METHODS, DEVICES AND SYSTEMS FOR ADAPTIVE DIAGNOSTICS BASED ON DISTRIBUTED MONITORING OF ENERGY ANOMALIES IN A POWER NETWORK

Abstract

Methods, devices, and systems for detecting an anomaly in a power network are described. One method for detecting an anomaly in a power network includes determining base power usage in the power network, receiving data indicative of active power usage in the power network, detecting an anomaly based on the difference between base power usage and active power usage, isolating a fault for an element in the power network that responds to the anomaly detection, and transmitting fault isolation information indicating the fault to a user device. Related devices and systems can perform operations of the method described herein.

Description

METHODS, DEVICES AND SYSTEMS FOR ADAPTIVE DIAGNOSTICS BASED ON DISTRIBUTED MONITORING OF ENERGY ANOMALIES IN A POWER NETWORK FIELD OF INVENTION Several modalities described in this description relate to devices, methods, and systems for a power network and, more particularly, to the adaptive diagnosis of power anomalies in a power network. BACKGROUND OF THE INVENTION A typical facility, such as a data center or manufacturing plant, consumes a substantial amount of energy during daily operations. A large facility often has hundreds of devices and subsystems with varying energy demand and consumption patterns depending on their functional importance. Frequently, a facility operates multiple shifts throughout the day and / or night, placing variable energy demands on the electrical grid depending on the type of production activities, the devices involved, environmental factors such as temperature, and / or the desired and achievable production intensity / speed during operation. SUMMARY OF THE INVENTION Several embodiments of the present invention are directed to QnRcnn / cznz / E / YiAi Ref. 344939 A method for detecting an anomaly in a power network. The method includes determining a base power usage in the power network, receiving an active power usage in the power network, detecting an anomaly based on a difference between the base power usage and the active power usage, isolating a fault for an element in the power network that responds to the detection of the anomaly, and transmitting fault isolation information indicating the fault to a user device. The element is in a lower level of the power network hierarchy than the anomaly. According to some modalities, isolating the fault may include identifying a major anomaly in a first element of a main power grid meter hierarchy that includes a plurality of first elements, searching for a lower anomaly in a plurality of second elements of the lower level hierarchy that is lower than the main meter hierarchy, where the plurality of second elements are associated with the first element, and identifying a second element from the plurality of second elements associated with the lower anomaly. According to some modalities, the lower anomaly may include a first lower anomaly, and the lower-level hierarchy includes a first lower-level hierarchy. The method for detecting an anomaly in a power grid may include searching for a second lower anomaly in a QnRcnn / cznz / E / YiAi plurality of third elements of a second lower-level hierarchy, where the plurality of third elements are associated with the second element, and identify a third element from the plurality of third elements associated with the second lower anomaly. The second lower-level hierarchy is at a lower level in the energy grid than the first lower-level hierarchy. According to some methods, detecting the anomaly may involve identifying an outlier in active energy usage, which corresponds to active energy usage outside a predetermined range of baseload energy usage, and determining that the outlier includes the anomaly in the energy grid. The method may include establishing feedback based on the anomaly and modifying baseload energy usage based on that feedback. According to some modalities, determining anomaly-based feedback may include generating data related to active energy use and the anomaly, and triggering the generation of feedback based on the data related to active energy use and the anomaly. Determining anomaly-based feedback may include providing anomaly-related information to a power grid user, receiving input from the user, and determining feedback based on the user input. According to some modalities, determining anomaly-based feedback may include monitoring active energy use over a period of time, determining a change in the properties of active energy use during that period, and determining feedback based on the change in the properties of active energy use during that period. The element may include a building, a section, a machine, or a component. According to some approaches, the method may include determining the type and / or severity of the anomaly. The type of anomaly may be associated with a repair action taken, which responds to the isolation of the fault. The severity of the anomaly may be obtained from a database that includes historical energy usage data from the power grid. Detecting the anomaly based on the difference between baseload energy usage and active energy usage may involve predicting expected energy usage over a future period based on historical energy usage data, generating a bounded uncertainty for the predicted expected energy usage, and detecting the anomaly, which corresponds to active energy usage that falls outside the uncertainty bound. According to some models, the use of energy The QnRcnn / cznz / E / YiAi active energy use may include a first active energy use in a first time period. The method may include determining a second active energy use in a second time period that is after the first time period, and modifying the base energy use, which responds to both the first active energy use and the second active energy use that is outside the uncertainty boundary. The expected energy use may be predicted based on determining a temporal pattern in the historical energy use data. The method may include determining whether the failure is related to a previously reported anomaly, and reporting the failure when the failure is not related to the previously reported anomaly. Several embodiments of the present invention relate to a device configured to detect an anomaly in a power network. The device includes a base energy usage circuit configured to determine base energy usage in the power network, an active energy usage circuit configured to determine active energy usage in the power network, an anomaly detection circuit configured to detect an anomaly based on a difference between base energy usage and active energy usage, a fault isolation circuit configured to isolate a fault to an element in the power network, responding to the detection of the anomaly, and a transmitter configured to transmit fault isolation information indicating the fault to a user device. The element is in a hierarchy of QnRcnn / cznz / E / YiAi lower level of the energy network than the anomaly. According to some modalities, the fault isolation circuit may include a major anomaly detection circuit configured to identify a major anomaly in a first element of a major power network meter hierarchy, where the major meter hierarchy includes a plurality of first elements, a lower anomaly detection circuit configured to search for a lower anomaly in a plurality of second elements of the lower level hierarchy that is lower than the major meter hierarchy and identify a second element from the plurality of second elements associated with the lower anomaly, where the plurality of second elements are associated with the first element. According to some modalities, the lower anomaly may include a first lower anomaly, and the lower-level hierarchy may include a first lower-level hierarchy. The fault isolation circuit may further include a second lower anomaly detection circuit configured to search for a second lower anomaly in a plurality of third elements of a second lower-level hierarchy and to identify a third element from the plurality of third elements associated with the second lower anomaly. The second lower-level hierarchy is QnRcnn / cznz / E / YiAi is at a lower level in the energy network than the first lower-level hierarchy, and the plurality of third elements are associated with the second element. The device may further include a database containing historical energy usage data, a prediction circuit configured to predict expected energy usage over a future time period based on historical energy usage data, and a bonded uncertainty generator configured to generate an uncertainty boundary for the predicted expected energy usage. The anomaly detection circuit may further be configured to detect anomalies, which are triggered by active energy usage that falls outside the uncertainty boundary. In some configurations, the device may include a feedback circuit configured to determine feedback based on the anomaly. The base power usage circuit may also be configured to modify base power usage based on the feedback. Several embodiments of the present invention relate to a power network that includes an anomaly detection device. The anomaly detection device includes a base power usage circuit configured to determine base power usage in the power network, an active power usage circuit configured to determine active power usage in the power network, and a detection circuit. QnRcnn / cznz / E / YiAi anomaly configured to detect an anomaly based on a difference between base power usage and active power usage, and a fault isolation circuit configured to isolate a fault for an element in the power network, responding to the anomaly detection. The anomaly detection device includes a reporting device that includes a transmitter configured to provide an indication of the fault to a display device. The element is in a lower level of the power network hierarchy than the anomaly. Skilled practitioners will appreciate other features, advantages, and details of the present invention from a reading of the figures and the detailed description of the preferred embodiments that follows; such description is merely illustrative of the present invention. It is observed that aspects of the invention's concepts described with respect to one embodiment may be incorporated into a different embodiment, even though they are not specifically described with respect to those embodiments. That is, all embodiments and / or features of any embodiment may be combined in any manner and / or combination. Other operations may also be performed in accordance with any of the embodiments described herein. These and other aspects of the invention's concepts are described in detail in the description presented below. QnRcnn / cznz / E / YiAi BRIEF DESCRIPTION OF THE FIGURES Figure 1 illustrates an energy measurement hierarchy, according to various modalities described in this description. Figure 2 is a block diagram of anomaly detection in a power network, according to various modalities described in this description. Figures 3A to 3C are a flowchart of operations for detecting an anomaly in a power network, according to various modalities described in this description. Figure 4 is an illustrative graph of energy consumption anomalies of a facility such as a manufacturing plant or data center, according to various modalities described in this description. Figures 5A to 5C are illustrative anomaly graphs at various energy measurement levels, according to various modalities described in this description. Figures 6 to 17 are flowcharts of operations for detecting anomalies in a power network, according to various modalities described in this description. Figure 18 is a block diagram of a device that can be included in the power network to perform the operations of the flow diagrams in Figures 2 to 3C and / or Figures 6 to 17, according to various modalities. QnRcnn / cznz / E / YiAi described in this description. Figure 19 is a block diagram of a power network that can perform the operations of the flow diagrams in Figures 2 to 3C and / or Figures 6 to 17, according to various modalities described in this description. DETAILED DESCRIPTION OF THE INVENTION Several modalities will be described more fully below with reference to the accompanying figures. Other modalities can take many different forms and should not be interpreted as limited to the modalities set forth in this description. Similar numbers refer to similar elements in all respects. In the figures, positions, connections, or relative features may be exaggerated for clarity. However, this invention can be characterized in many different ways and should not be interpreted as limited to the embodiments set forth herein. These embodiments are provided so that this description is exhaustive and complete, and will fully convey the scope of the invention to those skilled in the art. A power grid can include a facility such as a data center or manufacturing plant that has substantial energy demands. The facility may include a variety of devices or components, each of which operates QnRcnn / cznz / E / YiAi based on various electrical energies. The variety of devices in the facility can be arranged in a hierarchical structure. For example, as illustrated in Figure 1, a facility such as a data center or manufacturing plant may include several buildings, each of which may have unique energy requirements. Each building may be divided into several sections that have different energy requirements. Each section may include a variety of machines, which in turn may each include a variety of components. With reference to Figure 1, a facility 110 may include a plurality of buildings 120a to 120n. Each building may include a plurality of sections 130a to 130n. Each section may include a plurality of machines 140a to 140n. Each of the machines may include components 150a to 150n.As an example, in a data center, sections 130a to 130n could each be a group of servers, with one group of servers assigned to a client. Machines 140a to 140n could be individual servers, IT racks, chillers, compressors, or support equipment. Components 150a to 150n could be components of, for example, a chiller, such as a fan, compressor, evaporator, etc. In the illustrative hierarchy shown in Figure 1, each hierarchical level and subsequent elements within those hierarchical levels can have different demand patterns and QnRcnn / cznz / E / YiAi energy consumption according to its functional relevance. The facility may operate in several shifts throughout the day and / or night, which places variable energy demands on the electrical grid depending on the types of production activities, the devices in operation, and the intensity and / or speed of production at any given time. The hierarchy in Figure 1 is intended to represent a non-limiting example, such that fewer or more levels of hierarchy are possible without deviating from the scope of the various modalities. Several modalities described herein can arise from the recognition that anomalies in the power grid can indicate faults that need to be detected and isolated within the hierarchical structure. Anomalies can be identified at multiple levels in the power grid hierarchy, as will be discussed with respect to the various modalities described herein. Power anomaly detection in facilities such as data centers, factories, plants, commercial buildings, etc., is described herein. The power grid can learn typical energy consumption patterns for a given facility and discern deviations from those patterns. The power grid can identify the location of such energy anomalies in the facility's metering hierarchy. Power anomaly detection can QnRcnn / cznz / E / YiAi can occur at multiple levels in the facility's power chain hierarchy, from the individual machine, component, or subsystem level to the entire facility level. Once a power anomaly is detected, a downstream reporting mechanism can trigger a notification along with the likely source of the anomaly to provide fault isolation. The system can recommend an action to mitigate the fault. At the end of the cycle, the system can attempt to improve performance through explicit and / or implicit feedback mechanisms. An anomaly can occur when energy usage deviates from normal or baseline energy consumption. Anomalies in the energy consumption pattern of a machine, section, or facility are frequently the result of defective machine parts, operator errors, or faulty or degraded electrical components. Failure to diagnose an anomaly, assess its severity, or isolate the faults causing it can lead to increased energy costs, reduced productivity, unplanned downtime, and / or safety-related incidents. Therefore, it is important for a facility to diagnose anomalies in real time to take corrective action. Anomaly detection for homes and buildings Commercial QnRcnn / cznz / E / YiAi can occur when reading a single energy meter for the entire building. Energy monitoring and anomaly detection based on a single energy meter at the building hierarchy level can suffer from limitations such as low detection rates, a large number of false alarms, limited isolation capability, and / or a lack of recommended corrective actions. Energy meters read at a single hierarchical level, such as for an entire building, can produce a slow energy anomaly detection rate because thresholds such as lower and upper control limits may be used that miss contextual anomalies and / or slow degradation between control limits.Energy meters at a single hierarchical level can produce a large number of false alarms because the energy consumed in a facility depends on energy usage patterns, local weather patterns, operating schedules / shifts, process flow information, the functional characteristics of the machinery involved in production, and / or the types of computing resources being used at any given time. A large number of false alarms can be a significant nuisance for a customer or end user who relies on the power grid to supply energy to their facility. Energy meters at a single hierarchical level offer limited fault isolation. When monitored at the facility level, it is difficult to isolate anomalies. QnRcnn / cznz / E / YiAi occur at a lower level of the hierarchy, such as a building / section / machine within the facility, due to a lack of downstream information. Energy meters at a single hierarchical level may not provide sufficient information to recommend corrective actions to improve the energy performance of individual elements. Recommendations for corrective action would prevent increased electricity bills, lost productivity, or unplanned downtime. Additional challenges for anomaly detection in facilities can arise from seasonally varying energy consumption and other energy usage variations. For example, energy consumed at any given time depends not only on seasonally varying loads such as lighting and HVAC, but also on non-seasonal factors such as production demand, which can make it difficult to identify seasonal patterns in energy consumption. Compared to the residential and commercial building sectors, energy consumption in industrial settings can be highly variable due to different types of loads with variable energy characteristics and the non-seasonal factors that govern their operation. These variations can create challenges for anomaly detection, as anomalies may be attributed to variations in the data. Therefore, techniques for QnRcnn / cznz / E / YiAi anomaly detection may need to be robust when using available contextual information. According to several modalities described herein, data-driven techniques are applied to learn the normal energy consumption behaviors of devices, components, machines, sections, buildings, and facilities. The knowledge gained regarding energy consumption activities can be used to detect deviations in behavior during the daily operations of a facility, thereby classifying and / or identifying a potential energy anomaly in a power network. Anomaly detection procedures may attempt to pinpoint the location of the anomaly. Subsequently, fault isolation can be provided within the metering hierarchy. A potential solution to address a detected anomaly may be recommended. Several modalities described herein may use direct and indirect feedback mechanisms to analyze whether the flagged behavior has been corrected.Such feedback can be useful for adaptive learning. Adaptive learning may be necessary due to the dynamic nature of the underlying energy-consuming processes. In this sense, the feedback mechanism can close the loop in such a way that the learning of the QnRcnn / cznz / E / YiAi Energy system behavior occurs when the marked anomaly is observed to be part of normal operation under current circumstances. By using feedback information, the energy network may be able to autonomously correct itself to recognize future occurrences of similar behavior patterns as normal. This feedback mechanism, however, can be separated from the primary learning mechanism of the operations described herein, which learns a baseline energy usage behavior of the facility from historical data. In some modalities, the adaptive learning mechanism can augment the baseline energy usage behavior during online operation of the system. To address the challenges mentioned above in detecting anomalies in a power grid, several approaches describe operations for detecting anomalies and isolating faults in a power network. In addition to electrical energy consumption data, process data can be used, which may include information on variations such as non-seasonal variations. For example, non-seasonal variations in energy use might include the production rate of a factory, which depends on business requirements, or usage patterns in the servers of a data center. According to the various approaches described herein, anomaly detection can be performed using energy data. High-variance QnRcnn / cznz / E / YiAi data, due to its sensitivity to contextual characteristics, may require adequate variance capture. For example, when using production rate data, it may be possible to model expected variations in these contexts based on historical data. Deviations from historical data in these contexts can then be successfully identified and reported as an anomaly in the power grid. Figure 2 is a block diagram of anomaly detection in a power network, according to various modalities described herein. Referring to Figure 2, an installation 200 in a power network includes several energy-consuming elements. Anomaly detection can be achieved using an Internet of Things (IoT) architecture or an on-premises architecture such as a private cloud deployed within the data center. Information or data regarding useful energy consumption, processes, production, and the environment associated with installation 200 can be provided to one or more anomaly detection devices, which include one or more processors for determining anomalies in the power network. The anomaly detection system may include output terminal processing 220 and input terminal processing 250.Data associated with installation 200 can be provided to a. QnRcnn / cznz / E / YiAi predictive machine learning module 222 and / or a historical energy consumption database 230. The historical energy consumption database 230 can store information regarding the energy consumption of the facility 200 over time and provide this information as training information to the predictive machine learning module 222. The predictive machine learning module 222 can generate a time-series prediction model. The predictive machine learning module 222 can provide process data to the anomaly detection module 224. The anomaly detection module 224 identifies potential anomalies in the energy data and provides this information to an anomaly scoring module 226. The anomaly scoring module 226 can quantify how far the current active energy usage is from a baseline energy usage. Active energy relates to the actual energy consumed by the customer's devices and / or equipment. Baseline energy can be the expected or forecasted energy. The anomaly scoring module 226 can detect an anomaly based on the difference between baseline energy usage and active energy usage by, for example, weighting various factors or statistically differentiating the difference.According to some modalities, base energy use may come from the predictive model, while active / actual energy consumption. QnRcnn / cznz / E / YiAi originates from sensors associated with energy measurement. The anomaly scoring module 226 can provide an identified anomaly or an indication of a fault to the reporting module 234. The reporting module 234 can access the database 236 to obtain operational relationships between machines and / or other elements in the power distribution hierarchy of the facility 200. The adaptive learning module 228 can receive information from the predictive machine learning module 222 and / or other data related to the facility 200. In some modes, the adaptive learning module 228 can receive user feedback from a user feedback module 260.The adaptive learning module 228 can use data from the predictive machine learning module 222, data regarding the facility's energy use, and / or user feedback from the user feedback module 260 to provide feedback to the predictive machine learning module 222. The predictive machine learning module 222 can use this feedback to adapt operations for determining anomalies in the power network. Even with reference to Figure 2, a user 210 can take actions that affect the energy use of installation 200. User 210 can obtain information from the terminal QnRcnn / cznz / E / YiAi input 250 from the power network and provide input to guide the operation of the power network. The input terminal 250 may include a display and / or other indicators that provide anomaly detection information 252, fault isolation information 254, severity assessment 256 and / or recommended actions 258. User-provided input 210 that responds to particular anomalies may be stored in a database to help make future recommendations when faults are detected. Figures 3A to 3C are a flowchart of operations for detecting an anomaly in a power network. Figures 3A to 3C illustrate a detailed flowchart of the predictive machine learning module 22, the anomaly detection module 224, and the anomaly scoring module 226 from Figure 2. Figures 3A to 3C describe the operations involved in detecting anomalies in a power data stream. The power data can be preprocessed and fed into a pre-trained predictive deep learning-based regression model that predicts the expected normal range of the data. The actual power data values ​​received at the current time can be evaluated to determine if they fall within the normal range. If the current active power data usage is outside the normal range, an anomaly may have occurred. The anomaly is scored based on how far the value falls from the expected range. QnRcnn / cznz / E / YiAi actual in relation to the predicted interval. This mechanism of comparing current active energy data to a predicted interval is deployed at various levels in the facility's measurement hierarchy to attempt to identify a fault associated with the anomaly at a lower level of the hierarchy. With reference to Figure 3A, a training phase 300 may include preprocessing 304, based on historical energy data from a database 302. The preprocessed data is then used for training using a machine learning technique, such as short-term long-term memory learning (LSTM), to learn patterns in the historical data. The resulting model can be stored in block 308 and fed to a trained deep learning model 316. The current active energy usage data stream 314 can be preprocessed 312 and fed to the trained deep learning model 316. The output of the trained deep learning model 316 is used to determine the prediction and confidence limits for future energy usage in block 318.Some post-processing 320 can be performed at these prediction and confidence limits, and then a decision 322 can be made based on a comparison of the current active energy usage data stream 314 with the prediction and / or confidence limits. If the data stream. QnRcnn / cznz / E / YiAi 314 of current active energy usage is outside the confidence limits, an anomaly score can be determined based on an error distance, in block 324. The anomaly detection system can include two learning phases: an offline learning phase and an online learning phase. The offline learning phase involves using a learning base for a given installation based on its historical data. The offline learning phase can be referred to as the training phase, as shown in Figure 3A. The online learning phase involves the system applying changes to the base model according to changes in the energy consumption behavior of various elements in the installation. The online learning phase may include event-driven feedback and / or continuous learning. Event-driven feedback uses implicit (soft) or explicit (hard) feedback to trigger updates to the base model. Updates are triggered based on feedback received on anomalies reported through various feedback mechanisms. These feedback mechanisms may include explicit feedback directly from the user or operator, or implicit feedback by implicitly monitoring data for expected corrective changes. QnRcnn / cznz / E / YiAi Implicit feedback can be less effective compared to explicit feedback due to the time required for machine learning to converge. However, implicit feedback can be provided as redundancy so that explicit feedback is optionally available to the user. Implicit feedback, being an optional feature, ensures that the power system can operate without human intervention. Explicit and implicit feedback can both be used simultaneously, or one can be selected to control the operations described herein.Continuous learning monitors track changes in the statistical properties of the data and / or the accuracy of the model over time to decide whether the base energy usage needs to be updated to incorporate changes in behavior at a particular hierarchy level, such as a device, section, building, or facility. With reference now to Figure 3B, the prediction and confidence limits in block 318 are used to predict whether the confidence prediction is consistently low, or whether negative feedback repeatedly increases, in block 326. If this is the case, then the training pipeline is triggered with new data in block 328 using feedback information obtained from QnRcnn / cznz / E / YiAi accumulate implicit and / or explicit feedback in block 352 of Figure 3C. Monitoring for improved prediction confidence or negative feedback is performed in block 330. If a stable improvement is observed in block 332, then learning is stopped. If a stable improvement, i.e., a steady state, is not observed in block 332, then the system continues learning in block 334 to achieve improved base energy utilization. Reports and feedback are provided to allow the user to interact with the overall system. The user can receive reports on an interface or screen and may be able to provide feedback to the system, which will be used for adaptive learning. Figure 3C incorporates an additional machine learning model to further classify detected anomalies into types and severity levels. This classifier can be a pre-trained model trained on a database of anomaly types and severity levels, which can be obtained from the facility's historical records. This knowledge can be stored in a database that includes mappings from historical failures to their remedies or repair actions taken in the past. With reference now to Figure 3C, the anomaly score that was determined in block 324 can be evaluated for QnRcnn / cznz / E / YiAi determines if an anomaly is redundant, in block 336. If the anomaly is determined not to be redundant, the type and / or severity of the anomaly can be classified, in block 338. If the anomaly is determined to be critical, in block 340, an alarm notification can be sent, in block 342. The alarm notification can include suggested action items, in block 344, based on historical fault information and repair actions that were performed, which are stored in a 360 fault database. Information related to the anomaly can be provided or displayed, including a level in the hierarchy at which the anomaly occurred, fault isolation information that identifies an element or a lower level of the anomaly, severity information, and / or suggested actions, in block 346.If no user feedback is provided (364), anomaly detection continues to look for expected changes in energy data in response to the corrective action taken (block 348). Based on the collective action taken and the subsequent expected changes (block 350), positive or negative feedback is provided so that implicit and explicit feedback can accumulate, respectively (block 352). This implicit and / or explicit feedback can then be provided to activate the training pipeline (block 328). User feedback (QnRcnn / cznz / E / YiAi) 364 can also be a portion of the accumulated feedback in block 352. User feedback can be used to update the fault database (360) and / or retrain the fault classification system in block 362. In some cases, user feedback may conflict with the implicit feedback generated. The implicit feedback can be ignored or disabled in these cases. The fault database (360), which includes information on historical faults and subsequent repair actions, can be used to train a classification model (356) during the training phase. Facility hierarchy data from a database (354) can also be used to train the classification model (356). This classification model can be saved in block 358 and used to classify the type and / or severity of faults in block 338. Figure 4 is an illustrative graph of energy consumption anomalies for a facility such as a manufacturing plant. With reference to Figure 4, the energy meter data for the actual facility level are shown. The novel high peaks are the abnormally high energy consumption points identified through the operations described with respect to the flowcharts in Figures 2, 3A, 3B, and / or 3C. Since the interval is predicted In addition to normal energy usage, abnormally low consumption points can also be identified. These low points are identified in this illustrative graph, along with the high points. Anomalies can be determined at the power grid installation level, such as the installation level. These anomalies identified at the installation level in Figure 4 are then further evaluated to determine anomalies at lower hierarchy levels and potentially identify a fault in a particular element of the power grid, as will be discussed with respect to Figures 5A to 5C. Figures 5A through 5C are illustrative anomaly charts that occur at various energy measurement levels lower in the hierarchy than the main meter hierarchy in Figure 4. With reference to Figure 5A, anomalies can be determined at a main meter hierarchy in the energy network, such as at the building-level hierarchy for a building complex (120a to 120n) in Figure 1. Spot anomalies, labeled as detected anomalies in Figure 5A, can be detected due to a sharp increase in peak cross-power. For example, a detected anomaly, as labeled in Figure 5A, might be +107 above the predicted uncertainty, while a different contextual detected anomaly is -17 below the predicted uncertainty. These detected anomalies can QnRcnn / cznz / E / YiAi indicates that a subprocess in a lower-level hierarchy may not be functioning properly. For example, a section of the building may have an equipment failure resulting in an anomaly based on lower-than-expected active energy usage. As a further example, a thermostat setting may have been changed in a section of the building, resulting in substantially lower or higher energy usage, but it may not maintain an adequate operating temperature for the equipment in that section. According to some modalities, anomalies may be found at a lower level in the hierarchy, but the anomalies could be upstream, downstream, or neighboring devices. As such, anomalies can be diagnosed at a higher level in the hierarchy or at the same level. With reference to Figure 5B, anomalies are illustrated at a lower level of hierarchy than the main meter hierarchy in Figure 5A, such as in the building's mechanical section within the main meter hierarchy of Figure 5A. In this example, anomalies are observed three times. With reference to Figure 5C, anomalies are illustrated at an even lower level of hierarchy, which is lower than the mechanical section in Figure 5B. The lowest level of hierarchy could be the machine level or the component level in Figure 1. Although Figure 1 illustrates four level hierarchies as a non-limiting example, lower or higher levels are possible. QnRcnn / cznz / E / YiAi hierarchies. In the example in Figure 5C, energy usage in the lowest level of the hierarchy may indicate anomalies in one or more IT machines. These anomalies detected in Figure 5C can be isolated to a single IT machine in this example. This isolation of a single IT machine is achieved by discovering anomalies in the main level meter in Figure 5A and then working down the subsequent levels in the hierarchy in Figures 5B and 5C by performing the operations in the flowcharts of Figures 3A to 3C at the lower levels and finally isolating the fault that leads to an overall spike in the main level of the hierarchy at the lowest level.In the example shown in Figures 5A through 5C, an anomaly observed in the power demand measured by the main upper-level power quality meter on October 2nd at a particular data center was traced to a peak demand resulting from a specific sublevel within the facility. This example illustrates the ability of the modalities described herein to isolate anomalies to a specific sublevel without human intervention. Detecting energy anomalies can include operations such as using a predictive algorithm to estimate future energy consumption as a precursor to anomaly detection. A detection mechanism can be used that is based on the prediction and the actual consumption occurring at the time of the prediction. QnRcnn / cznz / E / YiAi A scoring mechanism can be used based on the difference between expected and actual energy consumption, as well as the criticality of the monitored process. Fault isolation activities can be used to pinpoint the location of the observed energy anomaly. A reporting mechanism may be necessary to determine when and how to report abnormal energy consumption. The type and severity level of the detected anomaly can be identified. An adaptive learning module can adjust to dynamic changes in the energy consumption characteristics of the facility, building, section, machine, and / or component. According to some modalities, a data-driven approach can be used to learn normal energy consumption patterns from historical plant data. The learning mechanism can be a temporal pattern learning algorithm based on a deep learning architecture. The learning algorithm can be trained on historical plant data and used as a predictive model during online deployment. The learning algorithm can function by predicting expected energy consumption in the next prediction horizon based on its acquired knowledge of installation behavior during training. The prediction horizon can be defined in terms of various time intervals such as the next hour, next day, next x minutes, etc. The term "predictive" can also describe an uncertainty boundary as the prediction. This uncertainty boundary can be a learned quality and, therefore, a function of the energy data available to the learning algorithm at the time of training. This uncertainty boundary can help determine the confidence of the learning algorithm's prediction of energy consumption based on past data. According to some models, the predictive algorithm learns a prediction interval rather than a single point. Anomalous energy consumption can be detected if the actual energy consumption, for which the prediction was made, falls outside the uncertainty interval surrounding that prediction. To rate anomalies that have been identified as a measure of their severity, the score can be determined as a function of two numerical quantities: DI and D2. DI is defined as a distance that describes how far the actual energy consumption is from the predicted consumption value, while D2 is defined as the distance from the nearest uncertainty limit. DI, D2, or a combination of these can be used to rate the anomalies. A mechanism is provided for reporting flagged anomalies. The decision to send a notification to the user or other affected authority is based on the anomaly score and the anomaly's severity. The objective QnRcnn / cznz / E / YiAi may be minimizing repeated reports of multiple anomalies where the anomalies occur at close intervals and are more likely to be triggered by a similar underlying cause. According to several methods described herein, it can be determined whether the facility's energy consumption behavior has deviated significantly from historically learned behavior. The uncertainty limit can serve as an input to determine whether a sustained increase in uncertainty in predicted energy consumption implies that the behavior of the monitored physical entity has changed significantly from what was observed during training. A sustained increase in uncertainty can trigger secondary learning, where adaptive learning is used to incorporate new knowledge of energy usage. This secondary learning can then be fed into the reporting mechanism, as a sustained change in behavior may imply degradation of the underlying physical component. Adaptive online learning learns new behavior as new energy data arrives and adjusts the parameters of the main predictive model to fit the new behavior in the energy system. Adaptive online learning should be used with caution because it alters system performance to adapt. QnRcnn / cznz / E / YiAi to the new data previously unseen, which represents the change in the definition of normal behavior, i.e., base energy use. An explicit feedback mechanism can be used to close the loop and learn from input received from a user or operator. An interactive, GUI-based dashboard can allow the user or facility manager to receive status information, anomaly and / or fault information, and provide input mechanisms. An implicit feedback mechanism without user involvement can be used to monitor energy data for changes after an anomaly is detected and subsequently reported. This implicit feedback can then be fed into adaptive online learning to update model parameters. When the operations described herein are applied to multiple levels of the facility's measurement hierarchy, the system can utilize information specific to those levels to detect anomalous energy consumption. For example, at the machine level, machine characteristics can be used as input to the operations described herein, resulting in improved accuracy due to the availability of additional information. In this way, fault isolation can be provided at various levels. Individual QnRcnn / cznz / E / YiAi, to identify at what level the anomaly has occurred and display such information to the user. Figures 6 through 17 are flowcharts of operations for detecting anomalies in a power network, according to various modalities described herein. With reference to Figure 6, detecting an anomaly in a power network may include determining base energy usage in the power network, in block 610. Detecting the anomaly may also include receiving data indicative of active energy usage in the power network, in block 620. Base energy usage may be determined based on a predictive model, which, in turn, learns from historical data. Historical data may be stored in a database. Active energy usage may be received from an energy meter or other monitoring device associated with a facility. Detecting the anomaly may also include detecting an anomaly based on a difference between base energy usage and active energy usage, in block 630.Detecting the anomaly may include isolating a fault for an element in the power network, which responds to the detection of the anomaly, in block 640. Detecting the anomaly may include transmitting fault isolation information indicating the fault to a user device, in block 650. With reference to Figure 7, detecting the anomaly may include identifying an outlier in energy use QnRcnn / cznz / E / YiAi active, which responds to active energy usage that is outside a predetermined range of base energy usage, in block 710. Detecting the anomaly may include determining that the outlier includes the anomaly in the power grid, in block 720. With reference to Figure 8, operations may include determining feedback based on the anomaly, in block 810. Operations may include modifying base energy usage based on feedback, in block 820. With reference to Figure 9, determining the feedback based on the anomaly may include generating data related to active energy use and the anomaly, in block 910. Determining the feedback may include activating the generation of feedback based on data related to active energy use and the anomaly, in block 920. With reference to Figure 10, determining feedback based on the anomaly may include providing anomaly-related information to a power grid user, in block 1010. Determining feedback may include receiving user input, in block 1020. Determining feedback may include determining feedback based on user input, in block 1030. QnRcnn / cznz / E / YiAi With reference to Figure 11, determining anomaly-based feedback can include monitoring active energy use over a period of time, in block 1110. Determining feedback can also include determining a change in active energy use properties over the period of time, in block 1120. Furthermore, determining feedback based on the change in active energy use properties over the period of time can be done in block 1130. Energy use properties can be shifted and / or updated to new properties (i.e., a mean change). Feedback can trigger model relearning. The element can be a building, a section, a machine, or a component. With reference to Figure 12, isolating the fault may include identifying a major anomaly in a first element of a plurality of first elements in a main energy network meter hierarchy, in block 1210. Isolating the fault may include searching for a lower anomaly in a plurality of second elements in a hierarchy lower than the main meter hierarchy, in block 1220. The second elements are associated with the first element. Isolating the fault may also include identifying a second element from the plurality of second elements associated with the lower anomaly, in block 1230. QnRcnn / cznz / E / YiAi With reference to Figure 13, isolating the fault may include searching for a second lower anomaly in a plurality of third elements of a second lower-level hierarchy, in block 1310. The third elements are associated with the second element. Isolating the fault may include identifying a third element from the plurality of third elements associated with the second lower anomaly, in block 1320. With reference to Figure 14, detecting an anomaly may include determining the type and / or severity of the anomaly in block 1410. The type of anomaly may be associated with a repair action taken in response to the fault isolation. The severity of the anomaly may be obtained from a database that includes historical energy usage data from the power grid. With reference to Figure 15, detecting the anomaly based on the difference between baseload energy use and active energy use may include predicting the expected energy use over a future time period based on historical energy use data, in block 1510. Detecting the anomaly based on the difference may include generating an uncertainty boundary for the predicted expected energy use, in block 1520. The anomaly may be detected in response to active energy use that is outside the uncertainty boundary, in block 1530. QnRcnn / cznz / E / YiAi According to some modalities, active energy usage may include an initial active energy usage in a first time period. With reference to Figure 16, the operations may further include determining a second active energy usage in a second time period that follows the first time period, in block 1610. The operations may include modifying the baseline energy usage, which responds to both the initial active energy usage and the second active energy usage that is outside the uncertainty boundary, in block 1620. Modifying the baseline energy usage may include modifying the predictive machine learning model. Expected energy usage can be predicted based on determining a temporal pattern in historical energy usage data. With reference to Figure 17, operations may include determining if the failure is related to a previously reported anomaly, in block 1710. The failure may be reported when the failure is not related to the previously reported anomaly, in block 1720. Figure 18 is a block diagram of a device that can be included in the power network to perform the operations described in the flowcharts of Figures 2 to 3C and / or Figures 6 to 17. With reference to Figure 18, the 1800 device is configured to detect an anomaly in a power network. The 1800 device includes a circuit of QnRcnn / cznz / E / YiAi Base Energy Usage 1810 configured to determine base energy usage on the power grid, Active Energy Usage 1870 configured to determine active energy usage on the power grid, Anomaly Detection 1820 configured to detect an anomaly based on a difference between base energy usage and active energy usage, and Fault Isolation 1880 configured to isolate a fault for an element on the power grid, which respond to anomaly detection. Device 1800 may include a Receiver 1805 that receives base energy usage and / or active energy usage information. Device 1800 may include a Transmitter 1875 that transmits fault isolation information indicating the fault to a user device. The fault isolation circuit 1880 may include a primary anomaly detection circuit 1885 configured to identify a primary anomaly in a first element of a primary meter hierarchy of the power network, where the primary meter hierarchy includes a plurality of first elements, and a lower anomaly detection circuit 1890 configured to search for a lower anomaly in a plurality of second elements of a hierarchy lower level than the primary meter hierarchy and to identify a second element from the plurality of second elements associated with the lower anomaly. The isolation circuit QnRcnn / cznz / E / YiAi fault may include a lower anomaly detection circuit 1895 configured to search for a lower anomaly in a plurality of third elements of a lower level hierarchy and identify a third element from the plurality of third elements associated with the lower anomaly. According to some configurations, the 1800 device may include an 1850 database that stores historical energy usage data, an 1830 prediction circuit configured to predict expected energy usage over a future time period based on historical energy usage data, and an 1860 uncertainty limit generator configured to generate an uncertainty limit for the predicted expected energy usage. The anomaly detection circuit may be further configured to detect anomalies, which are triggered by active energy usage that falls outside the uncertainty limit. The 1800 device may include an 1840 feedback circuit configured to determine anomaly-based feedback. In some modes, the 1810 base power usage circuit may also be configured to modify base power usage based on feedback. Figure 19 is a block diagram of a power network that can perform the operations of the flow diagrams in Figures 2 to 3C and / or Figures 6 to 17. With reference As shown in Figure 19, a power network 1900 may include an anomaly detection device 1910. The anomaly detection device 1910 may include a base energy usage circuit 1920 configured to determine base energy usage in the power network, an active energy usage circuit 1930 configured to determine active energy usage in the power network, an anomaly detection circuit 1940 configured to detect an anomaly based on a difference between base energy usage and active energy usage, and a fault isolation circuit 1950 configured to isolate a fault to an element in the power network, which respond to the detection of the anomaly. The power network 1900 may include a reporting device 1960 configured to provide an indication of the fault to a display device 1990.The 1900 power network may include a 1970 receiver and / or a 1980 transmitter that communicates with a facility, energy measuring devices, users, operators and / or a display device configured to provide fault information. Additional definitions: In the preceding description of various embodiments of this description, it should be understood that the terminology used herein is intended solely to describe particular embodiments and is not intended to be limiting to the invention. Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly known to a person skilled in the art to which this description pertains. It is further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning consistent with their meaning in the context of this description and the relevant art, and should not be interpreted in an idealized or overly formal sense unless expressly defined herein. When an element is described as being connected, coupled, responsive, or variants thereof to another element, it may be connected, coupled, or responsive to the other element, or intervening elements may be present. Conversely, when an element is described as being directly connected, directly coupled, directly responsive, or variants thereof to another element, no intervening elements are present. Similar numbers refer to similar elements in all respects. Furthermore, coupled, connected, responsive, or variants thereof as used herein may include wirelessly coupled, connected, or responsive. As used herein, the singular forms a, an The terms "el" and "la" are intended to include plural forms as well, unless the context clearly indicates otherwise. Known functions or constructions may not be described in detail for brevity and / or clarity. The term "y / o" includes any and all combinations of one or more of the associated listed articles. As used in the present description, the terms comprise, comprising, includes, having, or variants thereof are indefinite, and include one or more indicated features, wholes, elements, stages, components or functions, but do not exclude the presence or addition of one or more additional features, wholes, elements, stages, components, functions or groups thereof. The illustrative methods described herein are based on block diagrams and / or flowchart illustrations of computer-implemented methods, apparatus (systems and / or devices), and / or software products. It is understood that a block in the block diagrams and / or flowchart illustrations, and combinations of blocks in the block diagrams and / or flowchart illustrations, can be implemented by software instructions executed by one or more computer circuits. These software instructions can be provided to QnRcnn / cznz / E / YiAi a processor circuit of a general-purpose computer circuit, special-purpose computer circuit and / or other programmable data processing circuit to produce a machine, such that the instructions, which are executed through the computer processor and / or other programmable data processing devices, transform and control transistors, values ​​stored in memory locations and other hardware components within such circuit system to implement the functions / actions specified in the block diagrams and / or flowchart block(s) and, in this way, create means (functionality) and / or structure to implement the functions / actions specified in the block diagrams and / or flowchart block(s). These computer program instructions can also be stored on a tangible computer-readable medium that can direct a computer or other programmable data-processing device to operate in a particular manner, such that the instructions stored on the computer-readable medium produce a manufactured item that includes instructions implementing the functions / actions specified in the block diagrams and / or block or block flowcharts. A tangible, non-transient, computer-readable medium can include any electronic, magnetic, optical, electromagnetic, or semiconductor data storage system, apparatus, or device. More specific examples of computer-readable media include a laptop floppy disk, random-access memory (RAM) circuit, read-only memory (ROM) circuit, erasable programmable read-only memory (EPROM or Flash memory) circuit, portable compact disc (CD-ROM) read-only memory, and portable digital video disc (DVD / Blu-ray) read-only memory. The instructions of the computer program can also be loaded into a computer and / or other programmable data processing device to cause a series of operational steps to be performed on the computer and / or other programmable device to produce a computer-implemented process, such that the instructions executed on the computer or other programmable device provide steps for implementing the functions / actions specified in the block diagrams and / or flowchart block(s). Accordingly, the modalities of the present description can be incorporated into hardware and / or software (including firmware, resident software, microcode, etc.) that runs on a processor such as a digital signal processor, which may be collectively referred to as a circuit system, a circuit, or a QnRcnn / cznz / E / YiAi module, a unit or variants thereof. The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of computer systems, methods, and products according to various aspects of this description. In this sense, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, comprising one or more executable instructions to implement the specified logical function(s). It should also be noted that, in some alternative implementations, the functions indicated in the block may occur out of the order shown in the figures. For example, two blocks shown in succession may, in fact, execute substantially simultaneously, or the blocks may sometimes execute in reverse order, depending on the functionality involved.It will also be noted that each block in the block diagrams and / or flowchart illustration, and combinations of blocks in the block diagrams and / or flowchart illustration, can be implemented by special-purpose hardware-based systems that perform the specified functions or actions, or combinations of special-purpose hardware and computer instructions. It should also be noted that in some implementations As alternatives, the functions / actions indicated in the blocks may occur out of the order shown in the flowcharts. For example, two blocks shown in succession may, in fact, execute substantially at the same time, or the blocks may sometimes execute in reverse order, depending on the functionality / actions involved. Furthermore, the functionality of a given block in the flowcharts and / or block diagrams may be separated into multiple blocks, and / or the functionality of two or more blocks in the flowcharts and / or block diagrams may be at least partially integrated. Finally, other blocks may be added / inserted between the blocks shown. Additionally, although some of the diagrams include arrows in the communication paths to show a primary direction of communication, it should be understood that communication may occur in the opposite direction to the arrows shown. Many different embodiments have been described in the present description, along with the preceding description and figures. It would be unduly repetitive and obfuscating to describe and illustrate verbatim every combination and subcombination of these embodiments. Accordingly, the present description, including the figures, shall be construed as constituting a complete written description of various illustrative combinations and subcombinations of embodiments and of the manner and process of their manufacture and use, and shall support the claims to any such combination or QnRcnn / cznz / E / YiAi subcombination. Many variations and modifications can be made to the modalities without substantially departing from the principles of the present invention. It is intended that such variations and modifications be included in the present description within the scope of the present invention. It is hereby stated that, as of this date, the best method known to the applicant for putting the aforementioned invention into practice is the one that is clear from the present description of the invention. QnRcnn / cznz / E / YiAi a lower anomaly detection circuit configured to search for a lower anomaly in a plurality of second elements of the lower level hierarchy that is lower than the main meter hierarchy and identify a second element from the plurality of second elements associated with the lower anomaly, wherein the plurality of second elements are associated with the first element. 17. The device according to claim 16, characterized in that the lower anomaly comprises a first lower anomaly, wherein the lower level hierarchy comprises a first lower level hierarchy, and wherein the fault isolation circuit further comprises: a second lower anomaly detection circuit configured to search for a second lower anomaly in a plurality of third elements of a second lower level hierarchy and to identify a third element from the plurality of third elements associated with the second lower anomaly, wherein the second lower level hierarchy is at a lower level in the power network than the first lower level hierarchy where the plurality of third elements are associated with the second element. 18. The device according to claim 15, characterized in that it further comprises: QnRcnn / cznz / E / YiAi a database comprising historical energy usage data; a prediction circuit configured to predict expected energy usage over a future time period based on historical energy usage data; and an uncertainty boundary generator configured to generate an uncertainty boundary for the predicted expected energy usage, wherein the anomaly detection circuit can be further configured to detect the anomaly, which responds to active energy usage that is outside the uncertainty boundary. 19. The device according to claim 15, characterized in that it further comprises: a feedback loop configured to determine feedback based on the anomaly, wherein the base energy usage loop is further configured to modify the base energy usage based on the feedback. 20. An energy network characterized by comprising: an anomaly detection device comprising: a baseload power usage circuit configured to determine a baseload power usage on the power grid; an active energy usage circuit configured to determine active energy usage on the power grid; QnRcnn / cznz / E / YiAi an anomaly detection circuit configured to detect an anomaly based on a difference between base power usage and active power usage; and a fault isolation circuit configured to isolate a fault for an element in the power network, responding to the detection of the anomaly, wherein the element is in a lower level hierarchy of the power network than the anomaly; and a reporting device comprising a transmitter configured to provide an indication of the fault to a display device.

Claims

1. A method for detecting an anomaly in a power network characterized in that it comprises: determining a base energy usage in the power network; receiving data indicative of an active energy usage in the power network; detecting an anomaly based on a difference between the base energy usage and the active energy usage; isolating a fault for an element in the power network, which responds to the detection of the anomaly; and transmitting fault isolation information indicating the fault to a user device, wherein the element is in a lower level hierarchy of the power network than the anomaly.

2. The method according to claim 1, characterized in that isolating the fault comprises: identifying a major anomaly in a first element of a main power network meter hierarchy comprising a plurality of first elements; searching for a lower anomaly in a plurality of second elements of the lower level hierarchy that is lower than the main meter hierarchy, wherein the plurality of second elements are associated with the first element; and identifying a second element from the plurality of second elements associated with the lower anomaly.

3. The method according to claim 2, characterized in that the lower anomaly comprises a first lower anomaly, and wherein the lower level hierarchy comprises a first lower level hierarchy, the method further comprises: searching for a second lower anomaly in a plurality of third elements of a second lower level hierarchy, wherein the plurality of third elements are associated with the second element; and identifying a third element from the plurality of third elements associated with the second lower anomaly, wherein the second lower level hierarchy is at a lower level in the energy network than the first lower level hierarchy.

4. The method according to claim 1, characterized in that detecting the anomaly comprises: identifying an outlier in active energy usage, which corresponds to active energy usage that is outside a predetermined range of base energy usage; and determining that the outlier comprises the anomaly in QnRcnn / cznz / E / YiAi the energy network.

5. The method according to claim 1, characterized in that it further comprises: determining feedback based on the anomaly; and modifying the base energy usage based on the feedback.

6. The method according to claim 5, characterized in that determining the feedback based on the anomaly comprises: generating data related to active energy usage and the anomaly; and activating the generation of feedback based on the data related to active energy usage and the anomaly.

7. The method according to claim 5, characterized in that determining the feedback based on the anomaly comprises: providing anomaly-related information to a power network user; receiving an input from the user; and determining the feedback based on the user input.

8. The method according to claim 5, characterized in that determining feedback based on the QnRcnn / cznz / E / YiAi anomaly comprises: monitoring active energy usage over a period of time; determining a change in the properties of active energy usage over the period of time; and determining feedback based on the change in the properties of active energy usage over the period of time.

9. The method according to claim 1, characterized in that the element comprises a building, a section, a machine, or a component.

10. The method according to claim 1, characterized in that it further comprises: determining a type of anomaly and / or severity of the anomaly, wherein the type of anomaly is associated with a repair action taken, which responds to the isolation of the fault, and wherein the severity of the anomaly is obtained from a database comprising historical energy usage data from the power grid.

11. The method according to claim 1, characterized in that detecting the anomaly based on the difference between base energy use and active energy use comprises: predicting the expected energy use in a future time period based on historical energy use data; generating an uncertainty boundary for the predicted expected energy use; and detecting the anomaly, which corresponds to active energy use that is outside the uncertainty boundary.

12. The method according to claim 11, characterized in that the active energy use comprises a first active energy use in a first time period further comprises: determining a second active energy use in a second time period that is after the first time period; and modifying the base energy use, which responds to both the first active energy use and the second active energy use that is outside the uncertainty boundary.

13. The method according to claim 11, characterized in that the expected energy use is predicted based on determining a temporal pattern in historical energy use data.

14. The method according to claim 1, further comprising: determining whether the failure is related to a previously reported anomaly; and reporting the failure when the failure is not related to the previously reported anomaly. QnRcnn / cznz / E / YiAi 15. A device configured to detect an anomaly in a power network characterized in that it comprises: a base energy usage circuit configured to determine a base energy usage in the power network; an active energy usage circuit configured to determine an active energy usage in the power network; an anomaly detection circuit configured to detect an anomaly based on a difference between the base energy usage and the active energy usage; a fault isolation circuit configured to isolate a fault for an element in the power network, which responds to the detection of the anomaly; and a transmitter configured to transmit fault isolation information indicating the fault to a user device, wherein the element is in a lower level hierarchy of the power network than the anomaly.

16. The device according to claim 15, characterized in that the fault isolation circuit comprises: a main fault detection circuit configured to identify a main fault in a first element of a main power network meter hierarchy, wherein the main meter hierarchy comprises a plurality of first elements; and

Images