MANAGEMENT SYSTEM

The work leading to this invention has received funding from the European Union's Seventh Framework Programme (FP7/2007-2013) under grant agreement n° 285729.

The present invention relates to a method for monitoring a building management system, BMS, wherein observations of at least one data point within an observation window are recorded or provided, further comprising the following steps: statistically analysing the observations by a statistical measure, in order to obtain statistical data or a statistical characteristic of the observations, and detecting a logical and/or physical misconfiguration and/or deterioration of the BMS or of a sensor, actuator, or sub-system of the BMS, and/or detecting a change to the core and/or shell of the building, based on a deviating of the obtained statistical data or statistical characteristic of the observations from an expected statistical characteristic or feature.

Further, the present invention relates to a system for monitoring a building management system, BMS, comprising means for recording or providing observations of at least one data point within an observation window, the system further comprising: means for statistically analysing the observations by a statistical measure, in order to obtain statistical data or a statistical characteristic of the observations, and means for detecting a logical and/or physical misconfiguration and/or deterioration of the BMS and/or means for detecting a change to the core and/or shell of the building, based on a deviating of the obtained statistical data or statistical characteristic of the observations from an expected statistical characteristic or feature. At present, buildings and their BMSs, heating, ventilation, air conditioning, HVAC, and mechanical, electrical, and plumbing, MEP, systems are designed in a one-off process during the pre-construction design phase of the building. The BMS is typically installed and configured during the construction process and not reconfigured or re-calibrated afterwards, even if the usage patterns change or the HVAC or MEP systems are modified. As a result, poorly configured and even positively misconfigured BMS systems are a common phenomenon.

One of the causes for this is that the BMS of a complex building easily has several thousand sensors and actuators and that thus configuring and calibrating is time- consuming, expensive, and requires great skills and experience.

Such poorly configured, mis-configured or not well-understood buildings and characteristics are a major cause for inefficiencies and wasting of resources. However, as mentioned, getting a firm grasp of the dynamics and characteristics of a building and its equipment, appliances, and configuration is something that is rarely achieved.

Within the present document the following definitions have to be considered:

• data point = a sensor being measured or computed, i.e. a "virtual sensor", i.e. the variables for which statistics are derived. This explicitly includes data points being transformed by linear or non-linear operations such as averaging, applying the logarithm or even more complex operations such as Karhunen-Loeve transform, KLT.

• observation = a specific value of a data point at a specific time, at a specific event, or at another defined reference or trigger, e.g. when other, associated data points are read - observation window example see below.

• observation window = a series of consecutive observations. Here it must be noted that an Observation window' is not necessarily restricted to a time window but can be defined by any parameter, e.g. a start and stop temperature or pressure. Example: Let data point A be a temperature and let B be electrical energy. Then an observation window can e.g. range from T1 = 20° C to T2 = 40° C while observations of B are recorded at temperature intervals of 0.5° C. Due to the above mentioned complex of problems and difficulties of known BMSs, it is an object of the present invention to provide a method and system for monitoring a BMS which facilitate the assessment of a building ^' s performance and which provide information for a suitable configuration for saving energy.

In accordance with the invention, the aforementioned object is accomplished by a method for monitoring a BMS comprising the features of claim 1 and by a system for monitoring a BMS comprising the features of claim 26. According to the invention it has been recognized that a statistical analysis of observations of a data point by a suitable statistical measure provides the basis for an assessment of a building ^' s performance with high quality. Statistical data or a statistical characteristic of the observations obtained by the statistical measure are used for a further method step which comprises the detection of a logical and/or physical misconfiguration and/or deterioration of the BMS or of a sensor, actuator or sub-system of the BMS and/or the detection of a change to the core and/or shell of the building, based on a deviating of the obtained statistical data or statistical characteristic of the observations from an expected statistical characteristic or feature. The present invention greatly facilitates the assessment of a building ^' s performance and further supports owners and operators in understanding the building ^' s dynamics. The provided detection of a logical and/or physical misconfiguration and/or deterioration of the BMS or of a sensor, actuator or subsystem of the BMS and/or the detection of a change to the core and/or shell of the building can provide necessary information for a suitable configuration of the BMS for saving energy resources.

Within a preferred embodiment the expected statistical characteristic or feature can be derived from structural building data of a building or of a building information model, BIM, and/or from previously measured statistical reference data and/or metrics of a building or parts of the building or of a data point. In this way it is possible to identify changes of system data ^' s statistical characteristics deviating from expected statistical features in a simple way. For providing very meaningful information the previously measured statistical reference data and/or metrics can be measured over the entire life-cycle of the building or over definable parts of the entire life-cycle of the building. Thus, important parts of the entire life-cycle of the building can be used for the statistical analysis.

Within a further preferred embodiment the previously measured statistical reference data and/or metrics can be a previously obtained statistical data or statistical characteristic of the observations, so that the obtained statistical data or statistical characteristic of the observations is a n-th data set and the previously obtained statistical data or statistical characteristic of the observations is a corresponding (n-i)-th data set, wherein n and i are integers. Thus, definable previous measurements can be used for providing a very simple statistical measure and statistical analysis.

Within a further preferred embodiment the expected statistical characteristic or feature can represent expert knowledge on proper BMS behavior and/or proper BMS configuration, so that a deviation from said expected statistical characteristic or feature can indicate a deviation from an expected inter-system relation. This preferred embodiment provides the possibility of using expert knowledge during the statistical analysis. Thus, very reliable and meaningful detection of logical and/or physical misconfiguration and/or deterioration of the BMS and/or of a change to the core and/or shell of the building is possible. Within a further preferred embodiment the detecting step can comprise not only the detection but also the identifying of a misconfiguration and/or deterioration of the BMS and/or identifying of a change to the core and/or shell of the building. For such an identifying step previous statistical data or statistical characteristic of the observation can be used in an easy way.

Further preferred, an interrelation of a logical and/or a physical misconfiguration and/or a deterioration of the BMS can be detected. The disclosure of such an interrelation can help identifying causes for misconfigurations and/or deteriorations of the BMS. With regard to a widespread use of the obtained statistical data or statistical characteristic of the observations a database of this obtained statistical data or statistical characteristic of the observations can be built. Such a database can be accessed by specifically authorized persons or clients.

For providing a very meaningful data collection the database can comprise human expert knowledge and/or typical values for particular subsets of building statistics or for particular subsets of the obtained statistical data or statistical characteristic of the observations. Alternatively or additionally the database can comprise dependencies and/or misconfigurations and/or error sources of buildings, of their equipment and of BMS or control system. Thus, a very widespread use of the collected data is possible.

Within a further preferred embodiment the database can be cloud-based. This can provide a very simple access to the database for a lot of different users from various locations.

As a result the obtained statistical data or statistical characteristic of the observations and/or human expert knowledge can be retrieved from the database for use during statistically analysing the observations. This results in a very effective statistical analysis.

For providing a reliable locating of inefficiencies and changes in building physics a measurement statistic in form of the obtained statistical data or statistical characteristic of the observations can be related with BIM data for physically locating inefficiencies and changes in building physics. Thus, locating inefficiencies and changes in building physics can be provided by simply using BIM data. Within a further preferred embodiment a measurement statistic in form of the obtained statistical data or statistical characteristic of the observations can be related with BIM data and/or data from the database for rating the plausibility of the obtained statistical data or statistical characteristic or of subsets of the obtained statistical data or statistical characteristic. By such a rating of the plausibility obtained statistical data or statistical characteristic of the observations being not plausible or reasonable can be cancelled. This reduces the amount of data and simplifies any detecting step. Further preferred, the plausibility rating can be used to flag those obtained statistical data or statistical characteristic or those subsets of the obtained statistical data or statistical characteristic which are not sufficiently plausible for human expert inspection. This will simplify any detecting step and the method as a whole.

Within a further preferred embodiment the plausibility rating can be used to raise an alert. Such an alert can be provided or transmitted to a human expert and/or to a functional entity which can be enabled to provide any necessary measure for configuration or reconfiguration of the BMS or of a sensor, actuator, or sub-system of the BMS. As a result, a misconfiguration and/or a deterioration of the BMS and its functional entities can be removed.

For further enhancing the quality of the present method during statistically analysing the observations additional information regarding the weather conditions and/or from a suitable or relevant Information and Communication Technology, ICT, system or sensor, providing additional relevant information that helps to assess the building state, can be exploited.

Statistically analysing the observations by a statistical measure can comprise different steps depending on the individual situation. Thus, statistically analysing the observations can comprise at least one linear and/or non-linear and/or multivariate statistical measure. Alternatively or additionally statistically analysing the observations can comprise computing the pairwise correlation of two series of observations or applying analysis of variance for multiple series of observations.

Regarding the enhancement of quality of the present method the statistical measure can comprise repeatedly applying the statistical measure to the recorded and/or provided and/or obtained and/or measured and/or derived data or observations. A preferred embodiment of the method can comprise a step wherein, if a deviation of the obtained statistical data or statistical characteristic of the observations from an expected statistical characteristic or feature exceeds a definable threshold, a notification can be transmitted or indicated to a specified contact. Such a specified contact can be a device or a person which can be responsible for the configuration of the BMS.

Within a further preferred embodiment at least one scaling factor or other calculated parameter of the statistical measure can be stored within a database along with the obtained statistical data or statistical characteristic of the observations within the observation window.

Depending on an individual situation a data point can undergo at least one linear or non-linear transformation and/or calculation prior to the recording or providing of the observations or prior to the statistically analysing of the observations by the statistical measure.

Further depending on an individual situation a data point can be a result of a multivariate combination of other data points.

The present invention provides a system and method to aid understanding and identification of causes of building management systems ^' misconfiguration or misconfigurations as well as the support for identifying and physically locating the causes. Important aspects of preferred embodiments of the inventive method and system can be summarized as follows:

- A system and a method are provided to identify changes of system data's statistical characteristics deviating from expected statistical features as derived from structural building data, building information model, BIM, or previously measured statistical reference metrics over the entire life-cycle of a building or parts thereof. This step aims specifically at detecting system or BMS deterioration or undocumented changes to the building's core and shell. - A system and a method are built to detect logical and physical misconfigurations of building management systems and their interrelation based on changing / deviating data point measurement statistics.

- Measurement statistics can be related with BIM data for physically locating inefficiencies and changes in building physics.

- Measurement statistics can be related with BIM data in order to rate the plausibility of subsets of the statistical data and using these plausibility ratings to flag those subsets of a building's statistic which are not sufficiently plausible for human expert inspection.

- A cloud based database of building statistics can be built up representing a large pool of both human expert knowledge and typical values for particular subsets of building statistics which in turn can be used for plausibility checking of future assessment rounds and guidance for the experts.

Preferred embodiments of the inventive method can comprise the following steps:

1) Acquiring building data from e.g. building management system, BMS, or other sources.

2) Computing statistical metrics of the building data, representing the interrelations / correlations of multiple observations or of multiple series of observations or the autocorrelation of series of observations of a single data point.

3) Generating statistical metrics corresponding to those as described in 2) but based on structural building data, BIM, "blue print", i.e. the ideal synthetic building behaviour.

4) Storing computed real statistical metrics.

5) Comparing real metrics with either predicted or historical metrics in order to identify deviations which hint at undocumented changes, faults, or deterioration.

Further important features of embodiments of the present invention are:

- Data driven diagnosis of building system behavior. - Controllable and configurable thresholding mechanism to tune the sensitivity of the invention.

- Enhancing inaccuracy and inefficiency detection by drawing on additional BIM information for location.

- Both consecutive observation windows as well as past observation windows can be used to cope with seasonal changes and for errors compared to past phases.

- Observation windows are configurable.

- It is noteworthy that the comparison of consecutive observation windows with appropriate thresholds will track changes in system correlation statistics, e.g. due to seasons, without triggering superfluous notifications. It is also beneficial for this invention to rely on statistical measures to abstract from exceptional seasonal situations, e.g. a very cold winter, which would impact absolute measurement values but should not impact system inter-relations.

- Enhancing and facilitating human assessment step by pre-checking the plausibility of the building analysis by means of BIM data.

- Enhancing and facilitating human assessment step by pre-checking the plausibility of the building analysis by means of a cloud-based recommender system.

- Providing the option of feeding back current assessment results and ratings in the recommender system.

The present invention enhances the process of finding building system, e.g. energy, inefficiencies saving time and money. An exhaustive manual inspection of all building performance statistics is by no means feasible. This invention facilitates the automatic and manual assessment and interpretation of building performance statistics, the latter by pre-selecting only those values for human expert assessment which are significantly different to those as expected by querying the BIM model or the recommender system.

There are several ways how to design and further develop the teaching of the present invention in an advantageous way. To this end, it is to be referred to the patent claims subordinate to patent claim 1 on the one hand and to the following explanation of preferred examples of embodiments of the invention, illustrated by the drawing on the other hand. In connection with the explanation of the preferred embodiments of the invention by the aid of the drawing, generally preferred embodiments and further developments of the teaching will be explained. In the drawings

Fig. 1 is illustrating an embodiment of a method or system for monitoring a BMS according to the invention, which can identify malfunctions in buildings,

Fig. 2 is illustrating in a block diagram an embodiment of a method for monitoring a BMS according to the invention and

Fig. 3 is illustrating a further embodiment of the present invention in the form of a system for identifying poorly configured/misconfigured buildings and building management systems, supported by a cloud recommender system, BIM data, and human expert knowledge.

Fig. 1 is illustrating the overall concept of an embodiment of a method or system for monitoring a building management system, BMS, according to the invention. This embodiment of the present invention draws upon statistical analysis of measurements, observations, by the building management systems in question. The measurements may be either time-based or event-triggered. The more sensors are deployed and the higher the measurement frequency the more accurate the statistical analysis will be. Additional sources of information, e.g. weather databases on the public internet or other ICT systems such as access card readers may also be exploited in this statistical analysis. Furthermore, derived data may be considered as well, e.g. average values of several real sensors, calculated values, e.g. virtual sensors, etc.

The method according to the embodiment calculates and stores statistical characteristics of the observations. In this document, statistical analysis means the generic concept of evaluating data, e.g. by computing the pairwise correlation of two series of observations or applying - multivariate - analysis of variance for multiple series of observations by repeatedly applying these statistical methods to all these measured and derived data, Fig. 2, block A, the entirety of the building data will be included in the statistical analysis. For readability it is used the term "statistics" in the remainder of this document but it is explicitly noted that this includes various linear, non-linear or multivariate statistical measures such as Pearson product-moment correlation coefficient, Allison, Paul D. (1998) Multiple Regression: A Primer, distance correlation, Szekely, G. J. Rizzo, M. L. and Bakirov, N. K. (2007), Measuring and testing independence by correlation of distances, Annals of Statistics, 35/6, 2769-2794., the cross-correlation, Local Gaussian correlation, Dag Tjostheim, Karl Ove Hufthammer: Local Gaussian Correlation: A new measure of dependence (201 1 ), or multiple correlation, Allison, Paul D. (1998) Multiple Regression: A Primer, defined in the state of the art are considered as interchangeable for the purpose of this invention's use of calculations for statistical analysis of the data. The number of observations considered in the computation of the statistics is configurable, e.g. it can relate to a fixed time interval or can be event-based, e.g. 2 days before and 2 days after a scheduled conference in a hotel. Any new calculated statistical characteristics are compared to one or several of the previously stored statistical data sets, Fig. 2, blocks B, C, e.g. by computing the difference.

In general, the number of observations subjected to the statistical computations may vary among individual data-point pairs as well among data-points in general - standard techniques such as interpolation of data points may be used to facilitate the pair-wise or multivariate calculations.

In general, observation windows considered in the calculation of statistics may or may not overlap - resulting in different embodiments of the invention. Without loss of generality also observation windows of only a - configurable - subset of data points may be considered for computing the statistics.

Regarding Pearson's correlation and the distance correlation methods in particular - other statistical assessment methods work analogously -, the computed correlations represent to what degree two subsystems affect each other: This can be a positive correlation, i.e. if one reading changes so does the other, a negative one, in case of Pearson, i.e. the increase of one reading means also the drop of the other, or they can be uncorrelated. As input to our invention, a configurable tolerance threshold can be specified - only changes in value of correlation exceeding this threshold will trigger notifications to specified contacts - typically that of human experts, but could also be other expert systems - for further investigation. This mechanism indicates that the system behavior changed compared to the chosen reference correlation. The interpretation of the identified change in correlation is beyond the scope of the invention but, however, the invention supports the investigation in the following two ways:

• A notification of the identified change in correlation and the subsystem identifiers to the building operator, technical staff or other appropriate systems.

• If connected to a Building Information Model, BIM, structural data on the corresponding subsystems is provided in order to assist physically locating these, Fig. 2, block E. Also, cross system correlations may be detected that should not be expected based on the BIM information of the building layout - e.g. caused by a newly created doorway not modelled in the BIM data yet but indicated by a change in correlation values of measurement data among systems that should not be affecting each other based on the stored building physics.

Given that a current statistical characteristics data set can be compared to any of the historical, stored ones, different modes of operation are conceivable, resulting in different embodiments, the two most important ones being: · Comparing the current data set to a fixed historical one which is constant over time, e.g. created when the building took up routine operation. In this mode, a notification will be triggered as soon as difference exceeding a threshold is found between the current and the reference correlation data, no matter how long ago the reference data had been created. • Comparing the current data set (n) to a previous data set (n-i) in which case slowly changing system dynamics are not detected as they do not exceed the threshold value. For this mode of operation, a separate threshold value may be specified.

In another embodiment, scaling factors, e.g. the standard deviations of the data point pair considered for Pearson's correlation, and other calculated parameters of the chosen statistical method for computing the statistics are stored along with the data statistics for the observation window themselves. This allows undoing statistical operations of the considered observation window and applying e.g. stored parameters of the statistics of another - stored - observation window. This in particular allows to translate the statistics between the observation windows and is e.g. beneficial for tracking degradation of equipment over time. Note that this goes beyond the original definition of the chosen statistical method.

In another embodiment, the calculated statistical characteristics are compared to well formulated expectations about the statistical characteristics representing expert knowledge on proper system behavior and proper system configuration. Deviations from this expected statistical data indicate a deviation from the expected inter-system relations and thus may be indicative of malfunctions, misuse or exceptional events and thus should trigger further investigation. For example, it might be expected that for small time windows heating and cooling systems' power usage should have negative correlation as heating and cooling should ideally not be running simultaneously. Again, a thresholding or a tolerance mechanism may be used to suppress an over-sensitive system and to focus on real misconfigurations and inefficiencies. Note that the aforementioned expected correlation data may either be created on demand as shown in Fig. 2, block D or may have been created in advance and is treated as a specific historic reference correlation data set, block B.

Fig. 3 depicts the flow of computation of our invention. Generally, as mentioned in the definitions part, data points may undergo transformations and calculations prior to calculating the measurement statistics of the observation window. In particular as stated in the definitions section, data points are not limited to sensor measurements, but could also be a result of a multivariate combination of other data points.

Examples for transformations of data points are logarithm functions, the Karhunen-Loeve transform, KLT, or Principal Component Analysis, PCA, the Fourier Transform or other transformations known from the literature. These calculations and transformations apply prior to step A of Fig. 2.

To further assess the statistical data, to formulate expectations on the system's statistical characteristics, and to gather expert knowledge, see Fig. 2, block D and Fig. 3, an embodiment of the invention a) integrates structural data about the building to predict expected, plausible statistical characteristics. Again, using the correlation example, based on the structural building data, BIM, for all data point pairs a prediction can be made as to whether an interrelation is expected or not, e.g. a correlation is to be expected between {Window open/closed; A/C on/off} whereas no correlation is expected between {Light Store Room on/off; outside temperature},

b) matches the expected statistical characteristics with the real, computed statistical data and flags those entries which are implausible, the reason for this being that it cannot be reasonably expected that in a building with x>1000 data points a statistics matrix with x ² entries can exhaustively be assessed by a human expert so that only those parts of the real correlation data are flagged for manual human inspection which are significantly different from the automatically predicted ones,

c) presents these - pre-selected - implausible statistical items in an easily comprehensible manner to human experts in order to identify any inconsistencies or unexpected results in them, d) prompts human experts for their expert knowledge in order to assess and validate the plausibility of the automatically computed statistical data matrix of the building and its performance,

e) shares the assessment/rating of the validity and value of an interrelation between two data points with others, e.g. by uploading it into a cloud- based recommender system. The recommender system continually "learns" from the feedback provided by the expert assessments of buildings and thus creates a database of typical dependencies of building sub-systems, how these are affected by the geographic location of a building, e.g. Northern vs Southern Europe, by building type, e.g. hotel vs university, by size, and many other parameters. This accumulated expert knowledge can in future building assessments be used to complement the suggestions based on the BIM data alone,

f) downloads existing votes/assessments/ratings from the recommender cloud to further assist the assessment of the statistical metrics of the actual building being investigated, possibly also further filtering the information presented to a human expert in steps b - d and

g) creates a - cloud - database of typical dependencies and typical misconfigurations / error sources of buildings and their equipment and control systems which can be searched by e.g. building type, use pattern, geographical location, etc. representing a recommender system based on swarm ratings by a huge number of experts.

The below pseudo code listing for the case of Pearson's correlation value shows a basic variant of an embodiment of the present invention that is executed for configured observation windows. a) Calculate correlation matrix C for current data point observation window for selected data points

b) Load reference correlation matrix R for selected data points

c) Calculate difference correlation matrix D=C-R

a. For each two data points i and j, D has an entry d_ij

b. If d_ij > threshold t i. Access BIM entry of i and j, if available check for intersection

ii. Inform human of d_ij and available system information for data points i and j through appropriate means

d) Store C for later use as reference correlation matrix

Note that handling of data points being invariant during the entire observation window is considered to be implementation specific. These could be treated as 0, NAN or -Infinite or +lnfinite values with different implications on the d_ij calculations above.

Many modifications and other embodiments of the invention set forth herein will come to mind the one skilled in the art to which the invention pertains having the benefit of the teachings presented in the foregoing description and the associated drawings. Therefore, it is to be understood that the invention is not to be limited to the specific embodiments disclosed and that modifications and other embodiments are intended to be included within the scope of the appended claims. Although specific terms are employed herein, they are used in a generic and descriptive sense only and not for purposes of limitation.