Estimation of physical quantities through distributed measurement systems
By implementing local decision algorithms in a distributed power consumption measurement system, dynamically creating subsets participating in sampling, the problems of network bandwidth occupation and insufficient representation caused by conventional sampling techniques are solved, and high-precision and real-time physical quantity estimation is achieved. in large-scale distributed systems.
Patent Information
- Application Number
- CN201980071991.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2018-10-04
- Filing Date
- 2019-10-01
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2039-10-01
AI Technical Summary
In large-scale distributed power consumption measurement systems, conventional sampling technology causes the transmission of large-scale measurement values to occupy network bandwidth, increasing system-wide limitations, and static representative samples cannot effectively predict future measurement results, limiting the system expansion and accuracy .
By executing a local decision algorithm at each measurement device of the distributed measurement system in each sampling cycle, dynamically and randomly creating a subset of participating in the sampling, using the sending probability πi to determine whether the measurement device sends a measurement sample, achieving accurate and fast estimation of physical quantity.
It significantly reduces the number of measurements that must be sent, reduces network bandwidth usage, improves estimation accuracy and representativeness, allows for shorter sampling periods (such as second levels), and maintains the real-time and scalability of the system.
Smart Images

Figure CN112970165B_ABST
Abstract
Description
Technical Field
[0001] The present invention generally relates to the estimation of time-varying physical quantities by means of a distributed measurement system comprising a centralized platform and a set of multiple distributed measurement devices, each of which is capable of measuring local data representative of the physical quantity and remotely communicating with the centralized platform via a communication network. Background Art
[0002] One field of application to which the present invention is particularly, but not exclusively, directed is the management and / or monitoring of power consumption by electrical appliances (eg household appliances) located in the homes of a particularly large number of consumers and connected to an electricity distribution network.
[0003] In such a system, at least one electronic measuring device is usually installed in multiple locations (homes, commercial buildings, communities, etc.), which are locally connected to at least one monitored appliance so that the power consumption of this appliance can be measured, preferably continuously. Then, a dedicated centralized platform in the form of one or more servers obtains, preferably in real time, the power consumption measurement samples communicated by each local measuring device of the system. The centralized platform is thus able to determine the basic load curve, that is to say, the time variation of the relative consumption at each load or monitored appliance, and to derive the total load curve by aggregating these basic load curves.
[0004] For systems dedicated to monitoring only electricity consumption, the measuring devices distributed throughout the various sites may typically be "smart" meters capable of measuring the electricity consumed by one or more electrical appliances so as to periodically send electricity consumption measurement samples.
[0005] The applicant has also proposed a distributed system for regulating the consumption of a collection of a very large number of electrical appliances (such as water heaters, electric heaters, air conditioners, etc.) located in the homes of a large number of users, wherein the various load profiles obtained by the distributed measuring devices can in particular provide a distributed load shedding service, by which a subset of measuring devices can be selected in real time from the collection of all measuring devices locally connected to the electrical appliances, to which the system (in particular a centralized platform) sends a command to temporarily stop operation in order to reduce the power consumption of the collection by a given set value. Systems of this type are described, for example, in application WO2008 / 017754 or in application WO2012 / 172242. The system is also able to selectively regulate the power supply to certain appliances by analyzing the consumption of a large number of users, via stop and restart commands, which makes it possible to adapt the power consumption to the power generation available at a given time.
[0006] When measurement systems are distributed on a large scale, where the collection includes a large number of distributed measurement devices, the knowledge of physical quantities summarized from the measurement samples acquired by the various measurement devices requires the measurement devices to transmit a very large number of measurement values, typically hundreds of thousands (or even more). This takes up a large amount of network bandwidth, regardless of what communication network is used to transmit the data. In addition, for certain types of communication networks, such as cellular networks, the amount of data required for this transmission carries a huge cost. Therefore, this problem constitutes a limitation to increasing the scope of these systems (i.e., increasing the number of distributed measurement devices).
[0007] The above mentioned problem increases in large-scale distributed power consumption measurement systems such as described in the above cited applications WO2008 / 017754 and WO2012 / 172242 for which it is important to know exactly the aggregated load profile in real time based on the real-time transmission of a large number of consumption measurements.
[0008] In such a system, the centralized platform obtains power consumption measurement samples with a sampling period that is typically on the order of ten minutes. By reducing the sampling period, for example, by using a much reduced sampling period (typically on the order of ten seconds or even one second), a more accurately summarized load profile can be obtained. In this case, the number of measurements that must not only be collected but also summarized by the platforms in the set within the sampling period becomes too large.
[0009] In order to solve the problem caused by the large number of measurements collected (which directly depends on the number of measuring devices in the set and / or on the sampling frequency of the measurements), it is known practice to use a limited number of measurements collected within a subset (called a sampling subset) and then estimate the physical quantity (e.g. the total load curve) by extrapolation from the measurements collected in this subset. The accuracy of the estimate depends strongly on how the sampling subset is constructed.
[0010] However, the known solutions to date are based on conventional sampling techniques for which the formation of a representative sample (i.e., in this case, a subset of the measuring devices that are sampled) is performed routinely, once and for all, exhaustively and at the initiative of the sampler, completely randomly, by stratification, by cluster sampling or by quota sampling.
[0011] Therefore, these types of sampling only provide a pre-selection of a portion of the measuring instruments that constitute the set in the system, which belong to what is considered a representative sample due to observations of past behavior (consumption in the past hour, the past day, etc.) or a certain configuration (measured devices, subscriber price, location, etc.).
[0012] Regardless of the method of formation, the representative subset or representative sample is static or nearly static, ie, the composition of the sample is nearly constant with respect to the frequency at which measurements are made.
[0013] However, aside from the obvious difficulty of selecting a representative subset over an observation period, there is no guarantee that such representation observed in the past will adequately predict future measurements.
[0014] Experience has shown that for a subset to be representative of future observations, the size of the sampled subset must be very large, which greatly limits the advantage of sampling. Summary of the invention
[0015] It is an object of the present invention to overcome the disadvantages of systems using the aforementioned conventional sampling techniques.
[0016] To this end, a subject of the invention is a method for estimating a time-varying physical quantity by means of a distributed measurement system, the distributed measurement system comprising a centralized platform and a set consisting of a plurality of distributed measurement devices, each of the plurality of distributed measurement devices being capable of remotely communicating with the centralized platform via a communication network, the method comprising the following steps: a step of measuring local data representing the physical quantity by means of each measurement device in the set; a step of collecting measurement samples originating from a subset of the set within a predetermined sampling period T, wherein each measurement device belonging to the subset sends a measurement sample P of local data to the centralized platform via the communication network. i ; and based on the collected measurement samples P i The method comprises, in each sampling period T, the following phases: executing a local decision algorithm at each measuring device of the set, the local decision algorithm being able to calculate the transmission probability π associated with each measuring device based on the transmission probability π of the measuring device. i stochastically decides whether the corresponding measuring device is part of the subset, and the method is characterized in that, based on the collected measurement samples P, a calculation is performed according to the following relationship: i and the transmission probability π associated with each measuring device belonging to the subset i To estimate the physical quantity in each sampling period T:
[0017]
[0018] Wherein, N' is the number of measuring devices in the subset.
[0019] The invention thus enables the dynamic and random creation in each sampling cycle of a subset of measuring devices that will participate in sampling and, by aggregation, allow accurate and rapid estimation of physical quantities. In addition, the creation of the subset is performed only with the aid of such measuring devices that randomly decide whether to participate in sampling in each sampling cycle.
[0020] According to a possible implementation of the method according to the invention, the transmission probability π associated with at least one measuring device in the set i According to the obtained measurement sample P i and / or the conditions under which the measurement is performed and / or the measurement time; in this case, if the at least one measuring device is part of the subset, the at least one measuring device compares the transmission probability with the measurement sample P of the local data via the communication network in the acquisition step. i At the same time, it is sent to the centralized platform.
[0021] According to another possible embodiment of the method according to the invention, in the acquisition step, each measuring device belonging to the subset transmits the probability π via the communication network. i The measured sample P with local data i At the same time, it is sent to the centralized platform.
[0022] In one possible implementation, the method further comprises updating a local decision algorithm and / or a transmission probability π associated with at least one measuring device in the set. i steps.
[0023] In some cases, the sampling period T may be less than or equal to 10 minutes.
[0024] The sampling period T may particularly advantageously be substantially of the order of seconds or about ten seconds, which is particularly useful for systems that have to estimate physical quantities within the sampling period in real time.
[0025] Another subject of the present invention is a distributed measurement system for estimating a time-varying physical quantity, the system comprising a centralized platform and a set consisting of a plurality of distributed measurement devices, each of the plurality of distributed measurement devices being capable of remotely communicating with the centralized platform via a communication network, the system being configured such that in a measuring step, each measurement device in the set measures local data representing the physical quantity in a collecting step, and each measurement device belonging to a subset of the set sends a measurement sample P of the local data to the centralized platform via the communication network within a predetermined sampling period T. i, and in the estimation step, the centralized platform estimates the physical quantity within the sampling period based on the summary of the collected measurement samples, characterized in that each measuring device in the set comprises a software module configured to execute a local decision algorithm in each sampling period, the local decision algorithm being able to determine the local decision algorithm based on the transmission probability π associated with each measuring device i The centralized platform is configured to randomly determine whether the corresponding measuring device is part of the subset, and is characterized in that the centralized platform is configured to determine whether the corresponding measuring device is part of the subset by calculating according to the following relationship based on the collected measurement samples P i and the transmission probability π associated with each measuring device belonging to the subset i To estimate the physical quantity in each sampling period T:
[0026]
[0027] Wherein, N' is the number of measuring devices in the subset.
[0028] For example, the local data is the power drawn by a local load comprising one or more household appliances connected to the power distribution network, then the physical quantity is the total power consumed. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] The invention will be better understood by reading the following description given with reference to the accompanying drawings, in which:
[0030] Figure 1 An example of the architecture of a distributed measurement system capable of implementing the present invention is given;
[0031] Figure 2 It is shown that according to the invention, for example, Figure 1 The steps implemented in the method for estimating a physical quantity by a system. DETAILED DESCRIPTION
[0032] In the remainder of this disclosure, the invention will be described in the context of determining an overall load curve comprising an aggregation of basic load curves corresponding to the time evolution of the power consumption of a plurality of electrical appliances. Thus, in this case, the physical quantity to be estimated is the total power consumed by a plurality of electrical appliances. As mentioned above, the principles of the invention can still be applied to other physical quantities, for example, to other types of power consumption.
[0033] Figure 1The architecture of a distributed measurement system 1 is shown, which allows estimating the power consumed by electrical appliances (not shown) connected to a power distribution network (not shown) and located at a number N of sites (e.g. household appliances). The system comprises a centralized platform 1 hosted by one or more Internet servers and N measuring devices 21, ... 2 distributed at N sites. i , ……2 N Throughout the rest of the description, the index i is used for any reference to a measuring device 2i located at site i, i being an integer from 1 to a maximum value N.
[0034] Each measuring device 2 i It is possible to communicate remotely with the centralized platform 1 via the communication network 3. To this end, each measuring device 2 i including, for example, a wireless communication modem 20 i , which allows to establish a packet telephony type connection (such as GPRS, 3G or 4G) with a corresponding communication modem 10 at the centralized platform 1. Alternatively, the connection can be made via an Ethernet type link.
[0035] Each measuring device 2 i It is also possible to measure voltage and current via the module 22 i Measuring device 2 measures the power consumed locally by the connected appliances. i is, for example, a smart meter installed at each location i. In another possible architecture where the system can modulate power by providing distributed load shedding services, each measuring device 2 i Consists of one or more units, such as the driver unit and modulator box described in the above-referenced applications WO2008 / 017754 and WO2012 / 172242, incorporated by reference. Unlike the meters, these devices are also able to execute commands to stop / restart the operation of one or more appliances connected to them, these commands being received from a centralized platform for managing limited load shedding.
[0036] In either case, in order to avoid a large amount of data transmission, the system is configured to estimate the total power consumed based on measurement samples acquired by a subset or sampling subset of measurement devices that are considered representative. Specifically, the system is configured to implement the following steps:
[0037] A measuring step, wherein each measuring device 2 in the set of N devices i Measure the local power consumption Pi;
[0038] A collection step, wherein only N' measuring devices 2 corresponding to the sampling subset whose number is less than N are collected. iThe locally consumed power P is sent to the centralized platform 1 via the communication network 3 within a predetermined sampling period T. i of measurement samples; and
[0039] An estimation step, in which the centralized platform 1 estimates the total power consumed within the sampling period T based on a summary of the collected measurement samples.
[0040] Unlike systems using conventional sampling techniques with a fixed representative subset, the system according to the invention provides for changing the range of the sampling subset in each sampling period T, ie selecting the measuring devices whose measured values will actually be transmitted in the acquisition step.
[0041] To this end, according to the invention, the N measuring devices 2 i Each of the above executes a local decision algorithm in each sampling period T, the local decision algorithm being stored in the software module 21 i and can be based on each measuring device 2 i The associated sending probability π i , randomly decides whether the corresponding measuring device will measure its sample P i Sent to the centralized platform 1.
[0042] In other words, each measuring device 2 i Whether the measuring device will be part of the sampling subset that sends measurement samples within the sampling period T can be decided randomly.
[0043] In one possible embodiment, all measuring devices 2 i With the same predefined transmission probability π within period T i and the same local decision algorithm.
[0044] The local decision algorithm may follow a Bernoulli distribution and implement, for example, the following steps:
[0045] Generate a random number, whose value is between 0 and 100;
[0046] If the generated random number is less than the probability π assigned to the send i The percentage value (for example, less than 10), if the probability of sending i is equal to 10%, then the measuring device 2 i The measurement samples Pi are sent to the centralized platform 1 (in other words, the measurement device then forms part of the sampling subset).
[0047] Otherwise, measuring device 2 i Do not measure sample P i Sent to the centralized platform 1.
[0048] In another possible embodiment, the probability π is sent i It is possible to have a fixed value predefined within the period T, but it is different for each measuring device.
[0049] In another possible embodiment, for one or more measuring devices 2 in the set i (or even for all measuring devices) with a probability of sending π i The value of can be obtained according to the measurement sample P i and / or the conditions under which the measurements are made and / or the time of the measurements are variable. Thus, by way of non-limiting example, if the measurement samples P acquired within the sampling period in question i If the value of is equal to zero, it can be specified that the measuring device 2 i The sending probability π i In other words, the measuring device 2 i A zero measurement is not sent and will not form part of the sampling subset. i is part of a modulation system providing a distributed load shedding service, it can also be provided, for example, that if the measurement samples P taken within the sampling period in question i If the value of is not zero, then the measuring device 2 i The sending probability π i is equal to 100%, the command for disconnecting the appliance connected to the device should be executed. In other words, once the measuring device 2 i Considered abnormal, measuring device 2 i It can choose to send its measurements systematically.
[0050] As will become apparent below, the centralized platform must know the transmission probability π corresponding to any measurement sample it collects i Therefore, it is provided that in the acquisition step 120 the measuring device 2 i The probability of sending it via the communication network 3 is π i With the measured sample P i At the same time, the measurement device 2 is sent to the centralized platform 1. i belongs to a sampling subset and its transmission probability changes due to any reason.
[0051] In another possible embodiment, it is provided that in the acquisition step 120, each measuring device 2 belonging to the subset i A predefined transmission probability π is to be systematically transmitted via the communication network 3 i (regardless of whether the transmission probability is variable or immutable) and the measured sample P i At the same time, it is sent to the centralized platform 1.
[0052] It can be specified that at least one or more measuring devices in the set, or even all measuring devices, can execute a local decision algorithm and / or a predefined transmission probability π i This update may be sent by the centralized platform 1. As a variant, the update is performed based on information calculated or measured locally by a measuring device.
[0053] Once the centralized platform 1 has collected the various measurement samples P sent by the N′ measurement devices of the sampling subset i and the corresponding transmission probability π where applicable i , it will be possible to process these samples at the processing module 11 and, in particular, based on the acquired measurement samples P i and the transmission probability π associated with each measuring device belonging to the sampling subset i To estimate the total power P(T) consumed within the sampling period T.
[0054] In particular, it is shown that a good estimate of the total power P(T) consumed during a sampling period T can be obtained by the following relationship:
[0055]
[0056] It only considers N' measurement devices in the sampling subset.
[0057] The above relationship is of course based on the assumption that all the measured values sent by the N' devices are actually received. Otherwise, the relationship applies to the number of measured values actually collected.
[0058] Experience has shown that in the case of values to be estimated over time (time series), methods that dynamically generate sampling subsets in each sampling period T are much more accurate than "static" or "near-static" methods. This accuracy can be explained by the fact that selection biases resemble fluctuations due to the dynamic formation of sampling subsets and the mean value of the errors is then close to zero.
[0059] Figure 2 Summarizing the invention, it is possible to use, for example, N distributed measuring devices 2 i of Figure 1 The steps of the method for estimating a physical quantity by a system are implemented, and the physical quantity is the total power consumed within a sampling period T:
[0060] The estimation method comprises a first step 100 of measuring the N measuring devices 2 forming a starting set. i Each of the estimates represents local data of a physical quantity.
[0061] In step 110, the N measuring devices 2 of the set iEach of the two executes its local decision algorithm in order to randomly decide whether it sends the measurement sample P within the sampling period T in question. i This step 110 enables dynamic determination of the number of N' measuring devices 2 i The sampling subset formed, each N' measurement device 2 i When completing the execution of its local decision algorithm, it has decided to send the probability π i To send its measurement sample P i As mentioned above, the measurement device in question has a transmission probability π i In the case of multiple possible values of , this step also includes selecting (not shown) the transmission probability value that will actually be associated with the measurement.
[0062] Once the sampling subset has been determined for the sampling period T in question, the method continues with step 120 of acquiring, by means of the centralized platform 1, only the measurement samples P transmitted by the measuring devices 2i of the sampling subset. i As seen above, the transmission probability can be related to the corresponding measurement sample P i sent together, or in some cases must also be sent together with the corresponding measurement sample P i Send together.
[0063] The final step 130 comprises a step based on the maximum N' measurement samples P collected by the centralized platform 1 in step 120. i and the associated sending probability π i To estimate the physical quantity P(T) within the sampling period T, preferably by applying the relationship:
[0064]
[0065] The various steps are repeated in each sampling period so that a different subset of samples is obtained each time.
[0066] This can in particular affect all measuring devices and maintain representativeness over time.
[0067] The sampling period T may be of the order of 10 minutes. However, the use of dynamically created sampling subsets allows this sampling period to be significantly reduced. In particular, it has been shown that it is possible to accurately estimate the aggregated load profile of a large-scale distributed system comprising approximately one hundred thousand sites with a sampling period T of the order of about 10 seconds, or even one second, without incurring significant costs in terms of data communications.
[0068] The estimation method according to the invention makes it possible to significantly reduce the number of sites where measurement values must be sent (from 80% with traditional sampling methods with static subsets to around 10% according to the invention), while preserving the accuracy standards required for distributed solutions participating in the system service.
[0069] Compared to traditional sampling methods, in this case, there is no need to perform a preliminary analysis of the fleet of measuring devices in order to select a representative subset. There is also no need for prior knowledge of the fleet (location, pricing type, venue type, etc.), which is not always accessible and time-consuming and expensive to acquire. Moreover, the configuration of the fleet changes (e.g., home, work, etc.), and keeping such knowledge up to date requires constant allocation of resources.
[0070] Furthermore, since the sampling subset changes with each sampling period, it is not necessary to monitor the measurement devices selected from the sampling subset for potential failures.
[0071] Each measuring device is part of a sampling subset at one time or another, which improves the representativeness and, therefore, the quality of the estimate.
[0072] In the case of a measuring device with a variable transmission probability, the useful information is also increased for the same amount of transmitted data.
Claims
1. A method for estimating a time-varying physical quantity P(T) by means of a distributed measurement system, the distributed measurement system comprising a centralized platform (1) and a plurality of distributed measurement devices (2) each capable of remotely communicating with the centralized platform (1) via a communication network (3). i ), the method comprising: Measuring step (100), by means of each measuring device (2 i ) measuring local data representing the physical quantity; collecting step (120), collecting measurement samples from a subset of the set within a predetermined sampling period T, wherein each measurement device belonging to the subset sends the measurement sample P of the local data to the centralized platform (1) via the communication network (3) i and an estimation step (130), based on the collected measurement samples P i The physical quantity P(T) within the sampling period is estimated by summarizing the physical quantity P(T) within the sampling period, wherein the method comprises in each sampling period T: a stage (110), in which each measuring device (2 i ) executes a local decision algorithm at the location where the local decision algorithm generates a random number and if the random number is less than the number assigned to each measuring device (2 i )The sending probability π associated with i The corresponding measuring device then forms part of the subset and is further characterized in that, according to the calculation of the following relationship, based on the collected measurement samples P i and the transmission probability π associated with each measuring device belonging to the subset i The physical quantity P(T) is estimated in each sampling period T: Wherein, N' is the number of measuring devices in the subset.
2. The method according to claim 1, characterized in that with at least one measuring device (2 i )The sending probability π associated with i The value is based on the obtained measurement sample P i The value and / or the conditions under which the measurement is performed and / or the measurement time are variable, and is characterized in that if the at least one measuring device (2 i ) is part of the subset, then the at least one measuring device in the acquisition step (120) compares the transmission probability with the measurement sample P of the local data via the communication network (3) i At the same time, it is sent to the centralized platform (1).
3. The method according to claim 1, characterized in that In the acquisition step (120), via the communication network (3), each measuring device (2 i ) The sending probability π i The measured sample P of the local data i At the same time, it is sent to the centralized platform (1).
4. The method according to claim 1, characterized in that: The method further comprises: updating the local decision algorithm and / or communicating with at least one measuring device in the set (2 i ) is associated with the sending probability π i steps.
5. The method according to claim 1, characterized in that The sampling period T is less than or equal to 10 minutes.
6. The method according to claim 5, characterized in that The sampling period T is in the order of seconds or in the order of ten seconds.
7. A distributed measurement system for estimating a time-varying physical quantity P(T), the system comprising a centralized platform (1) and a plurality of distributed measurement devices (2) each of which is capable of remotely communicating with the centralized platform (1) via a communication network (3). i ), the system is configured such that in the measuring step (100), each measuring device (2 i ) measures local data representing the physical quantity, and in a collection step (120), each measuring device belonging to a subset of the set sends a measurement sample P of the local data to the centralized platform (1) via the communication network (3) within a predetermined sampling period T. i , and in the estimation step (130), the centralized platform (1) estimates the physical quantity P(T) within the sampling period based on the summary of the collected measurement samples, characterized in that, Each measuring device in the set (2 i ) includes software modules (21 i ), which is configured to execute a local decision algorithm in each sampling period, the local decision algorithm generates a random number, and if the random number is less than the number assigned to each measuring device (2 i )The sending probability π associated with i , the corresponding measuring device then forms part of the subset, and is characterized in that the centralized platform (1) is configured to calculate based on the collected measurement samples P according to the following relationship i and the transmission probability π associated with each measuring device belonging to the subset i The physical quantity P(T) is estimated in each sampling period T: Wherein, N' is the number of measuring devices in the subset.
8. The system according to claim 7, wherein: The local data is the power drawn by a local load comprising one or more household appliances connected to the power distribution network, and the physical quantity is the total power consumed.
Citation Information
Patent Citations
Method and system of real time management and modulation of electricity consumption
WO2008017754A2
Method for assisting in the installation of a home automation system, and related installation driver package
WO2012172242A1
Method and apparatus for finding candidate from data measured by sensor
CN103201621A
Measurement and modulation in real time of the electrical consumption of a plurality of electrical appliances
CN103828168A
