System and method for load detection

Abstract

System (1) for detecting loads (2) of a plant with a plant network (3) which is connected via a transfer point (4) to a single-phase or multi-phase power supply network, so that loads (2) of the plant are supplied with electrical power, with - at least one measuring unit (7) that monitors the current and voltage profile (I, U) of each phase (L) and transmits a power profile determined therefrom to a load detection unit (8) of the system, wherein the load detection unit (8) calculates a power requirement of the loads (2) based on the power profile produced for each phase (L) by at least one inverter (9) and the power profile determined by the measuring unit (7) for each phase (L), and thus automatically detects the loads (2) present within the system, and wherein the measuring unit (7) has a calculation unit which determines an active power curve (P) and a reactive power curve (Q) for each phase (L) from the monitored current and voltage curves, which are transmitted via an interface to the load detection unit (8), wherein the load detection unit (8) performs pattern recognition for each phase (L) based on the power profile produced by the inverter (9) and the power profile determined by the measuring unit (7) to identify the loads present individually for each phase (L), and where the pattern recognition is based on the magnitude of load jumps during switching-on or switching-off events of the calculated power demand of the loads (2).

Description

[0001] The invention relates to a system and a method for detecting loads of a plant with a plant network that is connected to a single- or multi-phase power supply network via a transfer point.

[0002] US 2015 / 0316592A1 describes a system for identifying electrical devices. This requires specialized equipment in which a bypass circuit manipulates the current to generate an identification signal. The change in current can be detected by a power sensor to extract the identification signal.

[0003] DE 10 2016 101 021 A1 relates to a method for identifying consumers in a supply network. This involves collecting and analyzing data from various sensors containing information about total consumption. The analysis is performed using a variety of algorithms that detect the behavior and power consumption of a device and calculate the system state. This information is evaluated using artificial intelligence to confirm or reject the detection of individual consumers.

[0004] The electrical networks of an industrial plant or a private household can contain a variety of different loads that consume energy during operation. These various loads can be supplied with power via one or more phases. While some electrical consumers are supplied with a single phase, others, such as three-phase motors, are typically powered by three phases. Increasingly, these networks also incorporate local power generation units that produce energy locally, which can then be fed into the plant's electrical network. Direct current (DC) sources are connected to the network via inverters.The inverter converts direct current (DC) from a DC power source in the system into alternating current (AC), which is then fed into the system's grid.

[0005] In many applications, it is helpful if the plant operator or user can identify which loads are present in the plant and are being operated there.

[0006] It is therefore an object of the present invention to create a system and a method for the reliable detection of loads within a plant.

[0007] This problem is solved according to the invention by a system and a method for detecting loads of a plant with the features specified in the independent claims.

[0008] The invention thus provides a system for detecting loads in a plant with a plant network that is connected at a transfer point to a single- or multi-phase power supply network, so that loads of the plant are supplied with electrical power, wherein the system comprises: at least one measuring unit that monitors a current and voltage profile of each phase and transmits a power profile determined therefrom to a load detection unit of the plant, wherein the load detection unit calculates a power requirement of the loads based on the power profile produced for each phase (L) by at least one inverter and the power profile determined by the measuring unit for each phase and thus automatically detects the loads present within the plant, and wherein the measuring unit comprises a calculation unit,which determines an active power curve and a reactive power curve for each phase from the monitored current and voltage curves, which are transmitted via an interface to the load detection unit, wherein the load detection unit performs pattern recognition for each phase based on the power curve produced by the inverter and the power curve determined by the measuring unit to identify the individual loads present in each phase, and wherein the pattern recognition is based on the magnitude of load jumps during switching-on or switching-off events of the calculated power demand of the loads.

[0009] In another possible embodiment of the system according to the invention, the load detection unit has an artificial intelligence module, in particular an artificial neural network or an implemented hidden Markov model, which is used for pattern recognition.

[0010] In another possible embodiment of the system according to the invention, the artificial neural network or the hidden Markov model implemented in the load detection unit is trained with training data.

[0011] In another possible embodiment of the system according to the invention, the artificial neural network implemented in the load detection unit comprises a deep neural network that has been trained by supervised machine learning, unsupervised machine learning and / or reinforcement machine learning.

[0012] In another possible embodiment of the system according to the invention, the load detection unit is integrated into the inverter of the system.

[0013] In another possible embodiment of the system according to the invention, the load detection unit is provided in a server which is connected to the inverter of the system via a data network.

[0014] In another possible embodiment of the system according to the invention, the inverter of the system converts a direct current originating from a direct current source of the system into an alternating current, which is fed into the system network by the inverter.

[0015] In another possible embodiment of the system according to the invention, the direct current source of the system comprises a photovoltaic unit, a wind power unit, a fuel cell or a battery.

[0016] In another possible embodiment of the system according to the invention, the load detection unit reports the detected loads to a user via a user interface and / or to a controller of the system via a data interface.

[0017] In another possible embodiment of the system according to the invention, the measuring unit has a scanning device which samples the current and / or voltage waveform for each phase at a predetermined sampling rate and temporarily stores the resulting current and voltage sampling values ​​in a data memory of the measuring unit.

[0018] In another possible embodiment of the system according to the invention, the measuring unit has a calculation device which determines current power values ​​based on the current and voltage sample values ​​temporarily stored in the data storage and transmits them to the inverter via an interface, provided with timestamps.

[0019] Furthermore, the measuring unit has a calculation unit that determines an active power curve and a reactive power curve for each phase from the monitored current and voltage curves, which are transmitted to the load detection unit via an interface.

[0020] In another possible embodiment of the system according to the invention, the load detection unit calculates an active power difference curve and a reactive power difference curve for each phase based on the active power curve and the reactive power curve transmitted by the measuring unit, as well as the active power curve and the reactive power curve fed into the plant network per phase by the inverter, which are evaluated by the load detection unit for pattern recognition.

[0021] The invention further provides a method for automatically detecting loads within a plant with a plant network that is connected at a transfer point to a single- or multi-phase power supply network that supplies loads of the plant with electrical power, wherein the method comprises the following steps: Measuring the current and voltage profile for each phase of the plant's network; Determining a power profile for each phase of the plant network based on the measured current and voltage profile of the respective phase, whereby an active power profile and a reactive power profile are determined for each phase from the monitored current and voltage profile; Transmitting the power profile determined for each phase to a load detection unit of the plant, whereby the determined active power profiles and reactive power profiles are transmitted to the load detection unit via an interface; Determining a power demand profile of the loads based on the power profile determined for each phase and a power profile produced for each phase; and recognizing the loads present within the plant based on the determined power demand profile of the loads,) wherein, for each phase, pattern recognition is performed based on the produced power profile and the determined power profile to identify the loads present individually for each phase, and wherein the pattern recognition is based on the magnitude of load jumps during switching-on or switching-off events of the calculated power demand of the loads.

[0022] Possible embodiments of the load detection system and the method according to the invention will be explained in more detail below with reference to the attached figures. Fig. Figure 1 shows a schematic representation of a possible embodiment of the load detection system according to the invention; Fig. Figure 2 shows a flowchart of a possible embodiment of the inventive method for the automatic detection of loads within a plant; Fig. Figure 3 shows a block diagram of another possible embodiment of the load detection system according to the invention; Fig. Figure 4 shows a block diagram of another possible embodiment of the load detection system according to the invention; Fig. Figure 5 shows a block diagram of another possible embodiment of the load detection system according to the invention; Fig. Figure 6 shows a block diagram of another possible embodiment of the load detection system according to the invention; Fig. 7, Fig. 8 schematic signal diagrams to illustrate pattern recognition performed in the load detection system according to the invention; Fig. 9, Fig. 10 further schematic signal diagrams to illustrate pattern recognition performed in the load detection system according to the invention; Fig. 11, Fig. 12 further signal diagrams to explain the functioning of the load detection system according to the invention.

[0023] Fig. Figure 1 shows a possible embodiment of a system 1 according to the invention for detecting loads within a system. This system can be an industrial plant or a private household system. The schematic representation according to Fig. 1. The various loads are aggregated in a block with the reference symbol 2. The loads of the plant are connected via a local plant network 3, which may include one or more power supply lines, to a transfer point 4 with a single-phase or multi-phase power supply network 5, as shown in Fig. 1 shown. In the Fig. In the embodiment shown in Figure 1, the power supply network 5 is a multi-phase power supply network SVN, which comprises three phases L1, L2, and L3. The loads 2 of the system are supplied with electrical power or electrical energy by the local system network 3. The loads 2 can be, for example, household appliances of a private household, such as washing machines, water pumps, vacuum cleaners, or clothes dryers, and the like. The loads 2 can also be single-phase or multi-phase motors or machines. The various loads 2 of the system are supplied with current I via the system network 3. Some electrical consumers or loads 2 can be connected to several power supply lines simultaneously in order to be supplied with power or current via several phases L1, L2, and L3 of the power supply network 5. In the embodiment shown in Figure 1, the loads 2 are connected to several power supply lines simultaneously in order to be supplied with power or current via several phases L1, L2, and L3 of the power supply network 5. Fig. In the example shown in Figure 1, the local plant network 3 of the plant has three local power supply lines 6-1, 6-2, 6-3. While some loads 2 are connected to all three power supply lines 6-1, 6-2, 6-3 to obtain three power phases L1, L2, L3, other power consumers or loads 2 may only be connected to one power line to obtain, for example, only one power supply phase L. Consequently, for each of the three in Figure 1, the following applies: Fig. The power supply lines 6-i of the plant network 3 shown in Figure 1 have their own superimposed power supply profile. This power supply profile depends on the number and types of loads 2 connected to the respective power supply line 6-i, as well as on the power sources connected to the respective power supply line.

[0024] At the in Fig. In the embodiment shown in Figure 1, the load detection system 1 has a measuring unit 7 which is connected on the input side to the power supply lines 6-i of the system network 3. The measuring unit 7 monitors both a current waveform I and a voltage waveform U for each power supply line 6-i or for each power supply phase L1, L2, L3 and determines an electrical power waveform for each phase L from this. The voltage waveform U can be measured with respect to a reference potential of a neutral conductor. Preferably, the measuring unit 7 determines both an active power waveform P and a reactive power waveform Q from the monitored current and voltage waveforms I, U of each phase L. The active power P and the reactive power Q determined by the measuring unit 7 are used in the Fig. In the embodiment shown in Figure 1, the load detection unit 8 of the load detection system 1 is transmitted to a data interface. Preferably, an active power value P and a reactive power value Q are transmitted for each power supply phase L1, L2, L3. In the embodiment shown in Figure 1, the active power value P and the reactive power value Q are transmitted. Fig. In the three-phase system 3 shown in Figure 1, the load detection unit 8 receives three active power values ​​P1, P2, P3 and three reactive power values ​​Q1, Q2, Q3 from the measuring unit 7 via the data interface. In one possible embodiment, a separate data channel is provided for each active and reactive power value. In another possible embodiment, the data transmission of the active and reactive power values ​​from the measuring unit 7 to the load detection unit 8 can be carried out according to a predefined data protocol.

[0025] In one possible embodiment, the active and reactive power values ​​are transmitted serially from the measuring unit 7 to the load detection unit 8 via a data line of a serial data interface. In an alternative embodiment, the active and reactive power values ​​can be transmitted directly in parallel from the measuring unit 7 to the load detection unit 8 via multiple data lines of a parallel data interface. In a further embodiment, the active and reactive power values ​​P, Q can also be transmitted as a payload in a data packet from the measuring unit 7 to the load detection unit 8. In the embodiment described in Fig. In the embodiment shown in Figure 1, the load detection unit 8 also receives active and reactive power values ​​P, Q from at least one inverter 9 of the system. In the illustrated embodiment, the inverter 9 is connected to a DC power source 10 and converts a DC current from the DC power source 10 into an AC current, which is fed into the local system grid 3 by the inverter 9. In the embodiment shown in Fig. In the embodiment shown in Figure 1, the inverter 9 supplies an alternating current (AC) for each phase L1, L2, L3 of the system network 3. The inverter 9 produces electrical power for each phase L1, L2, L3 and reports this to the load detection unit 8 of the load detection system 1 via a further data interface. Based on the electrical power or power curve produced by the inverter 9 for each phase L1, L2, L3 and the power or power curve determined by the measuring unit 7 for each phase L1, L2, L3, the load detection unit 8 calculates a power demand curve of the loads 2 and uses this to identify the loads 2 present within the system. Preferably, the load detection unit 8 performs pattern recognition separately for each phase L1, L2, L3 based on the electrical power curve produced by the inverter 9 and the electrical power curve determined by the measuring unit 7.In one possible embodiment, the load detection unit 8 can report or output the detected loads 2 via an interface 11 of the load detection system 1. In another possible embodiment, the load detection unit 8 can also report the detected loads 2 to a user or operator of the system via a user interface and / or to a local controller of the system via a data interface.

[0026] The measuring unit 7 of the load detection system 1 preferably comprises a scanning device which samples the current and voltage profiles I and U over time for each phase L1, L2, L3 at a predetermined sampling rate R and temporarily stores the generated current and voltage samples in a data memory. The voltage U can be measured with respect to a reference potential or neutral conductor.

[0027] Alternatively, voltage differences between the phase conductors (L1-L2, L2-L3, L3-L1) can also be measured by the measuring unit 7. In one possible embodiment, the data storage can be integrated into the measuring unit 7. The measuring unit 7 preferably has a processor or a calculation unit that determines or calculates current power values ​​based on the current and voltage samples temporarily stored in the data storage. The calculated power values ​​P, Q are preferably provided with timestamps TS (Time Stamp). The power values ​​with timestamps TS are preferably transmitted from the measuring unit 7 to the load detection unit 8 via a data interface. In one possible embodiment, the measuring unit 7 has a calculation unit that determines or calculates both an active power curve P and a reactive power curve Q for each phase L1, L2, L3 from the respective monitored current and voltage curves I, U.The load detection unit 8 calculates power values, preferably transmitted via an interface together with a timestamp TS. The load detection unit 8 preferably also has a calculation unit that calculates an active power difference ΔP and a reactive power difference ΔQ for each phase L1, L2, L3 based on the active power P and reactive power Q received by the measuring unit 7 for each phase L1, L2, L3, as well as on the active power P and reactive power Q fed into the grid by the inverter 9 for each phase L. The two power differences determined for each phase L, i.e., the active power difference ΔP and the reactive power difference ΔQ, are evaluated by the load detection unit 8 for the respective pattern recognition.

[0028] The recording of the measured or discharge values ​​by the measuring unit 7 and the recording of the power fed in by the inverter 9 per phase is preferably synchronized.

[0029] In one possible embodiment, phases (φ, cos φ) or phase differences between the current and voltage profile of a phase L and / or between different phases are additionally measured and transmitted to the load detection unit 8.

[0030] In one possible embodiment, the load detection unit 8 includes an artificial intelligence module (AIM) that performs pattern recognition. In one possible embodiment, the AIM is an artificial neural network (NN). In an alternative embodiment, the AIM is an implemented hidden Markov model (HMM). The artificial neural network or the hidden Markov model (HMM) implemented in the load detection unit 8 is preferably trained with training data. For this purpose, the AIM of the load detection unit 8 is preferably trained with training data in a training phase. In one possible embodiment, the training of the AIM implemented within the load detection unit 8 can be performed before delivery of the system to the customer or system operator.In an alternative embodiment, the load detection unit 8 is connected to a server 16 via a data interface and a data network, so that even after the load detection system 1 has been installed in the plant operator's facility, the artificial intelligence module (AIM) implemented in the load detection unit 8 can be trained remotely via the data network. In one possible embodiment, an artificial neural network (NN) in the form of a so-called deep neural network (DNN) is implemented in the load detection unit 8, which has several hidden layers of neural nodes. The training of the artificial neural network (NN) can be performed by supervised machine learning (ML), unsupervised machine learning (ML), or reinforcement machine learning. Supervised machine learning uses training data that has labels.In one possible embodiment, the training data and their labels can be stored in a database of the operator of the load detection system 1.

[0031] At the in Fig. In the embodiment shown in Figure 1, the load detection unit 8 is a self-contained unit that is connected to both the measuring unit 7 and the inverter 9 via a dedicated interface. In an alternative embodiment, the load detection unit 8 can also be integrated into the inverter 9 of the system. Furthermore, the load detection unit 8 can also be provided in a server 16, which is connected to the inverter 9 of the system via a data network.

[0032] A power meter for the system can be provided at the transfer point 4 of the power supply network 5.

[0033] Fig. Figure 2 shows a flowchart of a possible embodiment of a method according to the invention for the automatic detection of loads 2 within a plant. The plant has a plant network 3 which is connected via a transfer point 4 to a single-phase or multi-phase power supply network S.

[0034] In a first step S1, a current and voltage waveform I, U is measured for each phase L of the plant network 3. This is done, for example, with a measuring unit 7, which has a sampling device. The sampling device of the measuring unit 7 samples the current and voltage waveform I, U for each phase L at a predefined sampling rate R, whereby the generated current and voltage sampling values ​​can be temporarily stored in a data memory.

[0035] In a further step S2, an electrical power profile for each phase L of the plant network 3 is determined or calculated based on the measured current and voltage profiles I, U of the respective phase L. In one possible embodiment, current power values ​​are calculated based on the current and voltage samples temporarily stored in the data memory and preferably provided with corresponding timestamps TS. In another possible embodiment, for each phase L, both an active power profile P and a reactive power profile Q are determined from the monitored current and voltage profile, i.e., based on the current and voltage samples temporarily stored in the data memory.

[0036] In a further step S3, the power values ​​determined for each phase L are transmitted to a load detection unit 8 of the system. In one possible embodiment, this load detection unit 8 is integrated into an inverter 9 of the system. The transmission of the power determined for each phase L preferably takes place via a data interface.

[0037] In a further step S4, the power requirement of all loads 2 of the system is calculated based on the power profile determined or calculated for each phase L and the total power produced for each phase L. In one possible embodiment, the produced power is generated by an inverter 9 of the system.

[0038] In a further step S5, the loads 2 present within the system are automatically detected based on the determined power demand profile of the loads 2. In one possible embodiment, pattern recognition can be performed for each phase L based on the power produced by the inverter 9 per phase L and the power determined by a measuring unit 7 per phase. Based on the active power profile P and the reactive power profile Q measured for each phase L by a measuring unit 7, as well as the active power P and reactive power Q fed into the system grid 3 per phase L by the inverter 9 per phase L, an active power differential profile ΔP and a reactive power differential profile ΔQ can be calculated for each phase in one possible embodiment, both of which are evaluated for pattern recognition.In one possible embodiment, pattern recognition is preferably performed by an artificial intelligence module (KIM) that has been pre-trained with appropriate training data.

[0039] Fig. Figure 3 shows a block diagram of another possible embodiment of the load detection system 1 according to the invention. In the Fig. In the embodiment shown in Figure 3, the load detection unit 8 is connected to a data network 12 and communicates via an access point 13 with a portable mobile device 14 belonging to user U. User U can be the plant operator. If the plant is, for example, a private household, user U is, for instance, a resident of the household who operates the electricity consumers or loads 2 of the plant. The mobile device 14 can also be, for example, a mobile phone belonging to user U. In the embodiment shown in Figure 3, the load detection unit 8 is connected to a data network 12 and communicates via an access point 13 with a portable mobile device 14 belonging to user U. Fig. In the embodiment shown in Figure 3, the load detection unit 8 can transmit detected loads 2 to the user U's portable mobile device 14 via the data network 12, the access point 13, and a wireless air interface. This allows the loads 2 to be displayed to the user U via a graphical user interface. In one possible embodiment, the load detection unit 8 transmits the detected loads 2 and their current operating states to the user U for output. In the embodiment shown in Figure 3, the load detection unit 8 transmits the detected loads 2 and their current operating states to the user U. Fig. In the embodiment shown in Figure 3, user U can be located remotely from the system and still receive information about the loads 2 contained and operated within the system. For example, user U leaving their house can check whether a washing machine is currently operating in the house's electrical system. User U can, for instance, check whether they switched on the washing machine before leaving the house. If user U wanted to switch off the washing machine before leaving the house and recognizes from the transmitted information that the washing machine is currently operating as load 2, one possible embodiment allows user U to send a corresponding command to the load detection unit 8 via their terminal device 14 and the data network 12, for example, to switch off the washing machine or load 2. In the embodiment shown in Figure 3, the user U can then send a corresponding command to the load detection unit 8 via their terminal device 14 and the data network 12 to, for example, initiate the switching off of the washing machine or load 2. Fig. In the illustrated embodiment 3, the load detection unit 8 is also connected via a control line to a local control unit 15, which switches the relevant load 2 on or off via electronic or electromechanical switches according to the user command received. The successful switching on or off of the load 2 can be reported back to the user U.

[0040] In one possible embodiment, the load detection unit 8 can be integrated into the local control unit 15 of the system. The local control unit 15 of the system can also have a data interface to the inverter 9 and / or the DC power source 10.

[0041] Fig. Figure 4 shows an embodiment of the load detection system 1 according to the invention, in which the load detection unit 8 is integrated into the inverter 9. Based on the electrical power profile produced for each phase L by the inverter 9 and the power profile determined for each phase L by the measuring unit 7 at a transfer point 4, the load detection unit 8 detects a power demand profile of the loads 2, whereby the detected loads 2 can be output or reported via an interface 11. The load detection unit 8 can also control the corresponding detected loads 2 via the local control unit 15 of the system, as shown in Figure 4. Fig. 4 shown schematically.

[0042] Fig. Figure 5 shows a further embodiment of the load detection system 1 according to the invention. In the case of the Fig. In the embodiment shown in Figure 5, the load detection unit 8 is integrated into a server 16, which is connected to the data network 12. The server 16 has access to a central or distributed database 17. The server 16 can be a remote server connected to the building system via a data network 12, in particular the internet. In one possible embodiment, the database 17 can contain training data used to train an artificial neural network implemented in the load detection unit 8. In an alternative embodiment, the training data can also be transmitted via the data network 12 to a load detection unit 8 for training an artificial intelligence module, which is located, for example, in the inverter 9. In the embodiment shown in Figure 5, the load detection unit 8 is integrated into a server 16, which is connected to the data network 12. Fig. In the embodiment shown in Figure 5, the local control unit 15 of the system is directly connected to the data network 12 and can communicate with the load detection unit 8, which is implemented in the server 16. The loads 2 detected by the load detection unit 8 during the operation of the system can be transmitted via the data network 12 to the portable user terminal 14 of user U and output there to user U. The in Fig. 5 of the displayed servers (16) can be part of a cloud platform.

[0043] Fig. Figure 6 shows a further embodiment of the load detection system according to the invention. Communication interfaces or communication lines are shown with dashed lines. In the case described in Fig. In the embodiment shown in Figure 6, system 1 has several measuring units 7A, 7B, 7C. In the embodiment shown in Figure 6, the system 1 has several measuring units 7A, 7B, 7C. Fig. In the embodiment shown in Figure 6, the measuring unit 7A is connected to power supply lines 6-i at or near the transfer point 4. A further measuring unit 7 is provided at a power source 18, which feeds power into the local plant network 3 independently of the inverter 9. The power source 18 can be formed by another inverter that includes the measuring unit 7B. The power source or the additional inverter also feeds an alternating current (AC) into the local plant network 3. The power source 18 can also be a different AC source, for example, a diesel generator or the like. The number of additional power sources or inverters can vary depending on the application. Fig. In the embodiment shown in Figure 6, each additional power source 18 or inverter has an associated measuring unit 7B, as shown in Figure 6. Fig. Figure 6 shows the measuring unit 7B supplying data to the load detection unit 8 via an interface regarding the electrical power fed into the plant network 3 by the respective power source 18 for each current phase L. The communication interface can be formed by a MODBUS. Furthermore, the plant network 3 has the following features: Fig. Figure 6 illustrates an embodiment of consumption units or loads 19 that can be used to consume excess power or energy and can be controlled, for example, by a local control unit 15 of the system. In one possible embodiment, the loads or consumers 19 include a heating element control unit that uses excess photovoltaic power for hot water preparation. In one possible embodiment, the heating element control unit 19 can be used to maximize the self-consumption of locally generated energy. The heating element control unit 19 is preferably also connected to the MODBUS. The heating element control unit 19 contains heating elements as electrical consumers or loads that convert excess power into heat energy, for example, to heat water. The consumers of the heating element control unit 19 thus draw electrical power or energy from the system network 3 as needed.Energy to adjust the power per phase L such that neither electrical power is fed into the power supply network 5 nor is drawn from the power supply network 5 via the transfer point 4. In the case of... Fig. In the embodiment shown in Figure 6, the heating element control 19 has an associated measuring unit 7C, which supplies data to the load detection unit 8 via a data interface. In the embodiment shown in Figure 6, the heating element control 19 has an associated measuring unit 7C, which supplies data to the load detection unit 8 via a data interface. Fig. In the embodiment shown in Figure 6, the measuring unit 7A can record the total aggregated electrical energy consumption or energy consumption profile of the system and transmit the corresponding data to the load detection unit 8. The load detection unit 8 processes the data regarding the aggregated electrical power demand and subtracts from it the power fed in by the inverter 9 per phase L. If another power source is connected to the inverter 9, for example, the one in Fig. If the power source 18 shown in Figure 6 is present, the power data supplied to the load detection unit 8 is transmitted by the associated measuring unit 7B to the communication interface (e.g. MODBUS), as shown in Figure 6. Fig. Figure 6 shows this. Power consumers also use excess power, for example, those in Fig. The heating element control 19 shown in Figure 6, with heating elements for consuming electrical energy, each has an associated measuring unit 7C which transmits corresponding power consumption data to the load detection unit 8. Both active power data P and reactive power data Q can be transmitted for each phase L, so that the load detection unit 8 can perform automatic load pattern recognition for each phase L of the system. With the load detection system 1 according to the invention, it is possible to identify individual consumers or loads in a section to be supplied, for example within a household, from the recorded current and voltage profiles.The load detection unit 8 of the load detection system 1 is able to determine power flows per phase L and to derive specific power patterns from the existing power flows, which in turn provide information about which loads 2 in the plant have been operated or activated at different times or within different periods.

[0044] The measuring unit 7 preferably has a scanning device that samples the current and voltage waveforms for each phase L at a predefined sampling rate R. This sampling rate can vary depending on the application and is adjustable in one possible embodiment. In one embodiment, a relatively low sampling rate R of approximately 1 to 5 Hz is used to create corresponding load profiles. The relatively low sampling rate R minimizes the technical effort required to acquire raw data. Furthermore, the relatively low sampling rate R minimizes the storage space required for the sampled values. This allows the sampled data storage to be integrated into the measuring unit 7.On the other hand, the relatively low sampling rate R in this embodiment also results in a loss of information; for example, it cannot be guaranteed that current peaks during the switching on or off of loads 2 are recorded as samples. Therefore, in one possible embodiment, time windows containing a switching-on or switching-off event of a load 2 are located or determined. For pattern recognition, training data recorded for this time window and preferably marked or labeled accordingly can be used in one possible embodiment.

[0045] Fig. Figure 7 shows a power demand calculated by the load detection unit 8 over time t, for example an active power or reactive power demand curve for one of the current phases L.

[0046] Fig. Figure 8 shows three typical power curves of different known loads 2A, 2B, 2C. Using a pattern matching algorithm, the deviation between the values ​​shown in Figure 8 is calculated. Fig. 7 shown measured load profile and a combination of the in Fig. The 8 known load profiles shown are minimized. Various known load profiles or load distributions of different load types 2A, 2B, 2C, as shown in Fig. The values ​​shown in section 8 can be combined with each other and each with the actually calculated power requirement, as shown in Fig. Figure 7 is shown and can be compared. In one possible embodiment, the combination of known load profiles that results in the smallest deviation can be output as the detected combination of different loads 2 within the system. In this embodiment, a determined power demand profile is compared with combinations of known load profiles to identify the most likely combination of loads 2 with the smallest deviation. A linear combination of the load profiles of load types 2A, 2B, and 2C results in the measured load profile according to Figure 7. Fig. 7. The load distribution according to Fig. 7 can automatically be converted into a linear combination of known load profiles, such as those found in Fig. The 8 parts shown are disassembled.

[0047] In an alternative embodiment, pattern recognition takes place, as in the Fig. 9, Fig. Figure 10 illustrates this. In one possible embodiment, switching-on events can be aligned with corresponding switching-off events to detect loads 2. A switching-on or switching-off event, in which a load or consumer within the system network 3 is switched on or off, leads to load jumps. The magnitude of the load jump during switching on or off is typical or characteristic of the respective load 2. For example, switching on a washing machine results in a larger load jump than switching on a television within a household. A positive load jump when switching on a load or consumer corresponds to its magnitude or amplitude in the negative load jump when switching off the same load. As can be seen in Fig. As can be seen in Figure 9, the load profile or power flow shown begins with the first load 2-1 being switched on at time t1, leading to an increased power demand. At time t2, another load 2-2 is switched on, causing a slightly smaller positive load step. At time t3, a negative load step occurs, the magnitude of which corresponds to the positive load step at time t1. At time t4, a large positive load step occurs, i.e., a load or consumer 2-3 is switched on, significantly increasing the power demand. At time t5, a negative load step occurs, the magnitude of which corresponds to the positive load step at time t2. Finally, at time t6, a load is switched off, causing a negative load step, the magnitude of which corresponds to the positive load step at time t4. From the load profile according to Fig. 9, Fig. From equation 10, it can be deduced that a first load 2-1 was switched on at time t1 and switched off at time t3. Furthermore, the load profile shows that another load 2-2 was switched on at time t2 and switched off again at time t5. This second load 2-2 consumes significantly less power than the first load, which was switched on at time t1 and off at time t3. Finally, the load profile shows that a load 2-3 with high power consumption was switched on at time t4 and switched off again at time t6. Each load 2-i thus makes a typical contribution to the overall load profile for each phase L, so that by analyzing or decomposing the overall load profile based on characteristic features, such as load jumps, a combination of active loads 2-i at a specific time or period can be automatically identified.

[0048] The procedure according to Fig. 7, Fig. 8 can be done using the procedure according to Fig. 9, Fig. 10 can be combined to increase the accuracy of load detection.

[0049] A large number of measured load profiles can be recorded as training data and, in one possible implementation, used to train an artificial intelligence module. Using machine learning (ML), particularly by training a neural network (NN), automatic pattern recognition can be achieved. The accuracy of load detection can be further improved by performing pattern recognition separately for each power supply phase L1, L2, and L3, and by considering both an active power profile P(t) and a reactive power profile Q(t) for each phase L1, L2, and L3. This allows even loads with relatively low power requirements that result in relatively small load fluctuations to be clearly identified.The determination of the power requirements of loads 2-i in a system over a specified period can be carried out either by a local load detection unit 8 or by a load detection unit 8 implemented in a server. A user U or system operator thus has the ability to obtain information regarding which loads 2-i are activated, switched on, or switched off at a specific time or within a specific period. In one possible embodiment, the user U or system operator receives information regarding the detected loads 2-i in near real time. This allows the user U or system operator to intervene in the ongoing operation of the system, in particular to switch loads 2 on or off, or to activate additional power sources, especially inverters 9, if this is necessary to supply the loads 2.Decisions regarding this can also be made by a suitably trained neural network (NN). For example, the neural network (NN) can trigger or initiate specific measures based on a known combination of currently switched-on or switched-off loads (2-i) to avoid or prevent critical system states of the plant. For example, after detecting a specific load or power profile on one or more power supply lines (6-i) of the plant network (3), the neural network (NN) can proactively switch on additional power sources, such as a diesel generator, to reliably meet an expected increase in power demand within the plant.

[0050] Fig. Figure 11 shows an example of a load profile P / Q of a power supply phase at a transfer point PCC (Point of Common Coupling). In the Fig. In the load profile shown in 11, power dips LE are noticeable around noon, which may either originate from a switching of a load 2 of the system or be caused by, for example, clouds covering the sky, so that the power fed in by an inverter 9 is reduced.

[0051] Fig. Figure 12 shows the power curve of the inverter 9 over time. A drop in the power fed in by inverter 9, for example due to a cloud passing overhead, can be seen at a specific time t. From the in Fig. 11 measured power profile and the power profile of the inverter 9 according to Fig.12. A load profile of the loads 2 is generated. The load profile, or power demand profile, of the loads is calculated or determined by adding the load profile of the inverter 9 with the measured power profile at the transfer point. Based on the determined power demand or consumption, it can then be automatically identified which loads 2 were activated at which time. External influencing factors, such as passing clouds, are thus removed from the profile of the adjusted or net power demand of the loads 2, in order to prevent distortions of the load profile. This adjusted power profile shows typical load jumps of different loads, which allow the loads 2 present in the system to be automatically identified, or rather, which combination of loads 2 are switched on or off at a specific time or during a specific period.In the load detection system 1 according to the invention, external potentially distorting influences on the load profile or power consumption can be factored out in order to uniquely identify loads or load combinations based on a corrected power profile. This can be carried out separately for different power supply phases L1, L2, L3, thereby enabling even greater clarity in the identification of the various loads. The accuracy in the detection of the loads L is preferably further increased by taking into account both active power differences ΔP and reactive power differences ΔQ. The load detection system 1 according to the invention can thus detect with high accuracy, even with a relatively large number of different loads 2 within a system and even with loads 2 exhibiting relatively small load changes, which loads 2 are actively or passively switched at what time or during what period (e.g.,are in standby mode.

[0052] Further embodiments of the load detection system 1 according to the invention are possible. In one possible embodiment, the sampling rate R of the current and voltage waveforms, which are sampled by a scanning device of the measuring unit 7, is relatively low during normal operation (e.g., 1–5 Hz) in order to minimize the technical complexity and the required memory. In another possible embodiment, the sampling rate is flexibly adjustable and can be increased for specific periods (e.g., to 10 Hz or 100 Hz) to achieve higher load detection accuracy. Furthermore, if there is an increased density of load transitions on one or more power supply lines 6-i during a specific period, the sampling rate R can be dynamically increased temporarily in another possible embodiment to improve the precision of load detection in this complex load profile segment.

[0053] In another embodiment, the load detection unit 8 can perform not only a time-domain analysis but also a frequency-domain (spectrum) analysis of the load profiles. This is particularly possible when the load detection unit 8 is located on a server 16 that has the necessary computing resources.

[0054] In one possible embodiment, various loads 2 of the system can be selectively switched on and off by the local control 15 to generate training data. The training data generated in this way is stored, for example, in a database 17 and used to train a neural network (NN) of the load detection unit 8. In this embodiment, when implementing a system with a large number of different loads 2, various combinations of loads can be selectively switched on or off to record the resulting measurement data as training data. In this embodiment, the training data generated specifically for a system can be used to train the artificial intelligence module (AIM) implemented in the load detection unit 8, enabling automatic load pattern recognition during subsequent normal system operation.The artificial intelligence model KIM, or the neural network NN of the load detection unit 8, can be further trained during normal operation to further increase the precision of load detection. Furthermore, reference systems with standard load configurations can be used to generate training data, which is then used to train the artificial intelligence models KIM of systems implemented at a customer's site.

Claims

[1] System (1) for detecting loads (2) of a plant with a plant network (3) which is connected via a transfer point (4) to a single-phase or multi-phase power supply network, so that loads (2) of the plant are supplied with electrical power, with - at least one measuring unit (7) that monitors the current and voltage profile (I, U) of each phase (L) and transmits a power profile determined therefrom to a load detection unit (8) of the system, wherein the load detection unit (8) calculates a power requirement of the loads (2) based on the power profile produced for each phase (L) by at least one inverter (9) and the power profile determined by the measuring unit (7) for each phase (L), and thus automatically detects the loads (2) present within the system, and wherein the measuring unit (7) has a calculation unit which determines an active power curve (P) and a reactive power curve (Q) for each phase (L) from the monitored current and voltage curves, which are transmitted via an interface to the load detection unit (8), wherein the load detection unit (8) performs pattern recognition for each phase (L) based on the power profile produced by the inverter (9) and the power profile determined by the measuring unit (7) to identify the loads present individually for each phase (L), and where the pattern recognition is based on the magnitude of load jumps during switching-on or switching-off events of the calculated power demand of the loads (2). [2] System according to claim 1, wherein the load detection unit (8) comprises an artificial intelligence module, in particular an artificial neural network, ANN, or a hidden Markov model, HMM, which is used for pattern recognition. [3] System according to claim 2, wherein the artificial neural network, ANN, or the hidden Markov model, HMM, implemented in the load detection unit (8) is trained with training data. [4] System according to claim 3, wherein the artificial neural network, ANN, implemented in the load detection unit (8) comprises a deep neural network, DNN, trained by supervised machine learning, unsupervised machine learning and / or reinforcement machine learning. [5] System according to any one of the preceding claims 1 to 4, wherein the load detection unit (8) is integrated into the inverter (9) of the system. [6] System according to any one of the preceding claims 1 to 4, wherein the load detection unit (8) is provided in a server (16) which is connected to the inverter (9) of the system via a data network (12). [7] System according to any one of the preceding claims 1 to 6, wherein the inverter (9) of the system converts a direct current (DC) originating from a direct current source (10) of the system into an alternating current (AC) which is fed into the system network (3) by the inverter (9). [8] System according to claim 7, wherein the DC power source (10) of the system comprises a photovoltaic unit, a wind power unit, a fuel cell or a battery. [9] System according to any one of the preceding claims 1 to 8, wherein the load detection unit (8) reports the detected loads (2) to a user via a user interface or to a controller (15) of the system via a data interface. [10] System according to any one of the preceding claims 1 to 9, wherein the measuring unit (7) has a scanning device which samples the current and voltage waveforms for each phase (L) at a predetermined sampling rate and temporarily stores the current and voltage sample values ​​generated in a data memory of the measuring unit (7). [11] System according to claim 10, wherein the measuring unit (7) has a calculation device which determines current power values ​​on the basis of the current and voltage sample values ​​temporarily stored in the data storage and transmits them with timestamps via an interface to the load detection unit (8). [12] System according to one of claims 1 to 11, wherein the load detection unit (8) calculates an active power difference curve (ΔP) and a reactive power difference curve (ΔQ) for each phase (L) on the basis of the active power curve (P) and the reactive power curve (Q) transmitted by the measuring unit (7) and the active power curve (PWR) and the reactive power curve (QWR) fed into the plant network (3) per phase (L) by the inverter (9), which are evaluated by the load detection unit (8) for pattern recognition. [13] Method for automatically detecting loads (2) within a plant with a plant network (3) which is connected via a transfer point (4) to a single-phase or multi-phase power supply network (5) which supplies loads (2) of the plant with electrical power, comprising the following steps: (a) Measuring (S1) a current and voltage waveform (I, U) for each phase (L) of the plant network (3) of the plant; (b) Determining (S2) a power profile for each phase (L) of the plant network (3) of the plant based on the measured current and voltage profile of the respective phase (L), whereby for each phase (L) an active power profile (P) and a reactive power profile (Q) are determined from the monitored current and voltage profile; (c) Transmitting (S3) the power profile determined for each phase (L) to a load detection unit (8) of the plant, wherein the determined active power profiles (P) and reactive power profiles (Q) are transmitted to the load detection unit (8) via an interface; (d) Determining (S4) a power demand profile of the loads (2) based on the power profile determined for each phase (L) and a power profile produced for each phase (L); and (e) Identifying (S5) the loads (2) present within the plant on the basis of the determined power demand profile of the loads (2), wherein for each phase (L) a pattern recognition is performed to identify the loads present individually per phase (L) based on the produced power profile and the determined power profile, and wherein the pattern recognition is based on the magnitude of load jumps during switching-on or switching-off events of the calculated power demand of the loads (2).

Images