Classifications

• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
  • G06Q10/00—Administration; Management
  • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
  • G06Q10/063—Operations research, analysis or management
  • G06Q10/0635—Risk analysis of enterprise or organisation activities
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
  • G06F17/10—Complex mathematical operations
  • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N5/00—Computing arrangements using knowledge-based models
  • G06N5/04—Inference or reasoning models
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
  • G06Q10/00—Administration; Management
  • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
  • G06Q10/00—Administration; Management
  • G06Q10/20—Administration of product repair or maintenance
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
  • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
  • G06Q50/06—Energy or water supply

Definitions

• the disclosed embodiments generally relate to techniques for improving the reliability of electrical utility systems. More specifically, the disclosed embodiments relate to a technique that uses an irrelevance filter to facilitate efficient remaining useful life (RUL) analyses to improve the reliability of utility system assets in the field.
• Utility system assets such as power transformers, are critical for ensuring the uninterrupted delivery of electrical power from power generation facilities to consumers in electrical distribution grids.
• Electrical grids are typically “fault tolerant” with respect to losing generation assets, because whenever a power plant fails, replacement power can usually be pulled through the distribution grid to meet consumer demand.
• power transformer failures typically lead to “blackouts,” which may affect consumers in small areas comprising a few blocks, or may possibly affect consumers throughout a large service region comprising multiple square miles.
• the failure of a single transformer can potentially cause a very large voltage spike to be propagated throughout the distribution grid, which can cause other transformers to fail, and can lead to a large-scale, regional blackout affecting hundreds of square miles.
• Transformer explosions can also cause fires, which can result in significant property damage and loss of life.
• DGA dissolved gas analysis
• a DGA is essentially “reactive” and not “prognostic” because it detects the downstream symptoms of problems, well after the problems developed that caused hotspots that were sufficient to “bake out” the hydrocarbon gasses.
• RUL remaining useful life
• prognostic-surveillance techniques tend to fail infrequently. This means there may not exist sufficient historical failure data to determine whether an anomalous pattern of sensor signals is indicative of an impending failure, or is simply a new pattern of sensor signals, which is not correlated with an impending failure. This lack of historical failure data means that prognostic- surveillance techniques are likely to generate a high rate of false alarms, which leads to unnecessary maintenance operations, and may cause utility system assets to be prematurely replaced.
• the disclosed embodiments provide a system that estimates a remaining useful life (RUL) of a utility system asset.
• the system iteratively performs the following operations. First, the system receives a set of present time-series signals gathered from sensors in the utility system asset. Next, the system uses an inferential model to generate estimated values for the set of present time-series signals, and performs a pairwise differencing operation between actual values and the estimated values for the set of present time-series signals to produce residuals.
• the system then performs a sequential probability ratio test (SPRT) on the residuals to produce SPRT alarms.
• SPRT sequential probability ratio test
• the system applies an irrelevance filter to the SPRT alarms to produce filtered SPRT alarms, wherein the irrelevance filter removes SPRT alarms for signals that are not correlated with previous failures of similar utility system assets.
• the system uses a logistic-regression model to compute an RUL-based risk index for the utility system asset based on the filtered SPRT alarms.
• the risk index exceeds a risk-index threshold, the system generates a notification indicating that the utility system asset needs to be replaced.
• the system periodically updates the logistic-regression model and the irrelevance filter based on time-series signals from additional utility system assets that have failed.
• the RUL-based index is computed for a utility system asset only when more than a threshold number of filtered SPRT alarms were generated during a preceding time interval.
• the system receives an inferential training set of time-series signals gathered from sensors in the utility system asset during normal fault-free operation. Next, the system trains the inferential model to predict values of the time-series signals based on the inferential training set.
• the system receives an RUL training set comprising time-series signals gathered from sensors in similar utility system assets while the similar utility system assets are run to failure.
• the system also receives associated failure times for the similar utility system assets.
• the system uses the inferential model to generate estimated values for the RUL training set of time-series signals.
• the system then performs a pairwise differencing operation between actual values and the estimated values for the RUL training set of time-series signals to produce residuals.
• the system performs a SPRT on the residuals to produce SPRT alarms with associated tripping frequencies.
• the system trains the logistic-regression model to predict an RUL for the utility system asset based on correlations between the SPRT alarm tripping frequencies and the failure times for the similar utility system assets.
• the system also configures the irrelevance filter. During this process, the system identifies relevant SPRT alarms that were generated during a time interval before a utility system asset failed, and then configures the irrelevance filter to remove SPRT alarms that are not relevant.
• the system while training the logistic-regression model to predict the RUL for the utility system asset, the system only considers SPRT alarm tripping frequencies associated with relevant SPRT alarms.
• the time-series signals gathered from sensors in the utility system asset include signals specifying one or more of the following: temperatures; currents; voltages; resistances; capacitances; vibrations; dissolved gas metrics; cooling system parameters; and control signals.
• the inferential model comprises a Multivariate State Estimation Technique (MSET) model.
• MSET Multivariate State Estimation Technique
• the utility system asset comprises a power transformer.
• FIG. 2 presents a flow chart for a process that estimates an RUL for a utility system asset in accordance with the disclosed embodiments.
• FIG. 3 presents a flow chart illustrating a process for training an inferential model for a utility system asset in accordance with the disclosed embodiments.
• FIG. 4 presents a flow chart illustrating a process for training a logistic-regression model to predict an RUL for a utility system asset and for configuring an associated irrelevance filter in accordance with the disclosed embodiments.
• DETAILED DESCRIPTION [020] The following description is presented to enable any person skilled in the art to make and use the present embodiments, and is provided in the context of a particular application and its requirements.
• the computer-readable storage medium includes, but is not limited to, volatile memory, non-volatile memory, magnetic and optical storage devices such as disk drives, magnetic tape, CDs (compact discs), DVDs (digital versatile discs or digital video discs), or other media capable of storing computer-readable media now known or later developed.
• the methods and processes described in the detailed description section can be embodied as code and/or data, which can be stored in a computer-readable storage medium as described above.
• the computer system When a computer system reads and executes the code and/or data stored on the computer-readable storage medium, the computer system performs the methods and processes embodied as data structures and code and stored within the computer-readable storage medium.
• the methods and processes described below can be included in hardware modules.
• the hardware modules can include, but are not limited to, application-specific integrated circuit (ASIC) chips, field-programmable gate arrays (FPGAs), and other programmable-logic devices now known or later developed.
• ASIC application-specific integrated circuit
• FPGAs field-programmable gate arrays
• the hardware modules When the hardware modules are activated, the hardware modules perform the methods and processes included within the hardware modules.
• ASIC application-specific integrated circuit
• FPGAs field-programmable gate arrays
• the disclosed embodiments make use of a novel “irrelevance filter,” which mimics the functionality of the human brain’s basal ganglia to facilitate improved RUL prognostics for large populations of high-cost utility grid assets, especially high-voltage transformers.
• biomimicry that analyzes nature’s best ideas and adapts them for engineering use cases.
• the invention disclosed herein provides an example of biomimicry.
• Swedish researchers performing MRI studies on human brains discovered that the basal ganglia act as an “irrelevance filter,” which plays a crucial role in human memory and cognition. If the human brain tried to process and store all inputs coming in through the senses, the brain would be overwhelmed. The basal ganglia weeds out unnecessary information, thereby leaving only those details essential to form memories that contribute to survival of a species, such as memories associated with: acquisition of food; avoidance of danger; propagation of the species; and assurance that basic needs are met. It has been shown that humans with the best memories have highly active basal ganglia.
• This basal ganglia paradigm can be useful for facilitating certain types of engineering-related tasks.
• researchers are beginning to explore the possibility of using machine learning (ML), which is based on surveillance of time-series signals obtained from sensors in utility system assets, to facilitate the scheduling of maintenance operations.
• ML pattern-recognition techniques can be trained using sensor signals generated when an asset is deemed to be operating without faults, and can then be used to detect anomalous signal patterns for that asset, which can be used to schedule predictive maintenance to remediate the underlying causes of the anomalous signal patterns.
• ML machine learning
• it is extremely valuable for the asset operator to receive an alarm comprising an early warning about a potential problem.
• RUL estimation provides an estimate of how long that asset will be able to operate safely before the probability of catastrophic failure reaches a critical threshold (e.g. 95% probability of failure).
• a critical threshold e.g. 95% probability of failure
• two transformers in a utility grid may both issue early warning alerts.
• the service organization knows that a first transformer has an RUL estimate of 2 months, but a second transformer is likely to fail in the next 72 hours, it is more beneficial to schedule emergency remediation operations on the second transformer, and to wait for a “convenient maintenance window” to remediate the first transformer.
• an anomalous (but harmless) pattern of time-series signals may be associated with: a relatively new asset; an asset operating in an environment with large temperature fluctuations; or an asset operating in an environment with large fluctuations in electrical flow (e.g., from population changes or utility grid reconfigurations).
• Such alarms may be caused by new patterns in time-series data for individual assets, but may have no prognostic-health significance.
• an “irrelevance filter” that processes time-series signals for utility system assets that have been run to failure, and produces optimal weighting factors for an associated RUL methodology.
• Our new ML-based technique operates by processing data historian files. More specifically, when a population of utility system assets, such as high-voltage transformers, is monitored, the time-series telemetry signals are continuously stored in data historian files, wherein there exists one (logical) data historian file for each monitored asset.
• These data historian files can be “harvested” continuously (e.g., in 1 to 15 minute increments) and added to a large database, where they are processed to discover trends, anomalies, environmental problems, and other incipient problems.
• Our anomaly discovery process uses a systematic binary hypothesis technique called the “sequential probability ratio test” (SPRT) as an irrelevance filter for large volumes of time-series signals, and identifies small subsets of time-series signals that warrant further pattern- recognition analyses to facilitate anomaly detection.
• SPRT sequential probability ratio test
• our new technique substantially reduces RUL-analysis costs by systematically and safely filtering anomaly alerts generated for individual utility system assets so that RUL-analysis operations are only performed for “relevant” signature patterns that are likely to be associated with incipient fault conditions.
• FIG. 1 illustrates an exemplary prognostic-surveillance system 100 in accordance with the disclosed embodiments.
• prognostic-surveillance system 100 operates on a set of time-series sensor signals 104 obtained from sensors in a utility system asset 102, such as a power transformer.
• time-series signals 104 can originate from any type of sensor, which can be located in a component in utility system asset 102, including: a voltage sensor; a current sensor; a pressure sensor; a rotational speed sensor; and a vibration sensor.
• X [X 1 ,..., X m ]
• X(t) [X 1 (t),..., X m (t)] is the value of the time-series sensor signals at time t.
• time-series signals 104 feed into a time-series database 106, which stores the time-series signals 104 for subsequent analysis.
• the time-series signals 104 either feed directly from utility system asset 102 or from time-series database 106 into a non-linear, non-parametric (NLNP) regression model 108.
• NLNP regression model 108 Upon receiving the time-series sensor signals 104, NLNP regression model 108 performs a non- linear, non-parametric regression analysis on the samples (including a “current sample”). When the analysis is complete, NLNP regression model 108 outputs estimated signal values 110.
• NLNP regression model 108 uses a multivariate state estimation technique (“MSET”) to perform the regression analysis.
• MSET multivariate state estimation technique
• MSET User Planar System
• OLS Ordinary Least Squares
• SVM Support Vector Machines
• ANNs Artificial Neural Networks
• RMSET Regularized MSET
• NLNP nonlinear, nonparametric
• SVMs support vector machines
• AAKR auto-associative kernel regression
• LR simple linear regression
• the state matrix D acts as a fixed model of the system from which signal values are estimated. Suppose for the time being that the signal measurements represent linearly correlated phenomena.
• the estimation vector X est must represent an optimum estimation even if some elements of X obs fall outside the range of the same elements in D (i.e., when an observed signal value is less than the minimum or greater than the maximum value of the signal observed during the training phase). 3. If the observation vector X obs is identical to one of the column vectors in D, then the estimation vector X est must be identical to X obs . 4. The difference between X obs and X est must be minimized. Nonlinear operators that fulfill these conditions exist and have been shown to be successful in monitoring real systems. [042] Returning back to FIG. 1, NLNP regression model 108 is “trained” to learn patterns of correlation among the time-series signals 104.
• SPRT module 116 then performs a “detection operation” on the residuals 114 to detect anomalies and possibly to generate SPRT alarms 118.
• SPRT module uses the sequential probability ratio test (SPRT) proposed by Wald to detect subtle statistical changes in a stationary noisy sequence of observations at the earliest possible time. (See Wald, Abraham, June 1945, “Sequential Tests of Statistical Hypotheses,” Annals of Mathematical Statistics, 16 (2): 117– 186.)
• SPRT sequential probability ratio test
• Process signal Y is said to be degraded if the observations made on Y appear to be distributed about mean M with normal (Gaussian) distribution instead of mean zero, where M is a predetermined system disturbance magnitude.
• M normal (Gaussian) distribution instead of mean zero
• M a predetermined system disturbance magnitude.
• NLNP regression model 108 and difference module 112 work together to remove (filter) the dynamics in the signals X(t) so that the residual R(t) is a stationary random process when the system is in good condition.
• SPRT module 116 which generates corresponding SPRT alarms 118.
• SPRT module 116 applies a sequential probability ratio test to the residuals and produces an alarm when one or several residuals become statistically different from the residual corresponding to the undegraded condition of the system.
• the SPRT alarms 118 then feed through an irrelevance filter 120, which removes SPRT alarms for signals that are not correlated with previous failures of similar utility system assets to produce filtered SPRT alarms 124.
• Filtered SPRT alarms 124 feed into a logistic- regression model 126, which generates an RUL estimate 128, wherein the RUL estimate 128 can be expressed as a “quantitative risk index” as is described in more detail below.
• logistic-regression model 126 While calculating RUL estimate 128, logistic-regression model 126 records each instance an alarm in filtered SPRT alarms 124, and uses these instances to determine the current alarm-tripping frequency. As degradation progresses, the tripping frequency of the filtered alarms increases.
• FIG. 2 presents a flow chart for a process that estimates an RUL for a utility system asset in accordance with the disclosed embodiments.
• the system iteratively performs the following operations.
• the system receives a set of present time-series signals gathered from sensors in the utility system asset (step 202).
• the system uses an inferential model to generate estimated values for the set of present time-series signals (step 204), and then performs a pairwise differencing operation between actual values and the estimated values for the set of present time-series signals to produce residuals (step 206).
• the system then performs a sequential probability ratio test (SPRT) on the residuals to produce SPRT alarms (step 208).
• SPRT sequential probability ratio test
• the system applies an irrelevance filter to the SPRT alarms to produce filtered SPRT alarms, wherein the irrelevance filter removes SPRT alarms for signals that are not correlated with previous failures of similar utility system assets (step 210).
• the system uses a logistic-regression model to compute an RUL-based risk index for the utility system asset based on tripping frequencies of the filtered SPRT alarms (step 212). If the risk index exceeds a risk- index threshold, the system generates a notification indicating that the utility system asset needs to be replaced (step 214). Finally, the system periodically updates the logistic-regression model and the irrelevance filter based on time-series signals from additional utility system assets that have failed (step 216). [058] FIG.
• FIG. 3 presents a flow chart illustrating a process for training an inferential model in accordance with the disclosed embodiments.
• the system receives an inferential training set of time-series signals gathered from sensors in the utility system asset during normal fault-free operation (step 302).
• the system trains the inferential model to predict values of the time-series signals based on the inferential training set (step 304).
• FIG. 4 presents a flow chart illustrating a process for training a logistic-regression model to predict an RUL for an asset and for configuring an associated irrelevance filter in accordance with the disclosed embodiments.
• the system receives an RUL training set comprising time-series signals gathered from sensors in similar utility system assets while the similar utility system assets are run to failure (step 402).
• the system also receives associated failure times for the similar utility system assets (step 404).
• the process for determining which utility system assets are similar can involve automatically clustering the assets to form clusters comprising “like makes/models,” either from a list of asset makes/models, or empirically based on the numbers and types of internal sensors.
• the system uses the inferential model to generate estimated values for the RUL training set of time-series signals (step 406).
• the system then performs a pairwise differencing operation between actual values and the estimated values for the RUL training set of time-series signals to produce residuals (step 408).
• the system performs a SPRT on the residuals to produce SPRT alarms with associated tripping frequencies (step 410).
• the system then trains the logistic-regression model to predict an RUL for the utility system asset based on correlations between the SPRT alarm tripping frequencies and the failure times for the similar utility system assets (step 412).
• the system identifies relevant SPRT alarms that were generated during a time interval before a utility system asset failed (step 414), and then configures the irrelevance filter to remove SPRT alarms that are not relevant (step 416).

Landscapes

• Engineering & Computer Science (AREA)
• Business, Economics & Management (AREA)
• Physics & Mathematics (AREA)
• Human Resources & Organizations (AREA)
• Theoretical Computer Science (AREA)
• General Physics & Mathematics (AREA)
• Economics (AREA)
• Strategic Management (AREA)
• Tourism & Hospitality (AREA)
• General Business, Economics & Management (AREA)
• Marketing (AREA)
• Data Mining & Analysis (AREA)
• Operations Research (AREA)
• Entrepreneurship & Innovation (AREA)
• Mathematical Physics (AREA)
• Quality & Reliability (AREA)
• Health & Medical Sciences (AREA)
• General Engineering & Computer Science (AREA)
• Mathematical Analysis (AREA)
• Pure & Applied Mathematics (AREA)
• Mathematical Optimization (AREA)
• Computational Mathematics (AREA)
• Software Systems (AREA)
• Development Economics (AREA)
• Game Theory and Decision Science (AREA)
• Primary Health Care (AREA)
• General Health & Medical Sciences (AREA)
• Water Supply & Treatment (AREA)
• Public Health (AREA)
• Computing Systems (AREA)
• Evolutionary Biology (AREA)
• Evolutionary Computation (AREA)
• Probability & Statistics with Applications (AREA)
• Artificial Intelligence (AREA)
• Algebra (AREA)
• Databases & Information Systems (AREA)
• Computational Linguistics (AREA)
• Educational Administration (AREA)
• Bioinformatics & Computational Biology (AREA)
• Bioinformatics & Cheminformatics (AREA)

Priority Applications (4)

Applications Claiming Priority (2)

Publications (1)

Family

ID=71728879

Family Applications (1)

Country Status (5)

Cited By (1)

Families Citing this family (6)

Citations (1)

Family Cites Families (70)

• 2019
  • 2019-09-04 US US16/560,629 patent/US11341588B2/en active Active
• 2020
  • 2020-06-25 JP JP2022514593A patent/JP7578678B2/ja active Active
  • 2020-06-25 WO PCT/US2020/039630 patent/WO2021045835A1/en not_active Ceased
  • 2020-06-25 CN CN202080052760.0A patent/CN114175072A/zh active Pending
  • 2020-06-25 EP EP20743406.9A patent/EP4026024B1/en active Active
• 2022
  • 2022-05-11 US US17/741,709 patent/US12039619B2/en active Active
• 2024
  • 2024-05-16 US US18/666,387 patent/US20240303754A1/en active Pending
  • 2024-10-23 JP JP2024186615A patent/JP2025036409A/ja active Pending

Patent Citations (1)

Non-Patent Citations (3)

Also Published As

Similar Documents

Legal Events