404, of projected data sections distinguished by different colors (in Figure 4A the colors being replaced by gray levels) are shown. A selection of stub (s) or a group or a subgroup can be made by the user using a selection graphic tool 420 appearing for example in the form of a polygon or a a selection lasso. The selection of section (s) in the window of FIG. 4A can cause a selection of these same segments (s) on other windows of the dashboard and vice versa. The user U can thus navigate through the mass of data using a plurality of windows giving views complementary to that of the projection 20 illustrated in FIG. 4A. Specific sections 411 located at a given distance from the clusters may be outliers to eliminate that the user U can select and then delete. As indicated above, the respective sections of data groups 401, 402, 403, 404 may be associated with an identifier.
    [0029] In another window of the dashboard given in FIG. 4B, the data sections belonging to the first group 401 are identified as "working days". The sections of the first group 401 may for example correspond to days during which the site is in operation so that its power consumption is greater than a given threshold and is substantially constant over a given time interval of the day.
    [0030] The sections belonging to a second group 402 are identified as "closed days". The sections of the second group may correspond to days during which the site is not in operation so as to generate, for example, a power consumption below a given threshold and which is substantially constant over the aforementioned time interval. Sections belonging to a third group 403 are identified for example as "start-up days" and correspond to days during which the site is in operation for only part of the day. Segments belonging to a fourth group 404 are identified, for example, as "closing days" and correspond to days during which the site is only in operation for part of the day. Another window of the dashboard illustrated in FIG. 4C gives an energy consumption of the site, in particular its power consumption, over a period corresponding to that according to which the segmentation of the time series of data has been carried out previously. In the example of FIG. 4C, the curves 411, 412, 413, 414 are representative of typical power consumption profiles during one day, respectively for data originating from the first group 401, for data originating from the second group 402, for data from the third group 403, and for data from the fourth group 404. On another dashboard window given in FIG. 4D, measurement data from sensors of the measuring device 10 are shown for a range. analysis time selected for example using the window of Figure 4H listing the dates at which the measurements were made.
    [0031] The dashboard may comprise other windows as illustrated for example in FIGS. 4E, 4F, 4G giving complementary views, respectively in the form of a monthly histogram, of a histogram of the days of the week, and of a calendar on which the projected data is represented in a user-identifiable manner and making it possible to match the groups 401, 402, 403, 404 of the window of FIG. 4A.
    [0032] In the example of FIGS. 4E-4G, this correspondence can be established by the user using different hues (in this case grayscale) taking again those used in the window of FIG. 4A. The windows of Figures 4F and 4G provide the user with additional information with respect to the windows of Figures 4A and 4B. For example, they indicate that the second group 402 data identified as "closed day" is essentially data taken from the site on a Saturday. A graphical synthesis comprising one or more of the previously described windows 10 for one or more groups can then be integrated into a report. A backup in the form of project type files can also be performed. The results of the previously described processing may optionally be tabulated, for example in Excel spreadsheets that the tool 31 is configured to generate. The graphical interface of the software tool can be implemented so that, in addition to the first window of FIG. 4A, other windows from those previously described make it possible to create or edit groups of data sections.
    [0033] Recursive selection tools may also allow in the previously described dashboard to switch from a selected group of sections to a subgroup of that selected group. After the establishment of groups of data sections, a digital model establishment phase characterizing each group of sections can then be implemented. The construction of the individual model characteristic of a particular group can be realized by means of a learning algorithm in which for each new section of data is verified its belonging to this particular group or if this section is different from all that which has been seen previously.
    [0034] For this, it is possible to perform in particular an analysis of time profile (s) of evolution of measurement signals from the measuring device and which are contained in this new section. An example of a time profile gives an electrical consumption for the duration of a section, for example a daily electrical consumption when sections of duration equal to one day are produced. The learning algorithm of the characteristic individual model can for example be of kernel type with principal component analysis (also called "KerneIPCA"). A support vector machine type algorithm for a class (in English "one clans Support Vector Machine" or one-clans SVM) can also be used. This type of algorithm includes learning a decision function for a novelty detection. The learning algorithm can also be of the Sequential Minimal Optimization (SMO) type.
    [0035] A learning algorithm selection of the characteristic individual model can be performed by the user U using a control window of the graphical interface. Once established, the individual pattern characteristic of a group can be saved. In addition to the individual model characteristic of a group, a regression model can be implemented. The regression model is capable of establishing a data prediction of an element of the measuring device 10, for example one or more sensors (s) or one or more counters (s), from data from one or more several other elements (sensor (s) or meter (s)).
    [0036] To establish this regression model, the user can choose a second type of learning algorithm from among several learning algorithms. A Ridge Regression type algorithm based on a polynomial model, or a Lasso regression on a polynomial model, or a RVM relevance vector machine (acronym for "relevance 30 vector machine"). ") can be used. The learning algorithm selection of the regression model 3032786 can also be performed by the user U using a control window of the graphical interface. The tool 31 may be further configured to form for another projected data group another type of individual digital model called the individual calendar template. This model can use the naming help decision tree discussed above to determine which group of data legs a particular section belongs to by receiving a time indication associated with that particular data section. A calendar individual numerical model gives a condition of membership of new data to the first group from time metadata associated with said new data. Thus with the help of such a calendar model a group called "closed days" can be modeled for example by sections corresponding to "all Sundays AND French holidays EXCEPT easter". A computer system 50 for monitoring site performance as previously discussed in connection with FIG. 1 regularly collects, for example every day, data from the measuring device 10, to present them to a user under for example, dashboards, alarm tables, monitoring indicators, in particular Key Performance Indicators (KPIs).
    [0037] Such a performance monitoring system 50 may use one or more of the models described above to enable, for example, to evaluate possible energy savings made at the site under study or to detect any energy consumption drift of a site. An example of processing performed by the tracking system 50 when receiving new data will now be described in conjunction with FIG. 7. When new data is received by the tracking system, the tracking system 50 may calendar model to allow to determine, from the temporal metadata associated with these new data, which particular group these new data correspond (step 51).
    [0038] Then, the system 50 uses the characteristic profile pattern of that particular group to verify the membership of these new data to the particular group (step S2). In this case, either a membership of the new data 5 is detected in the particular data groups or the new data is classified as "abnormal" (step S3). In this case, a user of the performance monitoring system can then decide whether to classify these new abnormal data as outliers (step S31) or to associate them anyway with a group of data (step 10 S32). In the case where a match exists (step S4) and the membership of the new data to the particular data group is detected using both the calendar model and the characteristic individual model, a regression model associated with that group is used to monitor energy performance (step S5). For example, it is possible to detect through the calendar model that new data corresponds to a group of data sections named "working days" and to verify through the individual characteristic model that the measurement data profile associated with or contained in these new data correspond to a characteristic profile of measurements made during working days. In this case, it is then possible to use a corresponding regression model to obtain a nominal power consumption value (step S5), for example of E = 12 kWh of a working day and to be able to compare with this nominal value (step S6 ).
    [0039] A significant difference with the nominal value, for example detected with an absolute or relative value detection threshold (step S7), can then be interpreted as an error (step S71) or used to correct the corresponding model. working days (step S72).
    [0040] Such a system can be used to track the effect of an improvement that is made on the site to reduce its energy consumption or to follow an unintended deviation that deserves to be reported.
    权利要求:
    Claims (16)
    [0001]
    REVENDICATIONS1. Data processing method for analyzing the energy consumption of at least one site from measurement data, including electrical consumption measurement data as well as measurement data of one or more physical parameters, at least one measuring device (10) disposed at this site, the method comprising the steps of, through a computer processing system: a) selecting, within a selected time range of analysis, series temporal data from the measuring device, the time series of data being associated with explanatory metadata of said time series of data, b) segmenting the time series of data into a plurality of data sections of the same duration, c) performing a projection of the data sections selected and located in this analysis time range in a space with at least two dimensions in function n of their degree of similarity, d) displaying in a software graphical interface by means of a display device connected to the computer processing system, a graphical representation of said space including projections of the data sections, certain projections being grouped under one or more groups of data sections; e) selecting at least a first group of data sections; f) establishing one or more digital models characterizing the first group of data sections selected.
    [0002]
    2. Method according to claim 1, wherein among said models established in step f) is an individual numerical model characteristic of the first group constructed from an analysis of measurement time profiles contained in the data sections of a said groups. 3032786 25
    [0003]
    3. The method according to claim 2, further comprising, after step f), steps of: - acquiring new data, - using the characteristic individual numerical model to check the membership of the new data to a group of data.
    [0004]
    4. Method according to one of claims 1 to 3, wherein among said models established in step f) is a calendar individual numerical model giving a condition of membership of new data to the first group from associated time metadata to said new data.
    [0005]
    5. Method according to one of claims 1 to 4, wherein the measuring device (10) is provided with one or more measuring elements (s) including at least one counter (11) of electrical consumption and / or a or more sensors (13) for measuring physical quantities (s), the method further comprising establishing a regression model adapted to provide an estimation of the measurement data of at least one selected element of the measuring device to from data from one or more other measuring elements of the measuring device. 20
    [0006]
    6. Method according to one of claims 1 to 5, comprising the steps of: establishing for the first group a typical profile of magnitude (s) measured (s) by the measuring device (10) for a duration corresponding to that sections, 25 - display this profile type through the graphical interface.
    [0007]
    7. The method according to one of claims 1 to 6, further comprising the steps of: - selecting, through the software graphical interface, a particular group of data sections, 3032786 26 - associating this particular group to an identifier.
    [0008]
    The method of claim 7, wherein the identifier is generated using a naming algorithm implementing a decision tree and using metadata associated with the data stretches of the particular group.
    [0009]
    The method according to one of claims 1 to 8, wherein the selection of the time series of data in step a) comprises a selection of at least one type of metadata, in particular: a descriptor of area of the site for selecting time series of measurement data from one or more particular areas of the site from among a set of areas of the site, and / or a type of measurement device descriptor for selecting time series of data from one or more types of particular measuring means from a set of measurement means of the measuring device; and / or from a site measurement parameter type descriptor for selecting time series of relative data. one or more types of physical parameters out of a set of physical parameters that the measuring device is capable of measuring.
    [0010]
    The method of one of claims 1 to 9, further comprising selecting, through the software graphical interface and from among said section projections, one or more particular projections that do not belong to a group.
    [0011]
    The method according to one of claims 1 to 10, wherein in said graphical representation is located in a first window of the software graphical interface and wherein the sections are further represented in a form of another graphical representation in a second window of the graphical interface, the selection of the first group in said first window resulting in a selection and highlighting in said second window sections of the first group. 5
    [0012]
    12. The method of claim 11, wherein the selection of the time series of data performed in step a) or the selection of the first group performed in step e), is performed through this other graphical representation.
    [0013]
    13. The method according to one of claims 1 to 12, wherein the degree of similarity in step c) is established using a metric, the metric being selected through the software graphical interface. from a list of several different metrics.
    [0014]
    14. Method according to one of claims 1 to 13, comprising the steps of: - receiving new data, - detecting by means of a calendar model, from the temporal metadata associated with these new data, if these new data belong to a particular group among said groups, 20 - check the membership of these new data to the particular group using a characteristic individual model associated with the particular group, - use of a regression model to obtain a nominal energy consumption value, - compare the nominal consumption value with the new data.
    [0015]
    A computer program comprising program code instructions to enable the computer processing system to perform the steps of the method of any one of claims 1 to 14. 3032786 28
    [0016]
    16. A digital data medium usable by computer processing means, comprising code instructions of a computer program for carrying out the steps of the method according to any one of claims 1 to 14.
    类似技术:
    公开号 | 公开日 | 专利标题
    EP3059682A1|2016-08-24|Data-processing and modelling system for analysing the energy consumption of a site
    EP2696344B1|2019-01-02|Method and system for detecting acoustic events in a given environment
    US20170342962A1|2017-11-30|Synchronizing data from irregularly sampled sensors
    EP2000780B1|2017-02-15|Method and system for detecting and estimating the electric power consumption of a subscriber's facilities
    WO2010122056A2|2010-10-28|System and method for detecting abnormal audio events
    FR2963448A1|2012-02-03|METHOD AND SYSTEM FOR ANALYSIS OF FLIGHT DATA RECORDED DURING THE FLIGHT OF AN AIRCRAFT.
    US20100332437A1|2010-12-30|System For Generating A Media Playlist
    WO2018020181A1|2018-02-01|Method for determining a maximum allowable volume of water that can be removed over time from an underground water source
    EP3172548B1|2018-09-26|Method for detecting anomalies in a distribution network, in particular for drinking water
    EP3767467A1|2021-01-20|Method and device for determining a performance index value for prediction of anomalies in a computer infrastructure from performance indicator values
    US20180046599A1|2018-02-15|Automatic detection of outliers in multivariate data
    EP3343839A1|2018-07-04|Mechanism for monitoring and alerts of computer system applications
    KR102158100B1|2020-09-22|Auto monitoring method and apparatus by using anomaly detection
    EP3489831A1|2019-05-29|Method and device for monitoring a data generator process of a metric for the prediction of abnormalities
    EP3736743A1|2020-11-11|Method for designing a signature generator of an entity performing actions in a computer architecture, method for detecting abnormal behaviour, corresponding computer program and system.
    EP3767558A1|2021-01-20|Method and device for determining an estimated duration before a technical incident in a computer infrastructure from performance indicator values
    EP3846087A1|2021-07-07|Method and system for selecting a learning model within a plurality of learning models
    EP3252563A1|2017-12-06|Method for determining a mobility context of a user carrying a device provided with inertial sensors
    KR20190085795A|2019-07-19|On-line Risk Response Monitoring System through Analysis of Negative Buzz Media Transition and Amount of Conformity
    EP3767474A1|2021-01-20|Method for analysing resource consumption of a computer infrastructure, alert and sizing
    EP3846047A1|2021-07-07|Method and system for identifying relevant variables
    WO2015092279A1|2015-06-25|Method for locating an anomaly detected in a tree structure
    EP3748373B1|2021-08-04|Diagnosis of electrical distribution network
    FR3103611A1|2021-05-28|Method and device for monitoring the activity of a person with loss of autonomy
    WO2019002582A1|2019-01-03|System and method for managing risks in a process
    同族专利:
    公开号 | 公开日
    FR3032786B1|2017-03-24|
    EP3059682A1|2016-08-24|
    US10482204B2|2019-11-19|
    US20160371412A1|2016-12-22|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    CN107944146A|2017-11-28|2018-04-20|河海大学|Polynary hydrology Time Series Matching model building method based on principal component analysis|
    CN112102891A|2020-08-04|2020-12-18|广东工业大学|Horseshoe flame glass melting furnace energy consumption abnormity positioning method based on root cause analysis hierarchical clustering|US20030101009A1|2001-10-30|2003-05-29|Johnson Controls Technology Company|Apparatus and method for determining days of the week with similar utility consumption profiles|
    US20080177994A1|2003-01-12|2008-07-24|Yaron Mayer|System and method for improving the efficiency, comfort, and/or reliability in Operating Systems, such as for example Windows|
    US7146288B1|2002-07-18|2006-12-05|Johnson Controls Technology Company|System and method for estimating quantization error in sampled data|
    BRPI1012188B1|2009-05-08|2020-10-27|Accenture Global Services Limited|computer implemented method for energy analysis and energy analysis system.|
    US8788097B2|2009-06-22|2014-07-22|Johnson Controls Technology Company|Systems and methods for using rule-based fault detection in a building management system|
    US8239178B2|2009-09-16|2012-08-07|Schneider Electric USA, Inc.|System and method of modeling and monitoring an energy load|
    WO2012103485A2|2011-01-28|2012-08-02|The Board Of Regents Of The Nevada System Of Higher Education On Behalf Of The Desert Research Institute|Signal identification methods and systems|
    US10554077B2|2011-12-13|2020-02-04|Schneider Electric USA, Inc.|Automated monitoring for changes in energy consumption patterns|
    US20130151212A1|2011-12-13|2013-06-13|Schneider Electric USA, Inc.|Systems, methods and devices for determining energy conservation measure savings|
    US9116519B2|2013-03-15|2015-08-25|Gridpoint, Inc.|Method for implementing quality alarms in an energy management system|
    US20150276508A1|2014-03-27|2015-10-01|Palo Alto Research Center Incorporated|Computer-Implemented System And Method For Externally Evaluating Thermostat Adjustment Patterns Of An Indoor Climate Control System In A Building|WO2014145153A2|2013-03-15|2014-09-18|Neeley John|Automatic recording and graphing of measurement data|
    US9857778B1|2016-10-07|2018-01-02|International Business Machines Corporation|Forecasting solar power generation using real-time power data, weather data, and complexity-based similarity factors|
    US10819116B2|2017-02-28|2020-10-27|International Business Machines Corporation|Forecasting solar power generation using real-time power data|
    US10802673B2|2017-08-11|2020-10-13|Salesforce.Com, Inc.|Multi-selection on a chart|
    US11158046B2|2020-01-24|2021-10-26|Overjet, Inc.|Estimating measurements of craniofacial structures in dental radiographs|
    CN111860700B|2020-09-22|2020-12-15|深圳须弥云图空间科技有限公司|Energy consumption classification method and device, storage medium and equipment|
    法律状态:
    2016-02-09| PLFP| Fee payment|Year of fee payment: 2 |
    2016-08-19| PLSC| Publication of the preliminary search report|Effective date: 20160819 |
    2017-02-13| PLFP| Fee payment|Year of fee payment: 3 |
    2018-02-06| PLFP| Fee payment|Year of fee payment: 4 |
    2020-02-20| PLFP| Fee payment|Year of fee payment: 6 |
    2021-02-23| PLFP| Fee payment|Year of fee payment: 7 |
    2022-02-24| PLFP| Fee payment|Year of fee payment: 8 |
    优先权:
    申请号 | 申请日 | 专利标题
    FR1551302A|FR3032786B1|2015-02-17|2015-02-17|DATA PROCESSING AND MODELING SYSTEM FOR ANALYZING THE ENERGY CONSUMPTION OF A SITE|FR1551302A| FR3032786B1|2015-02-17|2015-02-17|DATA PROCESSING AND MODELING SYSTEM FOR ANALYZING THE ENERGY CONSUMPTION OF A SITE|
    US15/044,516| US10482204B2|2015-02-17|2016-02-16|System for processing data and modelling for analysis of the energy consumption of a site|
    EP16156010.7A| EP3059682A1|2015-02-17|2016-02-16|Data-processing and modelling system for analysing the energy consumption of a site|
    [返回顶部]