Facility electric vehicle charging detection

Abstract

The system includes a processor and a non-transitory computer-readable memory including instructions executable by the processor to cause the processor to perform operations. The operations include accessing facility consumption data in an electric distribution network. The facility consumption data includes an indication of facility resource consumption over a period of time. The operations also include applying a machine learning model to the facility consumption data. The machine learning model is trained to generate an output corresponding to an electric vehicle classification for the facility. The operations further include generating an electric vehicle classification for the facility using the output of the machine learning model and controlling generation of the electric distribution network based on the electric vehicle classification for the facility.

Description

[Technical Field]

[0001] The present disclosure relates generally to consumption detection in a resource allocation system. More particularly, but not by way of limitation, the present disclosure relates to detecting the charging of electric vehicles at a resource allocation system facility. [Background technology]

[0002] In a resource allocation system, such as a power grid that provides electricity, facilities may consume resources at different rates over time based on the devices within the facility that consume the resources of the resource allocation system over that time period. For example, electric vehicles charging within the facility may consume more power within the facility at different times of the day. To ensure sufficient power generation to meet the demands of the power grid, it is beneficial to accurately track facilities that regularly charge electric vehicles.

[0003] Demand on the grid may change over time as more electric vehicles require charging at different locations connected to the grid. For example, the number of electric vehicle charging facilities and the different types of chargers used at the facilities to charge electric vehicles may change over time. Accurately tracking electric vehicle charging facilities with minimal input from residents within the facilities can help drive generation management in resource allocation systems. Summary of the Invention

[0004] In one embodiment, a system includes a processor and a non-transitory computer-readable memory including instructions executable by the processor to cause the processor to perform operations. The operations include accessing facility consumption data in an electric distribution network. The facility consumption data includes an indication of facility resource consumption over a period of time. The operations also include applying a machine learning model to the facility consumption data. The machine learning model is trained to generate an output corresponding to an electric vehicle classification for the facility. The operations further include generating an electric vehicle classification for the facility using the output of the machine learning model and controlling generation of the electric distribution network based on the electric vehicle classification for the facility.

[0005] In another embodiment, a non-transitory computer-readable medium includes instructions executable by a processor to cause the processor to perform operations. The operations include accessing facility consumption data in an electric distribution network. The facility consumption data includes an indication of facility resource consumption over a period of time. The operations also include applying a machine learning model to the facility consumption data. The machine learning model is trained to generate an output corresponding to an electric vehicle classification for the facility. The operations further include generating an electric vehicle classification for the facility using the output of the machine learning model and controlling generation of the electric distribution network based on the electric vehicle classification for the facility.

[0006] In another embodiment, a computer-implemented method includes accessing facility consumption data for a facility in an electric distribution network. The facility consumption data includes an indication of facility resource consumption over a period of time. The method also includes applying a machine learning model to the facility consumption data. The machine learning model is trained to generate an output corresponding to an electric vehicle classification for the facility. The method further includes generating an electric vehicle classification for the facility using the output of the machine learning model and controlling generation of the electric distribution network based on the electric vehicle classification for the facility. [Brief explanation of the drawings]

[0007] [Figure 1]FIG. 1 illustrates an example physical topology of an electrical distribution network showing devices at various points or nodes on the network, according to some embodiments described herein. [Figure 2] FIG. 2 is a flowchart of a process for controlling power supply in the distribution network of FIG. 1 based on classification of electric vehicles within the distribution network facility, according to some embodiments described herein. [Figure 3] FIG. 3 is an example diagrammatic representation of data flow for identifying facilities performing electric vehicle charging operations, according to some embodiments described herein. [Figure 4] FIG. 4 is a flowchart of a process for training a machine learning model to identify facilities that perform electric vehicle charging operations, according to some embodiments described herein. [Figure 5] FIG. 5 is an example of a machine learning model environment used to identify facilities that perform electric vehicle charging operations, according to some embodiments described herein. [Figure 6] FIG. 6 is an example of an artificial neural network machine learning model according to some embodiments described herein. [Figure 7] FIG. 7 is an exemplary computing device used to detect charging activity of an electric vehicle according to some embodiments described herein. DETAILED DESCRIPTION OF THE INVENTION

[0008] These and other features, aspects, and advantages of the present disclosure will be better understood by reading the following detailed description in conjunction with the accompanying drawings.

[0009] The present invention describes techniques for providing detection of electric vehicle charging at facilities in a resource distribution system. In one example, a resource distribution system, such as an electric power grid, can provide consumable resources to facilities within the resource distribution system. For example, a facility may consume power at a different rate when electric vehicles are charging at the facility than when electric vehicles are not charging at the facility. As more electric vehicles are charging at facilities within the power grid, the power generation capacity of the power grid can be strained. Furthermore, because electric vehicles may be charging during periods traditionally considered "low demand," such as overnight, increased charging during these low-demand periods can create unexpected strain on the power grid if generators go offline during low-demand periods.

[0010] To maintain accurate information regarding numerous facilities performing electric vehicle charging operations, time-series power consumption data may be obtained from facilities consuming power from the power grid. The data may be processed by a trained machine learning model to identify facilities that are charging electric vehicles. For example, the trained machine learning model may be applied to the data to distinguish electric vehicle charging operations at the facilities from other types of charging operations. In some examples, the machine learning model may be trained to identify the type of charger (e.g., Level 1 or Level 2) performing electric vehicle charging operations at the facilities. Generally, grid power generation and grid planning may be controlled based on information identified by the machine learning model. For example, if additional electric vehicle charging is detected at facilities sharing a common transformer, a utility may proactively replace the common transformer with a higher-capacity version. Other components of the resource distribution network may also be replaced based on the likelihood of a change from general demand at the facilities served by the other components as additional electric vehicles are charged at the facilities served by the other components. In a further example, if additional charging operations at a facility are detected, the power generated may be increased at times when charging operations are most likely to occur. Similarly, the identified peak and low demand periods may change based on further electric vehicle charging operations, and the operation of the power generation equipment may be controlled to reflect the updated peak and low demand periods.

[0011] The illustrative examples are provided to introduce the reader to the general subject matter discussed herein and are not intended to limit the scope of the disclosed concepts. Various additional features and examples are described in the following sections with reference to the drawings, in which like numerals indicate like elements and directional descriptions are used to illustrate exemplary embodiments, but as such should not be used to limit the disclosure.

[0012] FIG. 1 illustrates an exemplary physical topology of an electric distribution network, showing devices at various points, or nodes, on the network. FIG. 1 illustrates a facility monitoring system 100 and an electric distribution network 110. In one example, the facility monitoring system 100 receives information from endpoint meters in the electric distribution network 110 to determine which facilities on the electric distribution network 110 are performing electric vehicle charging operations. While the facility monitoring system 100 and the electric distribution network 110 are described herein as part of an electric distribution environment, similar facility monitoring systems 100 may be incorporated into other utility systems. For example, the facility monitoring system 100 may be used in gas, water, or other utility distribution environments where it is desirable to monitor certain resource consumption operations.

[0013] The facility monitoring system 100 includes a facility monitoring application 101 and a head-end system 102. As shown in FIG. 7, the facility monitoring application 101 executes on a computing device. The facility monitoring application 101 can receive metering data, such as voltage, load, and power consumption, from meters installed at customer facilities. Other data from other data points may also be provided to the facility monitoring application 101. The facility monitoring application 101 receives metering data from the head-end system 102 or through an intermediate device that reads and aggregates the metering information. The meters can communicate metering information to the head-end system through additional network devices and networks, which are not shown in the figure for simplicity. In some examples, communication operations within the power distribution network 110 may be performed using radio frequency (RF) wireless communications, cellular communications, power line communications (PLC) communications, or other suitable communication technologies. In some examples, the facility monitoring application 101 may be incorporated into the grid edge. In such examples, some decisions regarding the power distribution network 101 may be made in a distributed manner (e.g., without going through the head-end system 102).

[0014] The power distribution network 110 includes a substation 112 and one or more feeders 120a-120n. The substation 112 distributes power received from a power source to the feeders 120a-120n. Examples of power sources include coal-fired power plants, wind turbines, and solar panel installations. The substation 112 may include a substation transformer 113. The substation transformer 113 steps down the voltage supplied to the substation 112 and outputs the lower voltage to the feeders 120a-120n. The substation 112 may distribute multi-phase (e.g., three-phase) power. Although FIG. 1 depicts a single substation 112, the power distribution network 110 may include multiple similar substations.

[0015] Each feeder 120a-120n can provide power to one or more facilities, such as facility 124 or facility 125, via one or more meters, such as meters 130 and 131. Facility 124 or 125 may include a residence, apartment, commercial building, or other end user of the power supplied by power distribution network 110. In one example, facilities 124 and 125 are coupled to meters 130 and 131. Meters 130 and 131 measure the power consumption by facilities 124 and 125, respectively, over time. Although only facilities 124 and 125 are shown in power distribution network 110, it will be understood that additional facilities, such as an entire network of facilities, may also be disposed in power distribution network 110.

[0016] The facility monitoring application 101 can derive a set of detailed information about the facilities 124 and 125 based on the power consumption data observed by the meters 130 and 131. For example, the meters 130 and 131 can transmit data regarding power consumption at the facilities 124 and 125 to the head-end system 102. In one example, the facility monitoring application 101 can determine the status of a particular facility of interest (e.g., no electric vehicles charging, level 1 electric vehicles charging, level 2 electric vehicles charging, etc.). For example, the facility monitoring application 101 can leverage data obtained from the meters 130 and 131 and other data sources in the distribution network 110 to predict the status of the facilities in the distribution network 110. In one example, the facility monitoring application 101 can apply one or more machine learning models to data obtained from the meters 130 and 131, the distribution network 110, other relevant data sources, or a combination thereof to generate a predictive display of the facility's status. In some examples, new machine learning models can be generated and trained for the purposes described herein. In other examples, data obtained from meters 130 and 131, distribution network 110, and other relevant data sources may be subjected to existing machine learning models trained using data similar to that collected by meters 130 and 131. For example, supervised regression machine learning models or reinforcement learning models may be used for the purposes described herein.

[0017] Using the facility's predicted status, the facility monitoring application 101 can use it for power generation scheduling or other resource provisioning management operations of the power distribution network 110. For example, the facility's predicted status can provide information used to predict power demand throughout the day. In such an example, a facility with electric vehicle charging facilities may consume more power at night than a facility without electric vehicle charging facilities. By tracking the facilities performing electric vehicle charging operations and the type of charging operation (e.g., Level 1 or Level 2), the power distribution network 110 can control power generation to meet the demand for electric vehicle charging operations on the power distribution network 110 during what are typically considered non-peak periods. By controlling power generation based on the facility's status, the power distribution network 110 can avoid burdening the network with overconsumption and avoid excessive power generation due to overestimation of power consumption. In a further example, the power distribution network 110 can intelligently schedule the charging of detected electric vehicles using common grid components to prevent the common grid components from becoming overloaded during charging operations.

[0018] If additional electric vehicles are detected being charged at facilities that share a common transformer (or if future charging activity is predicted at the facilities), the utility may proactively replace the common transformer with a higher capacity version of the transformer. Other components of the resource distribution network may also be replaced or updated based on potential changes from general demand at the facilities served by those other components as additional electric vehicles are charged at the facilities served by those other components.

[0019] In one example, the machine learning model of the facility monitoring application 101 may be trained based on a historical corpus of data obtained from one or more electric distribution networks. For example, training data and validation data including ground truth states may be used to train the machine learning model and to verify the accuracy of the trained machine learning model. The machine learning model may be trained in a manner such that historical data provided to the machine learning model results in an output that is consistent with a representation of the ground truth states. After initial training, as conditions in the electric distribution network 110 change over time, the machine learning model may be further adjusted using additional data points and additional ground truth states obtained from the facilities 125 and 125 of the electric distribution network 110.

[0020] In some examples, the trained machine learning model of the facility monitoring application 101 may be trained to predict future conditions of facilities in the distribution network 110. For example, the machine learning model can utilize data trends provided by the condition of facilities in the distribution network 110 over time to predict electric vehicle charging behavior at facilities in the distribution network 110 at future points in time. These predictions enable the utility to maintain a plan for capacity changes in the distribution network 110 over time.

[0021] In one example, a further machine learning model can be trained to generate predictions of electric vehicle charging activity within the power distribution network 110. In such an example, the output of the machine learning model over time that detects the presence of electric vehicle charging activity can be used as input to the further machine learning model. Thus, the further machine learning model may be trained to recognize trends that enable prediction of future electric vehicle charging activity within the power distribution network 110.

[0022] 2 is a flowchart of a process 200 for implementing the facility monitoring application 101 to control power sources in the power distribution network 110. In block 202, the process 200 includes accessing consumption data for a set of facilities 124 and 125 in the power distribution network 110. The data may be received at the head-end system 102 from metering devices 130 and 131 or from other data collection points in the power distribution network 110. Some data may be collected automatically by the meters 130 and 131, while other data may be collected manually in response to maintenance activities. For example, maintenance activities such as the installation of a distributed energy resource meter at a facility, the installation of an electric vehicle charger at a facility, or other related maintenance activities may be manually collected and reported to the facility monitoring application 101.

[0023] In block 204, process 200 includes applying a trained machine learning model to the collected data. The trained machine learning model may be trained to classify whether electric vehicle charging operations are occurring at individual facilities in the power distribution network 110. For example, the trained machine learning model may be applied to the consumption data obtained in block 202, which may be a time series of electricity consumption at each facility. In some examples, the time series of electricity consumption may represent instantaneous consumption over an interval. For example, the instantaneous consumption may occur at 10-minute or 15-minute intervals over a seven-day period. The trained machine learning model may be trained on similar data in the time domain to identify a set of facilities performing electric vehicle charging operations.

[0024] In block 206, process 200 includes generating a set of facility classification information based on the output of the trained machine learning model. In one example, the trained machine learning model may be trained to output an indication of a facility performing electric vehicle charging operations or a facility not performing electric vehicle charging operations. The facility's electric vehicle classification may be a one-hot encoding of a softmax function of a set of possible electric vehicle classifications. That is, the classification indicates one class of the set of possible classifications for each facility. In some examples, the classification of a facility performing electric vehicle charging operations may also include identifying the type of charger that performs electric vehicle charging operations. For example, the classification of a facility may include an indication that a Level 1 electric vehicle charger or a Level 2 electric vehicle charger is operating at the facility. For example, a Level 1 charger can provide approximately 1.2 kW and operate directly from a standard 120 VAC outlet. A Level 2 charger may be a charger in the 6.2-19.2 kW range and rely on a 208-240 V, 12-80 A circuit. In some examples, each facility within the power distribution network 110 may be classified based on whether electric vehicle charging operations occur within the facility.

[0025] At block 208, process 200 includes controlling a power source in the power distribution network 110 based on the classification of the facilities generated by the machine learning model. In one example, the classification generated by the machine learning model identifies the number of facilities performing electric vehicle charging operations within the power distribution network 110, and in some examples, the types of chargers performing electric vehicle charging operations. The generators supplying power to the power distribution network 110 may be controlled to meet the expected demand for the power distribution network 110 using information associated with the amount of facilities performing electric vehicle charging operations. For example, as the number of facilities performing charging operations increases, the generators may increase the power supply provided to the power distribution network 110 during periods when charging operations are likely to occur.

[0026] In one example, the power supply of the power distribution network 110 may be adjusted upon detecting new facilities performing electric vehicle charging operations. In some examples, the power supply of the power distribution network 110 may be adjusted only if a threshold number of new facilities performing electric vehicle charging operations is detected. For example, control of the power supply may occur if, since the last adjustment of the power supply, the total number of new facilities performing electric vehicle charging operations is expected to affect the ability of the power distribution network 110 to meet demand at the facilities. In such an example, the power supply may be adjusted if the total number of new facilities performing electric vehicle charging operations is expected to cause the power distribution network 110 to consume a threshold percentage (e.g., 1% more, 5% more, etc.) more power than after the last adjustment. Other threshold percentages may also be used, and other triggering events associated with the addition of facilities performing electric vehicle charging operations may trigger the power supply adjustment in block 208. In some examples, the classification information may be used by the power distribution network 110 to determine whether components of the power distribution network 110 should be updated for higher capacity. For example, transformers operating at facilities classified as performing electric vehicle charging operations may be upgraded to higher capacity models than typical standard capacity transformers. Other types of components of the electrical distribution network 110 may be upgraded similarly.

[0027] In some examples, process 200 can be implemented at any electric utility company with a sensor network that provides sensory data that can be leveraged by machine learning models to predict the electric utility company's electricity consumption needs.

[0028] 3 is an example diagrammatic representation of a data flow 300 for identifying facilities performing electric vehicle charging operations, according to some embodiments described herein. Electrical consumption data 302 may be received by facility monitoring system 100. In one example, electrical consumption data 302 is a series of time-domain consumption data for facilities in electrical distribution network 110. Electrical consumption data 302 may be obtained at 10-minute or 15-minute intervals over a seven-day period for each facility in electrical distribution network 110 for which a determination is sought as to whether the facility is performing electric vehicle charging operations. Other time periods and interval lengths may be used as well.

[0029] A trained machine learning model, such as a convolutional neural network, can be applied to the electricity consumption data 302 to detect low-level features 304 of the electricity consumption data 302. The low-level features 304 may include relatively coarse-grained features represented by the electricity consumption data 302. For example, the low-level features 304 may include peaks in consumption within a facility over time, floors in consumption within a facility over time, or other data anomalies associated with the electricity consumption data 302.

[0030] Upon detecting the low-level features 304, the machine learning model can detect high-level features 306 in the electrical consumption data 302. For example, the machine learning model can be used to interpret or classify the low-level features 304 in the electrical consumption data 302. In some examples, the high-level features 306 may be identified by a layer of the machine learning model that is closer to the output layer than the layer used to generate the low-level features 304.

[0031] The temporal ordering 308 of the electricity consumption data 302 may also be used by the machine learning model. Because the electricity consumption data 302 is provided in the time domain, various features of the electricity consumption data 302 may be related to the temporal ordering of the electricity consumption data 302. For example, the machine learning model may use the temporal ordering 308 to identify features of the electricity consumption data 302 that may occur consecutively with a particular event or that may occur at a particular time on a particular day. Electric vehicle charging occurring at a particular time (e.g., overnight) may be an important indicator used by the machine learning model to identify electric vehicle charging operations within a facility. Furthermore, the length of time of increased electricity consumption identified by the electricity consumption data 302 may also be an indicator used by the machine learning model to identify the type of charger used at a facility for electric vehicle charging operations. For example, charging operations performed at a Level 1 charger may increase electricity consumption for a longer period of time than charging operations performed at a Level 2 charger.

[0032] Based on the low-level feature detection 304, the high-level feature detection 306, and the temporal ordering 308, a classification 310 may be generated for each of the facilities represented in the electricity consumption data 302. In one example, the classification may include an indication that the facility is not performing electric vehicle charging operations, an indication that the facility is performing Level 1 electric vehicle charging operations, or an indication that the facility is performing Level 2 electric vehicle charging operations. Other classifications 310 are possible, such as identifying other types of chargers, identifying the charging capacity of the electric vehicle being charged, or detecting other types of charging operations (e.g., stationary batteries at the facility). Once the classification 310 is determined by the machine learning model, an output 312 may be generated for use in controlling power generation operations to meet the demand needs of the facilities in the electricity distribution network 110.

[0033] FIG. 4 is a flowchart of a process 400 for training a machine learning model to identify facilities performing electric vehicle charging operations, according to some embodiments described herein. At block 402, process 400 includes accessing a corpus of training and validation electricity consumption data. In one example, the corpus of data may be divided into a set of training data and a set of validation data used to validate the machine learning model trained using the set of training data. The corpus of training data may be a National Renewable Energy Laboratory (NREL) dataset labeled with electric vehicle charging operation information measured at 10-minute intervals at 600 facilities over a 52-week period. In some examples, the NREL dataset may be further divided into subsets of consumption data measured at 10-minute intervals over a 7-day period, and at block 406 below, the machine learning model may be trained using these subsets of the NREL dataset. The training labels may be one-hot training labels that identify individual datasets as not performing electric vehicle charging operations, performing Level 1 charging operations, or performing Level 2 charging operations. Other datasets may be used as training and validation data as well.

[0034] At block 404, the process 400 includes normalizing the training and validation consumption data. In one example, normalization may include assigning a value between 0 and 1 corresponding to kW values ​​in the consumption data. For example, 0 may represent 0 kW, 1 may represent 35 kW, and all values ​​between 0 kW and 35 kW may be assigned a corresponding value between 0 and 1. Normalization of the training and validation data may be performed using min-max scaling. In one example, the facility consumption data to which the trained machine learning model is applied may also be normalized using a min-max scaling operation. Other normalization scaling operations may be used as well.

[0035] At block 406, process 400 includes training a machine learning model to classify the presence of electric vehicle charging activity using the normalized training data. Using the labeled and normalized training data, the machine learning model can be trained to recognize facilities performing electric vehicle charging activity. In one example, the training data is provided to the machine learning model in the time domain. Thus, the trained machine learning model can be applied to the consumption data to classify whether electric vehicle charging activity is occurring at the facility, and if so, the type of charging activity, also in the time domain.

[0036] At block 408, the process 400 includes validating the trained machine learning model using normalized validation data. The normalized validation data may be a subset of the NREL dataset corpus that was not used to train the machine learning model. Validating the trained machine learning model can evaluate the accuracy of the machine learning model. In some examples, the validation process may include determining that the classification accuracy for the validation data exceeds an accuracy threshold. For example, if the validation accuracy is greater than 95%, the trained machine learning model may be implemented. Other threshold accuracy percentages may be used as well.

[0037] At block 410, process 400 includes updating the trained machine learning model using additional real-world consumption data. In some examples, the trained machine learning model can continuously learn based on additional data received by the machine learning model. In such examples, the machine learning model can be updated based on additional real-world consumption data that is provided to the machine learning model for further updates. Such continuous learning can be preferable as electric vehicles and electric vehicle charging behaviors evolve over time.

[0038] Exemplary Machine Learning Environment 5 is an example of a machine learning environment used to identify facilities performing electric vehicle charging operations, according to some embodiments described herein. Training vectors 502 are displayed along with queries 504 and known responses 506. By way of example, the queries may be requests to classify the state of charging of electric vehicles at facilities in the power distribution network 110. For example, query 504 may be a classification problem, such as an indication of whether a facility is performing a charging operation, and if so, the type of charging operation being performed at the facility. For simplicity of explanation, only two training vectors 502 are shown, but the number of training vectors may be much larger, such as 10, 50, 100, 1,000, 10,000, 100,000, or more.

[0039] The training vectors 502 may be used by a learning module 508 to perform training 510 of a model 512. The learning module 508 may optimize parameters of a model 512, such as a machine learning model, to achieve a quality metric (e.g., accuracy of the model 512) for one or more specified criteria. The accuracy may be measured by comparing known responses 506 with the predicted output of the model 512. The parameters of the model 512 may be iteratively varied to improve accuracy. Determining the quality metric may be implemented for any function, including all risk, loss, utility, and set of decision functions.

[0040] In some training embodiments, gradients of how parameter changes affect the cost function can be determined, which can provide a measure of the accuracy of the current state of the model 512. Gradients can be used in conjunction with a learning step (e.g., a measure of how much the parameters of the model 512 need to be updated at a particular time step in the optimization process). Thus, parameters (which may include weights, matrix transformations, probability distributions, etc.) can be optimized to provide an optimal value of the cost function, which may be measured, for example, by being above or below a threshold (i.e., exceeding a threshold), or by the cost function not changing significantly over several time steps. In other embodiments, training can be performed using methods that do not require Hessian matrices or gradient calculations, such as dynamic programming or evolutionary algorithms.

[0041] The prediction stage 514 can provide a predicted response 516 to a query vector 518 based on a new query 520. The new query 520 can be of a similar type to the query 504 of the training vector 502. If the new query record is of a different type, a transformation can be performed on the data to obtain data in a format similar to that of the training vector 502. The predicted response 516 may correspond to the question encoded in the query vector 518. In some examples, the predicted response 516 may be an indication of whether a facility 124 of the electricity distribution network 110 performs electric vehicle charging operations. In further examples, the predicted response 516 may also indicate the type of charging operation performed at the facility (e.g., Level 1 or Level 2 electric vehicle charging). The model 512 may be trained to generate other predictions related to identifying electric vehicle charging on the electricity distribution network 110.

[0042] The model 512 may include machine learning models such as deep learning models, neural networks (such as deep learning neural networks), kernel-based regression, adaptive basis regression or classification, Bayesian methods, ensemble methods, logistic regression and augmentation, Gaussian processes, support vector machines (SVMs), probabilistic models, probabilistic graphical models, etc. In embodiments using neural networks, widened and tensorized deep architectures, convolutional layers, dropout, various neural activations, and regularization steps can be used.

[0043] FIG. 6 is an example of a machine learning model, an artificial neural network 600, according to some embodiments described herein. As an example, the model 512 may be an artificial neural network 600 including a number of neurons 602 (e.g., adaptive basis functions) organized in layers. The layers include an input layer 608, a first hidden layer 604, a second hidden layer 610, and an output layer 612. Other layer arrangements are also contemplated. For example, the artificial neural network 600 may have more or fewer hidden layers than the two shown in FIG. 6. The neurons 602, i.e., nodes, may be connected by edges 606. Training the artificial neural network 600 may involve iteratively searching for an optimal configuration of the neural network's parameters for feature recognition, classification, and / or predictive performance. Various numbers of layers and nodes may be used. Those skilled in the art will readily recognize variations in neural network design and the design of other machine learning models.

[0044] Exemplary Computing Devices Used for Facility Monitoring FIG. 7 illustrates an exemplary computing device used to detect charging activity of an electric vehicle in accordance with some embodiments described herein. Any suitable computing system may be used to perform the operations described herein. The illustrated example computing device 700 includes a processor 702 communicatively coupled to one or more memory devices 704. The processor 702 executes computer-executable program code 730 stored in the memory devices 704, accesses data 720 stored in the memory devices 704, or both. Examples of processors 702 include a microprocessor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other suitable processing devices. The processor 702 may include any number of processing devices or cores, including a single processing device. The functionality of the computing device may be implemented in hardware, software, firmware, or a combination thereof.

[0045] The memory device 704 includes any suitable non-transitory computer-readable medium for storing data, program code, or both. Computer-readable media include electronic, optical, magnetic, or other storage devices capable of providing computer-readable instructions or other program code to a processor. Non-limiting examples of computer-readable media include flash memory, ROM, RAM, ASICs, or other media from which a processing unit can read instructions. The instructions may include processor-specific instructions generated by a compiler or interpreter from code written in any suitable computer programming language, including, for example, C, C++, C#, Visual Basic, Java, or a scripting language.

[0046] Computing device 700 may include a number of external or internal devices, such as input devices and output devices. For example, computing device 700 is shown with one or more input / output ("I / O") interfaces 708. I / O interface 708 can receive input from input devices and provide output to output devices. Computing device 700 also includes one or more buses 706. Bus 706 communicatively couples each of one or more components of computing device 700.

[0047] Computing device 700 executes program code 730 that configures processor 702 to perform one or more operations described herein. For example, program code 730 causes processor 702 to perform the operations described in Figures 1 through 6.

[0048] Computing device 700 also includes a network interface device 710. Network interface device 710 includes any device or group of devices suitable for establishing a wired or wireless data connection to one or more data networks. Network interface device 710 may be a wireless device and may have an antenna 714. Computing device 700 can communicate over a data network using network interface device 710 with one or more other computing devices implementing computing devices or other functions.

[0049] Computing device 700 may include a display device 712. Display device 712 may be an LCD, LED, touch screen, or other device capable of displaying information about computing device 700. For example, the information may include the operating status of the computing device, network status, etc.

[0050] While the present invention has been described in detail with respect to specific embodiments thereof, it will be understood that those skilled in the art, upon reading the foregoing and following disclosure, may readily make alternatives, variations, and equivalents of such embodiments. Accordingly, it should be understood that the present disclosure has been presented for purposes of illustration and not limitation, and is not intended to exclude the inclusion of modifications, variations, and / or additions to the present subject matter as would be readily apparent to one of ordinary skill in the art.

Claims

1. a processor; a non-transitory computer-readable memory containing instructions executable by the processor; wherein the instructions cause the processor to: accessing facility consumption data for facilities in an electrical distribution network, the facility consumption data including an indication of facility resource consumption over a period of time; applying a machine learning model to the facility consumption data, the machine learning model being trained to generate an output corresponding to an electric vehicle classification for the facility; generating the electric vehicle classification for the facility using the output of the machine learning model; and controlling generation of the power distribution network based on the electric vehicle classification of the facility; performing an action including system.

2. The operation is training the machine learning model to generate the output corresponding to the electric vehicle classification of a plurality of additional electric distribution network facilities using training vectors of ground truth data of the additional electric distribution network facilities. The system of claim 1 .

3. The operation is updating the machine learning model with additional consumption data from a plurality of facilities of the electricity distribution network. The system of claim 2 .

4. 10. The system of claim 1, wherein the electric vehicle classification of the facility includes an indication that the facility is not charging electric vehicles, an indication that the facility is charging the electric vehicles using a first type of electric vehicle charger, or an indication that the facility is charging the electric vehicles using a second type of electric vehicle charger.

5. The system of claim 1 , wherein the electric vehicle classification for the facility comprises a one-hot encoding of a softmax function of a set of possible electric vehicle classifications.

6. The system of claim 1 , wherein the display of facility resource consumption over a period of time includes a time series of power consumption by the facility over multiple days at regular time intervals.

7. The operation is normalizing the facility consumption data using min-max scaling that is also used to normalize training consumption data used to train the machine learning model. The system of claim 1 .

8. The system of claim 1 , wherein the facility consumption data includes time domain data.

9. A non-transitory computer-readable medium containing instructions executable by a processor, the instructions causing the processor to: accessing facility consumption data for facilities in an electrical distribution network, the facility consumption data including an indication of facility resource consumption over a period of time; applying a machine learning model to the facility consumption data, the machine learning model being trained to generate an output corresponding to an electric vehicle classification for the facility; generating the electric vehicle classification for the facility using the output of the machine learning model; and controlling generation of the power distribution network based on the electric vehicle classification of the facility; performing an action including Non-transitory computer-readable medium.

10. The operation is training the machine learning model to generate the output corresponding to the electric vehicle classification of a plurality of additional electric distribution network facilities using training vectors of ground truth data of the additional electric distribution network facilities. The non-transitory computer-readable medium of claim 9.

11. The operation is updating the machine learning model with additional consumption data from a plurality of facilities of the electricity distribution network. The non-transitory computer-readable medium of claim 10.

12. 10. The non-transitory computer-readable medium of claim 9, wherein the electric vehicle classification of the facility includes an indication that the facility is not charging electric vehicles, an indication that the facility is charging the electric vehicles using a Level 1 electric vehicle charger, or an indication that the facility is charging the electric vehicles using a Level 2 electric vehicle charger.

13. 10. The non-transitory computer-readable medium of claim 9, wherein the electric vehicle classification for the facility comprises a one-hot encoding of a softmax function of a set of possible electric vehicle classifications.

14. The operation is applying a further machine learning model to the electric vehicle classification at the facility; and generating a prediction of future electric vehicle charging activity using the output of the further machine learning model; and 10. The non-transitory computer-readable medium of claim 9, further comprising:

15. accessing facility consumption data for facilities in an electrical distribution network, the facility consumption data including an indication of facility resource consumption over a period of time; applying a machine learning model to the facility consumption data, the machine learning model being trained to generate an output corresponding to an electric vehicle classification for the facility; generating the electric vehicle classification for the facility using the output of the machine learning model; and controlling generation of the power distribution network based on the electric vehicle classification of the facility; 11. A computer-implemented method comprising:

16. applying a further machine learning model to the electric vehicle classification of the facility and to a plurality of further electric vehicle classifications of further facilities in the electricity distribution network; generating a prediction of future electric vehicle charging activity in the electricity distribution network using the output of the further machine learning model; and The computer-implemented method of claim 15 further comprising:

17. 16. The computer-implemented method of claim 15, wherein the electric vehicle classification of the facility includes an indication that the facility is not charging electric vehicles, an indication that the facility is charging the electric vehicles using a Level 1 electric vehicle charger, or an indication that the facility is charging the electric vehicles using a Level 2 electric vehicle charger.

18. The computer-implemented method of claim 15 , wherein the electric vehicle classification for the facility comprises a one-hot encoding of a softmax function of a set of possible electric vehicle classifications.

19. The computer-implemented method of claim 15 , wherein the display of facility resource consumption over a period of time includes a time series of power consumption by the facility over multiple days at regular time intervals.

20. The computer-implemented method of claim 15 , wherein the facility consumption data comprises time domain data.

Images