Method and system for monitoring and / or operating power system assets - Patents.com
A hybrid data-driven and model-based system for power system assets addresses the limitations of traditional methods by systematically capturing long-term trends and reducing human intervention, enhancing maintenance efficiency and accuracy.
Patent Information
- Application Number
- JP2023526611
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2020-11-06
- Filing Date
- 2021-11-05
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2041-11-05
AI Technical Summary
Traditional model-based approaches for monitoring power system assets fail to systematically capture long-term trends and require significant human intervention, leading to inefficiencies and errors in maintaining and operating these assets.
A hybrid data-driven and model-based system that periodically updates parameter values using sensors to establish a parameter evolution model, allowing for systematic trend analysis and reduced human involvement.
This system provides accurate, automated monitoring and forecasting of power system asset conditions, enabling better maintenance planning and reducing human error by capturing long-term trends and predicting future performance.
Smart Images

Figure 0007580879000009 
Figure 0007580879000010 
Figure 0007580879000011
Abstract
Description
[Technical field]
[0001] FIELD OF THEINVENTION The present invention relates to a method and system for monitoring and / or operating an electric power system asset.The present invention relates to a method and system for monitoring and / or operating an electric power system asset that uses a model of the electric power system asset. [Background technology]
[0002] 2. Background of the Invention Electric power generation, transmission, and distribution systems include multiple power system assets, such as rechargeable energy storage systems, transformers, generators (e.g., distributed energy resources (DERs)), etc. It is important to monitor the condition of such power system assets to ensure reliable and stable system operation, trigger maintenance or human inspection tasks, and / or establish the operating points of the power system assets.
[0003] Techniques for monitoring and / or operating power system assets may be model-based. To illustrate, models of power system assets may have parameterization, and parameter values for various parameters may be established based on measurements captured during field operations.
[0004] WO 2015 / 160745 discloses a method for predicting the performance of an electrical energy storage system, the method comprising the steps of receiving as input a time series of data points relating to a plurality of operating parameters of the electrical energy storage system. The input time series is provided to a model simulating the dynamics of the electrical energy storage system. A plurality of performance indicators of the electrical energy storage system can be calculated based on the simulated dynamics of the electrical energy storage system.
[0005] U.S. Patent Application Publication No. 2007 / 0194756 discloses a method of monitoring a battery that includes measuring an open circuit voltage on the battery to determine a measured state of charge after at least one use, obtaining a calculated state of charge based on the state of charge before at least one use and discharging the battery during at least one use, and determining a state of the battery based at least in part on a difference between the calculated state of charge and the measured state of charge.
[0006] EP 2 337 184 A1 discloses a grid-connected energy storage system and a method for controlling the system that allows increasing the efficiency of the converter.
[0007] Traditional model-based approaches for monitoring power system assets have various drawbacks. To illustrate, the state of a power system asset may no longer be adequately reflected by a power system asset model if the state of the power system asset significantly degrades in the time interval that has elapsed since the parameter values of the power system asset model were established. Iteratively redetermining the parameter values can alleviate this problem, but does not systematically capture long-term trends in the evolution of the power system asset state. Trend analysis is difficult when using traditional model-based approaches for monitoring power system assets, especially when considering that there is a vast number of possible operating conditions. Traditional model-based approaches often require significant involvement of human engineers, e.g., for the configuration process, which makes these approaches prone to error. Summary of the Invention [Problem to be solved by the invention]
[0008] overview There is a need in the art for improved methods and systems that mitigate at least some of the shortcomings of conventional model-based techniques for monitoring and / or operating power system assets. In particular, there is a need in the art for methods and systems that enable systematic identification of trends, facilitate trend analysis, and / or reduce the involvement of human experts. [Means for solving the problem]
[0009] According to an embodiment of the invention, parameter values of the power system asset model are determined. This is done iteratively, e.g., periodically. A parameter evolution model is established that determines the evolution of one, some, or all of the parameter values of the power system asset model. This can include determining a set of second model parameter values for parameters of the parameter evolution model that define how one, some, or all of the parameter values of the power system asset model evolve.
[0010] The parameter evolution model is separate from the power system asset model and can be used to systematically capture and quantify trends in the time evolution of parameter values of the power system asset model.
[0011] Determining the parameter values of the power system asset model may be performed using a data-driven approach, for example, based on measurements. The measurements may be captured by one or several sensors, including sensors mounted in or on the power system asset.
[0012] A second set of model parameter values for the parameters of the parameter evolution model that defines how one, some, or all of the parameter values of the power system asset model evolve depends directly or indirectly on the captured measurements. By way of example, the parameter evolution model can be trained using as input a set of first model parameter values, each set being determined for the power system asset model using the measurements, each set being associated with a different time or time interval.
[0013] Thus, methods and systems according to embodiments provide a hybrid data-driven and model-based power system asset monitoring system that provides end users with insight into the actual current state of power system assets and can predict their future state.
[0014] In the trend analysis, multiple different operating scenarios (e.g., multiple different hypothetical future control actions) can be accommodated. By way of example, the parameter evolution model can be used to predict the future evolution of the power system asset model for one or several operating scenarios.
[0015] Control signals and / or control parameters can be automatically set in the control unit to track operating scenarios, including temperature control, protection systems (e.g., circuit breaker status), charge and discharge rates, etc.
[0016] A method for monitoring and / or operating a power system asset according to one embodiment includes receiving measurements indicative of a state of the power system asset from at least one data acquisition unit. The method includes processing the received measurements to determine a sequence of a first set of model parameter values for a power system asset model, each of the first set of model parameter values being determined for a different time or time interval. The method includes using the sequence of the first set of model parameter values to determine a second set of model parameter values for a parameter evolution model different from the power system asset model, the parameter evolution model describing the evolution of one, some, or all of the first model parameter values of the power system asset model. The method includes generating and providing an output in response to at least one of the first set of model parameter values for the power system asset model and the second set of model parameter values for the parameter evolution model.
[0017] The parameter evolution model may describe the evolution of one, some, or all of the first model parameter values of the power system asset model as a function of operating and / or ambient conditions.
[0018] Generating the output may include performing an impairment diagnosis using the parameter evolution model. The output may be generated as a function of a result of the impairment diagnosis.
[0019] The method may further include using the parameter evolution model to generate a forecast of future evolution of one, some, or all of the first model parameter values of the power system asset model.
[0020] The forecasts can be used to generate outputs that can include at least one of a remaining useful life (RUL), a future power system asset state, and a future performance of the power system assets, determined based on the forecasted future developments.
[0021] The parameter evolution model may describe the evolution of one, some, or all of the first model parameter values of the power system asset model as a function of ambient and / or operating conditions.
[0022] The method may include intermittently adjusting ambient and / or operating conditions of the power system assets to determine a second set of model parameter values for the adjusted ambient and / or operating conditions.
[0023] Intermittently adjusting the ambient and / or operating conditions may include adjusting the ambient and / or operating conditions in response to user input.
[0024] Intermittently adjusting the ambient and / or operating conditions may include issuing a command to the control unit to adjust the ambient and / or operating conditions, and after a set of at least two or at least three first model parameter values have been determined for the adjusted ambient and / or operating conditions, issuing a command to the control unit to return the ambient and / or operating conditions to their previous values.
[0025] Intermittently adjusting the ambient and / or operating conditions may include, but is not limited to, one or more of: performing temperature control, controlling at least one protection system (e.g., circuit breaker state), controlling at least one switch state, controlling a charge rate, and controlling a discharge rate.
[0026] A set of second model parameter values can be determined for parameterization of the parameter evolution model (which can take the form of coefficients of a linear or non-linear regression model or a machine learning model) for each of a plurality of ambient and / or operating conditions, respectively. This can involve intermittently adjusting the ambient and / or operating condition to a particular value and determining a set of second model parameter values associated with that ambient and / or operating condition, respectively. Adjusting and determining the set of second model parameter values can be repeated until parameter values for parameterization of the parameter evolution model are determined for several different ambient and / or operating conditions.
[0027] The method may include using the parameter evolution model to generate a number of forecasts of future evolution of one, some, or all of the first model parameter values of the power system asset model, the several forecasts being established for a number of different future ambient and / or operating conditions.
[0028] The method may further include receiving user input specifying at least a subset of the future ambient and / or operating conditions, and / or automatically generating at least a subset of the future ambient and / or operating conditions.
[0029] The method may further include using the power system asset model to perform a diagnosis of events occurring on the first time scale.
[0030] The method may further include using the parameter evolution model to identify trends occurring at a second time scale that is greater than the first time scale.
[0031] The first time scale may be equal to, smaller than or similar to the sampling interval over which measurements are received or the time interval over which the first set of model parameter values are determined, respectively.
[0032] At least one of the multiple sets of first model parameter values of the power system asset model may be used in combination with the measurements to calculate a current power system asset state, a residual between the measurements and an observable estimated from the current power system asset state may be calculated, and a diagnosis of an event occurring on a first time scale may be performed based on the residual.
[0033] The output may include one or more of information output via a human machine interface (HMI) and / or control signals output to a controller.
[0034] The output may include power system asset health information. The output may include an alarm or warning.
[0035] The control signal may be output to a microgrid controller. The first set of model parameter values and the second set of model parameter values of the parameter evolution model may be determined separately for each of the multiple batteries or battery strings.
[0036] Determining the first set of model parameter values may include solving a discrete-time linear differential equation.
[0037] Determining the second set of model parameter values may include at least one of machine learning, fitting, or regression analysis, in particular linear regression.
[0038] The computing system comprises an interface operative to receive measurements indicative of a state of the energy storage system and at least one integrated circuit coupled to the interface, the at least one integrated circuit operative to process the received measurements to determine a sequence of a first set of model parameter values for a power system asset model, each of the first set of model parameter values being determined for a different time or time interval, determine a second set of model parameter values for a parameter evolution model different from the power system asset model using the sequence of the first set of model parameter values, the parameter evolution model determining the second set of model parameter values describing an evolution of one, some or all of the first model parameter values of the power system asset model, and generate and provide an output in response to at least one of the first set of model parameter values for the power system asset model and the second set of model parameter values for the parameter evolution model.
[0039] The computing system is operable to perform a method according to one embodiment.
[0040] The system comprises a power system asset, a data acquisition unit operative to collect measurements indicative of a condition of the power system asset, and a computing system according to an embodiment coupled to the data acquisition unit.
[0041] The power system assets may be, but are not limited to, transformers, distributed energy resources (DERs), and rechargeable energy storage systems. By determining a second set of model parameter values of the parameter evolution model, the long-term trends of the power system asset state evolution can be systematically determined. The first set of model parameter values can be stored and evaluated to observe longer-term degradation in the system as reflected by the parameter evolution model.
[0042] A method for monitoring and / or operating a rechargeable energy storage system (ESS) according to one embodiment includes receiving measurements indicative of a condition of the ESS from at least one data acquisition unit. The method includes processing the received measurements to determine a sequence of a first set of model parameter values for an ESS model, each of the first set of model parameter values being determined for a different time or time interval. The method includes using the sequence of the first set of model parameter values to determine a second set of model parameter values for a parameter evolution model different from the ESS model, the parameter evolution model describing the evolution of one, some or all of the first model parameter values of the ESS model. The method includes generating and providing an output in response to at least one of the first set of model parameter values for the ESS model and the second set of model parameter values for the parameter evolution model.
[0043] The parameter evolution model may describe the evolution of one, some, or all of the first model parameter values of the ESS model as a function of operating and / or ambient conditions.
[0044] Generating the output may include performing an impairment diagnosis using the parameter evolution model. The output may be generated as a function of a result of the impairment diagnosis.
[0045] The method may further include using the parameter evolution model to generate a forecast of future evolution of one, some or all of the first model parameter values of the ESS model.
[0046] The forecast can be used to generate an output that can include at least one of a remaining useful life (RUL), a future ESS condition, and a future performance of the ESS, determined based on the predicted future developments.
[0047] The parameter evolution model may describe the evolution of one, some, or all of the first model parameter values of the ESS model as a function of ambient and / or operating conditions.
[0048] The method may include intermittently adjusting ambient and / or operating conditions of the ESS to determine a second set of model parameter values for the adjusted ambient and / or operating conditions.
[0049] Intermittently adjusting the ambient and / or operating conditions may include adjusting the ambient and / or operating conditions in response to user input.
[0050] Intermittently adjusting the ambient and / or operating conditions may include issuing a command to the control unit to adjust the ambient and / or operating conditions, and after a set of at least two or at least three first model parameter values have been determined for the adjusted ambient and / or operating conditions, issuing a command to the control unit to return the ambient and / or operating conditions to their previous values.
[0051] Intermittently adjusting the ambient and / or operating conditions may include, but is not limited to, one or more of: performing temperature control (e.g., including controlling the temperature of an enclosure in which the ESS is mounted), controlling at least one protection system (e.g., a circuit breaker state), controlling at least one switch state, controlling a charge rate, and controlling a discharge rate.
[0052] A set of second model parameter values can be determined for parameterization of the parameter evolution model (which can take the form of coefficients of a linear or non-linear regression model or a machine learning model) for each of a plurality of ambient and / or operating conditions, respectively. This can involve intermittently adjusting the ambient and / or operating condition to a particular value and determining a set of second model parameter values associated with that ambient and / or operating condition, respectively. Adjusting and determining the set of second model parameter values can be repeated until parameter values for parameterization of the parameter evolution model are determined for several different ambient and / or operating conditions.
[0053] The method may include using the parameter evolution model to generate a number of forecasts of future evolution of one, some, or all of the first model parameter values of the ESS model, the several forecasts being established for a number of different future ambient and / or operating conditions.
[0054] The method may further include receiving user input specifying at least a subset of the future ambient and / or operating conditions, and / or automatically generating at least a subset of the future ambient and / or operating conditions.
[0055] The method may further include using the ESS model to perform a diagnosis of an event occurring on the first time scale.
[0056] The method may further include using the parameter evolution model to identify trends occurring at a second time scale that is greater than the first time scale.
[0057] The first time scale may be equal to, smaller than or similar to the sampling interval over which measurements are received or the time interval over which the first set of model parameter values are determined, respectively.
[0058] At least one of the multiple sets of first model parameter values of the ESS model may be used in combination with the measurement values to calculate a current ESS state, a residual between the measurement values and an observable estimated from the current ESS state may be calculated, and a diagnosis of an event occurring on a first time scale may be performed based on the residual.
[0059] The output may include one or more of information output via a human machine interface (HMI) and / or control signals output to a controller.
[0060] The output may include ESS health information. The output may include an alarm or warning.
[0061] The control signal may be output to a battery management system (BMS) or a microgrid controller.
[0062] The control signal may control at least one of a charge rate, a discharge rate, an ESS ambient temperature, a discharge depth, and a protection system.
[0063] The ESS may be an electrochemical ESS, in particular a rechargeable Battery Energy Storage System (BESS).
[0064] A BESS may include multiple batteries or multiple battery strings. The battery may be a lithium ion battery.
[0065] The first set of model parameter values and the second set of model parameter values of the parameter evolution model may be determined separately for each of the multiple batteries or battery strings.
[0066] Determining the first set of model parameter values may include solving a discrete-time linear differential equation.
[0067] Determining the second set of model parameter values may include at least one of machine learning, fitting, or regression analysis, in particular linear regression.
[0068] The first set of model parameter values may include resistance and / or capacitance and / or inductance.
[0069] The first set of model parameter values may include a parameter value for self-discharge resistance. The first set of model parameter values may include parameter values of a first RC element for modeling charging and discharging hysteresis.
[0070] The first set of model parameter values may include parameter values of a second RC element for modeling a long term transient phenomenon.
[0071] The first set of model parameter values may include parameter values of a third RC element for modeling short-term transients.
[0072] The computing system comprises an interface operative to receive measurements indicative of a state of an energy storage system (ESS) and at least one integrated circuit coupled to the interface. The at least one integrated circuit is operative to process the received measurements to determine a sequence of a set of first model parameter values of an ESS model, each of the sets of first model parameter values being determined for a different time or time interval. The at least one integrated circuit is operative to use the sequence of the sets of first model parameter values to determine a set of second model parameter values of a parameter evolution model different from the ESS model. The parameter evolution model describes the evolution of one, some, or all of the first model parameter values of the ESS model. The at least one integrated circuit is operative to generate and provide an output in response to at least one of the first model parameter values of the ESS model and the second model parameter values of the parameter evolution model.
[0073] The parameter evolution model may describe the evolution of one, some, or all of the first model parameter values of the ESS model as a function of operating and / or ambient conditions.
[0074] The computing system can be operative such that generating the output can include performing an impairment diagnosis using the parameter evolution model. The computing system can be operative such that the output can be generated as a function of a result of the impairment diagnosis.
[0075] The computing system may be operable to use the parameter evolution model to generate a forecast of future evolution of one, some, or all of the first model parameter values of the ESS model.
[0076] The computing system may be operable to use the forecast to generate an output that may include at least one of a remaining useful life (RUL), a future ESS condition, and a future performance of the ESS, determined based on the forecasted future developments.
[0077] The parameter evolution model may describe the evolution of one, some, or all of the first model parameter values of the ESS model as a function of ambient and / or operating conditions.
[0078] The computing system may be operable to intermittently adjust ambient and / or operating conditions of the ESS and determine a second set of model parameter values for the adjusted ambient and / or operating conditions.
[0079] The computing system may be operable such that intermittently adjusting the ambient and / or operating conditions may include adjusting the ambient and / or operating conditions in response to user input.
[0080] The computing system may be operable such that intermittently adjusting the ambient and / or operating conditions may include issuing a command to the control unit to adjust the ambient and / or operating conditions, and after a set of at least two or at least three first model parameter values have been determined for the adjusted ambient and / or operating conditions, issuing a command to the control unit to return the ambient and / or operating conditions to their previous values.
[0081] The computing system may be operable to intermittently adjust ambient and / or operating conditions to include, but are not limited to, one or more of: performing temperature control (e.g., including controlling the temperature of an enclosure in which the ESS is mounted), controlling at least one protection system (e.g., circuit breaker state), controlling at least one switch state, controlling a charge rate, and controlling a discharge rate.
[0082] The computing system may be operable to determine, for each of a plurality of ambient and / or operating conditions, a respective set of second model parameter values for parameterization of the parameter evolution model (which may take the form of coefficients of a linear or nonlinear regression model or a machine learning model). This may involve intermittently adjusting the ambient and / or operating condition to a particular value and determining a set of second model parameter values associated with that ambient and / or operating condition, respectively. The adjustment and determination of the set of second model parameter values may be repeated until parameter values for the parameterization of the parameter evolution model have been determined for a number of different ambient and / or operating conditions.
[0083] The computing system may be operable to use the parameter evolution model to generate a number of forecasts of future evolution of one, some, or all of the first model parameter values of the ESS model, the several forecasts being established for a number of different future ambient and / or operating conditions.
[0084] The computing system may be operable to receive user input specifying at least a subset of the future ambient and / or operating conditions and / or to automatically generate at least a subset of the future ambient and / or operating conditions.
[0085] The computing system is operable to use the ESS model to perform a diagnosis of an event occurring on a first time scale.
[0086] The computing system may be operable to use the parameter evolution model to identify trends occurring at a second time scale that is greater than the first time scale.
[0087] The first time scale may be equal to, smaller than or similar to the sampling interval over which measurements are received or the time interval over which the first set of model parameter values are determined, respectively.
[0088] The computing system is operable to use at least one of a plurality of sets of first model parameter values of the ESS model in combination with measurements to calculate a current ESS state, calculate a residual between the measurements and an observable estimated from the current ESS state, and perform a diagnosis of an event occurring at a first time scale based on the residual.
[0089] The computing system may be operable such that the output includes one or more of information output via a human machine interface (HMI) and / or control signals output to a controller.
[0090] The computing system may be operable such that the output includes ESS health information.
[0091] The computing system may be operable such that the output includes an alarm or warning.
[0092] The computing system may be operable such that the control signal is output to a battery management system (BMS) or a microgrid controller.
[0093] The computing system is operable such that the control signal controls at least one of a charge rate, a discharge rate, an ESS ambient temperature, a discharge depth, and a protection system.
[0094] The ESS may be an electrochemical ESS, in particular a rechargeable Battery Energy Storage System (BESS).
[0095] A BESS may include multiple batteries or multiple battery strings. The battery may be a lithium ion battery.
[0096] The computing system may be operable such that a first set of model parameter values and a second set of model parameter values of the parameter evolution model may be determined separately for each of a plurality of batteries or battery strings.
[0097] The computing system may be operable such that determining the first set of model parameter values includes solving a discrete-time linear differential equation.
[0098] The computing system may be operable such that determining the second set of model parameter values may include at least one of machine learning, fitting, or regression analysis, in particular linear regression.
[0099] The computing system is operable to determine a set of first model parameter values including resistance and / or capacitance and / or inductance.
[0100] The computing system is operable to determine a first set of model parameter values that includes a parameter value for self-discharge resistance.
[0101] The computing system is operable to determine a first set of model parameter values including parameter values of a first RC element for modeling charging and discharging hysteresis.
[0102] The computing system is operable to determine a first set of model parameter values including parameter values of a second RC element for modeling a long term transient phenomenon.
[0103] The computing system is operable to determine a first set of model parameter values including parameter values of a third RC element for modeling a short-term transient phenomenon.
[0104] The system comprises a battery energy storage system (BESS), a data acquisition unit operative to collect measurements indicative of a condition of the BESS, and a computing system according to an embodiment coupled to the data acquisition unit.
[0105] The system may include an enclosure in which the BESS is mounted. The computing system may be mounted within an enclosure.
[0106] The system may include a battery management system mounted within the enclosure.
[0107] When a data-driven model is implemented that describes the evolution of parameter values(s) of the power system asset model with respect to operating and / or ambient conditions, it is possible to assess the impact of various operating scenarios and / or ambient conditions on both performance and remaining useful life.
[0108] By combining the power system asset model with the parameter evolution model of the power system, diagnosis, fault prediction, and optimization can be performed. Thus, the system and method can be quickly and easily deployed with higher accuracy, limiting the risk of human error due to incorrect parameter setting, and reducing system maintenance.
[0109] The methods and systems may be used in connection with, but are not limited to, rechargeable energy storage systems (particularly BESS), transformers, energy generation systems, or other power system assets in power generation, transmission and / or distribution systems.
[0110] BRIEF DESCRIPTION OF THE DRAWINGS The subject matter of the present invention will now be explained in more detail with reference to preferred exemplary embodiments illustrated in the accompanying drawings. [Brief description of the drawings]
[0111] [Figure 1] FIG. 1 is a schematic diagram of a system according to one embodiment. [Diagram 2] FIG. 1 is a schematic diagram of a computing system according to one embodiment. [Diagram 3] FIG. 1 is a schematic diagram of a system according to one embodiment. [Figure 4] 1 is a flowchart of a method according to one embodiment. [Diagram 5] 1 is a graph illustrating the operation of a method and computing system according to one embodiment. [Figure 6] 1 is a flowchart of a method according to one embodiment. [Figure 7] FIG. 1 is a circuit diagram of a battery energy storage system (BESS) model. [Figure 8] 1 is a flowchart of a method according to one embodiment. [Figure 9] 1 is a flowchart of a method according to one embodiment. [Figure 10]1 is a graph illustrating identification of events on a first time scale based on measurements and a power system asset model. [Figure 11] 1 is a flowchart of a method according to one embodiment. [Figure 12] 1 is a flowchart of a method according to one embodiment. [Figure 13] 1 is a flowchart of a method according to one embodiment. [Figure 14] 1 is a flowchart of a method according to one embodiment. [Figure 15] FIG. 1 is a schematic diagram of a system according to one embodiment. [Figure 16] FIG. 1 is a schematic diagram of a system according to one embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0112] Detailed Description of the Embodiments Exemplary embodiments of the invention are described with reference to the drawings, in which identical or similar reference numbers indicate identical or similar elements. Some embodiments are described in the context of a battery energy storage system (BESS), although the invention is not limited thereto. Although embodiments of the invention are applicable to energy storage systems, each based on a lithium-ion battery, the invention is not limited thereto and may be applied to other types of BESS or other power system assets of power generation, transmission and / or distribution systems.
[0113] Features of the embodiments can be combined with each other unless otherwise stated. The same or similar reference numbers are used in the drawings to denote the same or similar elements.
[0114] 1 is a schematic representation of a system 10. The system 10 comprises a rechargeable battery energy storage system (BESS) 11. The rechargeable BESS may comprise multiple parallel battery strings. Each of the multiple parallel battery strings may comprise multiple batteries. The multiple parallel batteries may be coupled to a common coupling point via a converter and / or transformer. The battery strings may each include at least one series connection of batteries, e.g., lithium ion batteries. Some or all of the strings may include several series connections of batteries coupled in parallel with each other.
[0115] The BESS 11 may be coupled to a grid at a point of common coupling. The grid may be a national or regional power grid. The grid may be an island power grid, such as a microgrid.
[0116] As used herein, the term "battery" includes rechargeable batteries. The BESS 11 may operate to provide ancillary services to the grid, such as providing frequency stabilization, load regulation, and / or voltage support. BESS 11 operating to provide such ancillary services are often sized to accommodate peak demand, but operate at a much lower power or load than it is designed for throughout most of its operation. The techniques detailed herein are particularly suited to such BESSs that must be able to provide certain performance characteristics, e.g., due to contractual obligations, or must be able to accommodate worst case scenarios, but are typically operated at a much lower load than the rated load for which they are designed.
[0117] The BESS 11 is equipped to provide measurements indicative of its operating conditions (such as voltage, current and / or temperature) using one or more measurement devices (e.g. sensors). The outputs of the one or more measurement devices may be connected to a data acquisition unit 12.
[0118] The data acquisition unit 12 may be or include, but is not limited to, an analog-to-digital converter, a merging unit, or a gateway. The data acquisition unit 12 may be provided with a first setting P1 that defines the process of converting an analog waveform into a discrete signal, specifically the sampling rate and length of the signal that undergoes the conversion. The first setting P1 may be fixed or may be specified by a user. The first setting P1 may include multiple numerical values.
[0119] The data acquisition unit 12 is connected to a computing system 13. The computing system 13 comprises a data processing unit 14 and a communication module 17. Implemented within the data processing unit 14 are a data storage module 15 and a hybrid data-driven and model-based battery monitoring system module 16, which will be described in more detail below.
[0120] The data processing unit 14 may also execute additional processes or modules, such as threshold comparisons for comparing current or forecast conditions of the BESS that trigger the generation of an alarm, warning, or other output, and / or modules for generating and outputting control signals to the BESS 11 and / or devices operatively coupled to the BESS 11 (e.g., a protection system, a temperature control system, etc.).
[0121] The data processing unit 14 may use or act according to the settings P2. The settings P2 may include or consist of user input. The settings P2 may be provided to the computing system 13. The settings P2 may be stored in a memory for use in subsequent steps. The computing system 13 may comprise memory such as a read access memory (RAM) and a read only memory (ROM).
[0122] The computing system 13 may comprise or be connected to an output unit 18 capable of presenting the results of the condition monitoring to a user. The output unit 18 may comprise an optical output unit, such as a monitor, a printer, or any useful device for presenting the results of the condition monitoring as described in more detail herein. Alternatively or in addition to an output unit capable of operating as a user interface, the condition monitoring may trigger the generation of a control signal that may be output via the communication module 17 to a device of the BESS 11 or the power system associated therewith.
[0123] The data acquisition unit 12 may take the form of a dedicated data acquisition system or may form part of another device having data acquisition functionality, for example a battery management system (BMS) or a microgrid controller.
[0124] The computing system 13 may take the form of a dedicated computer system, or may form part of another device with data acquisition capabilities, for example a battery management system (BMS) or a microgrid controller.
[0125] 2 is a schematic diagram of a computing system 13. The computing system 13 comprises an interface 31 operable to receive measurements (such as voltage, current and / or temperature of the BESS 11). The computing system 13 comprises one or several integrated circuit(s) (IC(s)) 14.
[0126] The IC(s) 14 may include one or more of a processor, a microprocessor, a controller, a microcontroller, an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or any combination thereof.
[0127] The IC(s) 14 may be operable to execute a first parameter determination module 32. The first determination module 32 operates to determine a set of parameter values of a power system asset model (e.g., a BESS model) as a function of measurements received at the interface 31.
[0128] The parameterization of the power system asset model may be fixed. By way of example, the power system asset model may include two, three, four, or more than four parameters. The parameters may represent characteristics of the electrical circuit model (e.g., resistance, capacitance, etc.), mechanical properties, chemical properties (e.g., gas concentration of insulating oil, etc.), or combinations thereof. The power system asset model may be predefined, and the determination module 32 determines some or all of the parameter values of the power system asset model.
[0129] The measurements may be received at the interface 31 at sampling times. The first determination module 32 may operate to process the measurements (e.g., voltages and currents) measured in a time series and / or received at the interface 31 at the time series sampling times into a set of first model parameter values. By way of example, the first determination module 32 may operate to determine a set of first model parameter values for parameters of the power system asset model by solving a time-discrete differential equation. Measurements (e.g., voltages and currents) associated with at least two, at least three, or four or more consecutive sampling times may be processed to determine a set of first model parameter values, respectively.
[0130] The first determination module 32 iteratively repeats the determination of the set of first model parameter values of the power system asset model. The determination may be repeated periodically or aperiodically, for example in an event-driven manner. The measurements (e.g., voltage and current) received at R≧2 time series sampling times may be processed into one set of first model parameter values of the power system asset model. The measurements (e.g., voltage and current) received at R time series sampling times may be processed into another set of first model parameter values of the power system asset model. This may continue iteratively. Thus, all sets of first model parameter values of the power system asset model are associated with a time or time interval.
[0131] Each set of first model parameter values may include a number of parameter values, n≧2, for the n parameters of the power system asset model. The number of parameter values in each set of first model parameter values depends on the parameterization of the power system asset model. One or some of the parameter values may be defined by user input, in which case these parameter values may, but need not, be updated based on measurements. By way of example, a nominal battery capacitance or an initial value of the battery capacitance may be specified by input received from a user. The number of parameter values may be n≧3, n≧4, n≧5, or more, depending on the complexity of the power system asset model.
[0132] The parameter values in the different sets are assigned to the same parameters but may vary because the power system asset state, its operating conditions, and / or its ambient conditions may change. Computing system 13 is operative to quantify the evolution of one, some, or all of the parameter values included in the first model parameter value set of the power system asset model.
[0133] The IC(s) 14 may execute a second determination module 33. The second determination module 33 may determine a set of second model parameter values of a parameter evolution model different from the power system asset model. The parameter evolution model describes the evolution of one, some, or all of the first model parameter values of the ESS model.
[0134] The second determination module 33 may determine the second set of model parameter values using various techniques. By way of example, the second determination module 33 may perform a regression analysis on one, some, or all of the parameter values included in the series of the first set of model parameter values of the power system asset model. The second set of model parameter values may be determined by at least one of machine learning, fitting, or regression analysis, in particular linear regression, using some or all of the determined set of first model parameter values as inputs.
[0135] A second set of model parameter values (which may be coefficients of a linear regression analysis, weights of a neural network, etc.) may be established to quantify the time evolution of the parameter values of the power system assets.
[0136] The IC(s) 14 may execute an output generation module 34. The output generation module 34 may generate an output based on the sequence of the first set of model parameter values and the second set of model parameter values. The output may be based on measurements, for example when residuals are calculated.
[0137] The output may include any one or any combination of the following: Information about the current and / or future state of power system assets One or several Key Performance Indicators (KPIs) established for the current and / or future state of power system assets Information regarding the past, present and / or future performance or remaining useful life (RUL) of power system assets Alarms, warnings, or other indicators generated based on the past and / or current and / or future conditions of power system assets. Control signals that are generated based on the past and / or present and / or future states of power system assets.
[0138] The computing system 13 may be provided locally to an enclosure of the power system asset or may be integrated within the enclosure. The computing system 13 may be associated with multiple power system assets. Fleet learning may be utilized by the computing system 13 to determine a set of second model parameter values for the parameter evolution model.
[0139] 3 illustrates a system 40 including a plurality of power system assets 11a-11c (in this case, but not limited to, a plurality of BESSs), a plurality of local control systems 41a-41c each associated with one of the power system assets 11a-11c, and a computing system 13 communicatively coupled to the plurality of local control systems 41a-41c. The computing system 13 includes: a sequence of first model parameter value sets of a power system asset model for each of the assets 11a-11c, each of the first model parameter value sets being determined for a different time or time interval (parameter values may be different for different ones of the assets 11a-11c even when the same parameterization, i.e., the same power system asset model, is used); and at least one set of second model parameter values for the parameter evolution model; The parameter evolution model describes the evolution of first model parameter values of one, some, or all of the power system asset models of the assets 11a-11c.
[0140] 4 is a flow chart of a method 50 according to one embodiment. The method may be performed automatically by the computing system 13.
[0141] In step 51, measurements are received. The measurements may be received at time series sampling times. The measurements represent time series data.
[0142] In step 52, the measurements are processed into a sequence of a set of first model parameter values for the power system asset model. Each of the first model parameter value sets may be determined for a different time or time interval. The different time intervals may overlap, but do not coincide.
[0143] In step 53, the first set of model parameter values is processed to determine the parameter evolution. The parameter evolution can be modeled by a parameter evolution model. The parameter evolution model can model the time evolution of one, some, or all of the parameter values included in the first model parameter values of the power system asset model. The parameterization of the parameter evolution model can be fixed, and the second set of model parameter values can be determined for the fixed parameterization of the parameter evolution mode. Alternatively or additionally, the parameterization of the parameter evolution model can be determined automatically by the computing system (e.g., by using a machine learning model that responds to changes in the machine learning architecture, such as by changing the number of hidden layers and / or nodes in each layer of the artificial neural network).
[0144] In step 54, an output is generated based on the sequence of the first set of model parameter values and the second set of model parameter values. The output may also depend on measurements, for example by using residual techniques. The output may take any one of the forms described elsewhere herein.
[0145] 5 is a diagram illustrating the processing of measurements in one embodiment. Each solid circle along the time axis represents a measurement at a respective sampling time. Measurements (e.g., battery voltage and current) received at one or several time-series (e.g., consecutive) sampling times can be processed into one first set of parameter values.
[0146] The power system asset model is parameterized such that it has n≧2 power system asset model parameters that are established based on measurements. Parameter values for these n power system asset model parameters are established based on measurements at one or R time series sampling times. As shown in FIG. 5 , measurements 61-63 received at two, three or more time series sampling times may be processed into one first set of parameter values associated with a time interval defined by the sampling times of the measurements 61-63. Each first set of parameter values may generally be an ordered n-tuple, denoted as follows:
[0147]
number
[0148] Here, j denotes the time or time interval to which the n-tuple is associated. A first set of parameter values 71(p 1,1 ,...,p 1,n ) is determined from measurements 61-63 at sampling times received within a time interval. Another set of parameter values 72 (p 2,1 ,...,p 2,n ) is determined from measurements at sampling times received within another time interval. k+1,1 ,...,p k+1,n ) is determined from measurements at sampling times received within further time intervals.
[0149] FIG. 5 shows scenarios determined for measured values received during each set of parameter values over a time interval, where the time intervals do not overlap, although they may overlap. For illustration, the measured value 62 may be processed both to determine a first set 71 of model parameter values and to determine another set 72.
[0150] Two or more of the first sets 71-73 of model parameter values can be processed to determine a second set 79 of parameter development model parameter values. The second set 79 of model parameter values may generally be an ordered m-tuple as shown below.
[0151] [Number]
[0152] The number m may be less than n, may be equal to n, or may be greater than n. For illustration, if the parameter development model is based on linear regression and the time development of all n of the first model parameter values is quantified by the parameter development model, then m = n. If the parameter development model is based on linear regression and the time development of only one or only a subset of the n first model parameter values is quantified by the parameter development model, then m < n. When more complex techniques are used (such as non-linear regression models, machine learning where the number of weights between nodes in an artificial neural network exceeds n), the number m of the second model parameter values may exceed the number n of the first model parameter values within each set.
[0153] As shown in FIG. 5, the second set 79 of model parameter values need not be determined based on all available sets 71, 72, 73 of the first model parameter values. The second set 79 of model parameter values may be updated iteratively as new sets of the first model parameter values are determined.
[0154] In an exemplary embodiment, the second set of model parameter values 79 may represent coefficients of a linear or non-linear function of time. It will be appreciated that time may be measured in a discrete manner, such as by a discrete index that may label the time of the first measurements 61 that are processed into the respective first sets of parameter values 71.
[0155] Determining the set of second model parameter values 79 includes determining whether the function F is minimized or maximized. i The second set of model parameter values 79 may include determining
[0156]
number
[0157]
number
[0158]
number
[0159] or
[0160]
number
[0161] Efficient techniques are known for solving the extremum-finding problem of equations (3)-(6), with or without regularization. The second set of model parameter values 79 may be determined based on an iterative numerical technique that identifies a minimum or maximum of the objective function, as described with reference to equations (3)-(6).
[0162] The parameter evolution model, and more specifically the determined set of second model parameter values, can be used in a variety of ways: The second set of model parameter values can be used to quantitatively assess changes in power system asset conditions that occur over a longer, particularly much longer, time period than the time interval over which a new set of first model parameter values 71, 72, 73 are subsequently determined.
[0163] The second set of model parameter values may be used to determine, but is not limited to, a future power system asset state, a remaining useful life (RUL), a future performance, a future performance degradation, a future KPI, a forecast of a future KPI degradation, the second set of model parameter values may be used to provide failure prediction information to a user via a user interface (e.g., by providing an alarm, warning, or failure prediction information to a user or by automatically generating a control signal) and / or to initiate appropriate countermeasures before a performance degradation occurs.
[0164] The set of second model parameter values can be determined for a variety of different operating and / or ambient conditions, not just one. This allows for fault prediction information to be determined for a variety of different operating and / or ambient conditions. To determine the set of second model parameter values for a variety of different operating and / or ambient conditions, the computing system 13 can trigger intermittent changes in the operating and / or ambient conditions (e.g., by causing a temperature control unit in an enclosure of the power system assets to temporarily increase or decrease the temperature), determine at least two sets of first model parameter values while the operating and / or ambient conditions are so adjusted, and based thereon, determine the set of second model parameter values associated with the adjusted operating and / or ambient conditions.
[0165] 6 is a flow chart of a method 60. The method 60 may be performed automatically by a computing system 80.
[0166] In step 81, one or several sets (possibly multiple) of first model parameter values are used to identify events occurring on a first time scale. The first time scale may correspond to a sampling interval, i.e. the time interval between receipt of successive measurements.
[0167] Identifying an event occurring on the first time scale may include calculating an observable based on one or several sets(s) of first model parameter values and comparing the calculated observable with a measurement. A threshold comparison may be performed on the deviation. A deviation between the calculated observable based on the power system asset model and the actual measurement indicates an event occurring on the first time scale. Such a deviation may trigger an immediate action, such as an alarm, a warning, a status information, or an output of a control signal.
[0168] In step 82, the second set of model parameter values is used to determine a forecast(s) of power system asset states over a second time scale that exceeds the first time scale. The second time scale may be longer than the time interval over which the new set of first model parameter values is calculated.
[0169] Step 82 may include further processing. By way of example, the RUL or other KPI may be predictively determined. This may be done for various operating and / or ambient conditions, optionally using a second set of model parameter values determined for the various operating and / or ambient conditions. Suggested control actions to enhance the RUL or slow degradation of the other KPI may be automatically determined (e.g., by identifying optimal operating and / or ambient conditions with respect to mitigating future performance degradation while meeting desired performance of power system assets).
[0170] In step 83, an output is generated and provided based on the events identified over the first time scale and / or the forecast(s) made over a longer second time scale.
[0171] For further illustration, reference is made below to a rechargeable ESS, and in particular a BESS, although it will be understood that the disclosed techniques are applicable to other power system assets as well.
[0172] 7 is an example model circuit diagram 90 of a BESS. The circuit 90 includes an ohmic resistor R series , and four RC elements that model the following:
[0173] Battery: The ability of a battery to store energy is determined by the capacitor C b It is modeled as R b denotes the self-discharge resistance of the battery.
[0174] Hysteresis effect: Hysteresis is caused by the resistor R to introduce a nonlinear difference between the open circuit voltage (OCV) during charging and discharging for the same state of charge (SOC) of the battery. h and capacitance C h The input is modeled as an RC circuit with
[0175] Long-term dynamic element: C p and R p The RC element, including the R, introduces long term transients in the battery.
[0176] Short-term dynamic element: C s and R s The RC element, including the R1, introduces short-term transients in the battery.
[0177] Capacitance C b can be obtained from the BESS nameplate data (battery capacity in kWh). b is C h , C p and C sOther models may be used, particularly less complex models having fewer parameters.
[0178] 8 is a flow chart of the method 100. The method 100 may be performed automatically to determine parameter values for a BESS model.
[0179] In step 101, an approximate solution of the discrete-time model may be determined. This may include solving a linear differential equation. For illustrative purposes, θ may be calculated in response to measured voltages and currents determined for times j=k, k−1, and k−2. 1,...,6 The value of is
[0180]
number
[0181] where V meas [j] denotes the measured voltage at time j, and I meas [j] indicates the measured current.
[0182] In step 102, θ 1,...,6 Once determined, the OCV can be determined therefrom, which may be done using, but is not limited to, the techniques of X. Tang et al., "Li-ion Battery Parameter Estimation for State of Charge," American Control Conference, San Francisco, 2011.
[0183] In step 103, α 1,...,6is determined. The OCV in step 102 is used as one of the variables in a linear differential equation, for example, according to G. Aggoun et al., "SOC Estimation Based on OCV and Online Identification Parameters Of Lithium Ion Batteries with ADALINE," 8th International Conference on Modelling, Identification and Control, Algiers, 2016. For illustrative purposes, α 1,...,6 teeth,
[0184]
number
[0185] It may be determined that: α 1,...,6 From this, the resistance and / or capacitance of the BESS model can be determined. series , R p , C p , R s , and C s Without being limited thereto, the technique of X. Tang et al., "Li-ion Battery Parameter Estimation for State of Charge" (American Control Conference, San Francisco, 2011) may be used.
[0186] The determined set of first model parameter values of the BESS model allows a current battery condition to be estimated, which may include determining a SOC or state of health (SOH).
[0187] For example, the measured terminal voltage can be considered as an input to a BESS model whose parameter values have been previously established. The BESS model can then be used to estimate the current, open circuit voltage, and voltage drop of the battery.
[0188] An unknown input observer can be used to estimate the current of the battery for known terminal voltages and parameters. An exemplary technique that can be used is described in DC Cambron et al., "A Lithium-Ion Battery Current Estimation Technique Using an Unknown Input Observer" (IEEE, 2017).
[0189] 9 is a flow chart of a method 110 that can be used to determine observables of a current BESS state. The method 110 may be performed automatically by the computing system 13. The method 110 can be used to calculate the observables in step 81 of the method of FIG.
[0190] In step 111, a first set of model parameter values for the BESS model is determined. This can be done as described with reference to FIG.
[0191] An unknown input observer procedure is performed to estimate the battery current given the known terminal voltage and BESS model parameters, step 112. Other observables may also be determined.
[0192] The battery state, including the estimated battery current, may be estimated at step 113. Observables included in or derived from the battery state (such as the battery current) may be compared to the measurements to identify sudden events.
[0193] FIG. 10 is a graph illustrating how calculated observables derived from or included in the BESS states can be used to identify a sudden event. FIG. 10 illustrates the measured battery current 121 as a dashed line and the estimated battery current 122 as a solid line. The estimated battery current 122 is computationally determined according to the techniques described with reference to FIGS. 4-9. In certain situations, the measured battery current 121 and the estimated battery current 122 match, but only to a small deviation, but at certain time intervals there is a more significant deviation between the two quantities. Such a significant deviation indicates a sudden event and can trigger the output of an alert, warning, status information, or control signal by the computing system 13.
[0194] The techniques described with reference to Figures 1-10 can be used to systematically assess different long-term behaviors of a BESS or other power system asset. This can be done for various operating and / or ambient conditions. Providing information on forecasts of RUL or other KPIs for various operating and / or ambient conditions, rather than just one particular operating and ambient condition, can aid in the selection of suitable operating and / or ambient conditions that mitigate power system asset degradation.
[0195] The various operating and / or ambient conditions can be investigated in various ways. Information regarding the operating and / or ambient conditions can be received via a user interface, and the computing system 130 can automatically determine a set of second model parameter values for the specified operating and / or ambient conditions. Alternatively or additionally, the computing system 130 can determine a set of second model parameter values for each of the various operating and / or ambient conditions, respectively, and can output information regarding KPIs (such as RULs) and / or suggested control actions for the various operating and / or ambient conditions. These techniques are described in more detail with reference to FIGS. 11 and 12.
[0196] 11 is a flow chart of the method 130. The method 130 may be performed automatically by the computing system 13.
[0197] In step 131, information regarding one or several operating condition(s) and / or ambient condition(s) is received. The information may be received via a user interface. The information regarding the operating condition(s) may specify a charge or discharge current, a depth of discharge, or other information related to the operation of the BESS. The information regarding the ambient condition(s) may specify an ambient temperature within an enclosure in which the BESS is housed.
[0198] In step 132, computing system 13 may optionally generate and output a control signal. The control signal(s) may be generated to at least intermittently adjust the operating and / or ambient conditions according to the information received in step 131. The control signal may be selectively generated only if a second set of model parameter values has not been determined for the operating and / or ambient conditions specified in step 131 or if a second set of model parameter values has not been determined for similar operating and / or ambient conditions as those specified in step 131.
[0199] Control signals may be generated to at least intermittently adjust operational parameters (e.g., charge rate, discharge rate, discharge depth, protection system status) and / or ambient parameters (e.g., temperature and / or air humidity within an enclosure of the power system assets) according to the operating and / or ambient conditions specified in step 131.
[0200] While the operating conditions and / or ambient conditions are adjusted according to the information received in step 131, the measurements may be processed into at least two sets of first model parameter values of the power system asset model associated with two consecutive times or time intervals. A set of second model parameter values of the parameter evolution model may be established for these adjusted operating conditions and / or ambient conditions based on these at least two sets of first model parameter values of the power system asset model.
[0201] When several different operating conditions (e.g., different charge and / or discharge rates, different depths of discharge, different protection system states) and / or ambient conditions (e.g., different temperatures) are specified in step 131, control signals may be generated in step 132 to successively adjust operational and / or ambient parameters of the power system assets according to the different operating conditions and / or ambient conditions specified in step 131. In this manner, for each operating condition and / or ambient condition specified in step 131, at least two sets of first model parameter values of the power system asset model and a second set of model parameter values of the parameter evolution model may be determined.
[0202] In step 133, the second set(s) of model parameter values may be used to calculate a forecast of future developments of power system asset model parameter values or quantities derived therefrom, such as RUL or other KPIs, for each of a variety of operating and / or ambient conditions.
[0203] In step 134, outputs may be generated. The outputs may include information regarding the future development of power system asset model parameter values or quantities derived therefrom, such as RULs or other KPIs, for each of the various operating and / or ambient conditions specified in step 131.
[0204] The computing system 13 can use the forecast to identify the most suitable operating and / or ambient conditions. By way of example, the computing system 13 can automatically process the forecast to determine which operating and / or ambient conditions maximize one or more KPIs, such as RUL. The computing system can generate and output information about these operating and / or ambient conditions and / or provide information about suggested control actions that maximize one or more KPIs.
[0205] 12 is a flow chart of the method 135. The method 135 may be performed automatically by the computing system 13.
[0206] Possible operating and / or ambient conditions for the power system assets may be automatically generated by the computing system at step 136. By way of example, the computing system may automatically generate a set of charge and / or discharge rates, power system asset ambient temperatures, etc.
[0207] In steps 132-134, a set of second model parameter values may be determined for each of the possible operating and / or ambient conditions generated in step 136. The computing system 13 may generate and output control signals for intermittently varying the operating and / or ambient conditions of the power system assets to determine a set of second model parameter values of the parameter evolution model for each of the possible operating and / or ambient conditions generated in step 131. In step 134, the computing system 13 may use the set of second model parameter values associated with the different operating and / or ambient conditions to output information regarding the RUL or other KPIs for each of the ambient and / or operating conditions determined in step 136, or automatically identify the most suitable operating and / or ambient conditions. By way of example, the computing system 13 may automatically process the set of second model parameter values associated with the different operating and / or ambient conditions to determine which of the operating and / or ambient conditions maximizes one or several KPIs, such as the RUL. The computing system may generate and output information regarding these operating and / or ambient conditions and / or may provide information regarding suggested control actions to maximize one or several KPIs.
[0208] 13 is a flow chart of a method 140. The method 140 may be performed automatically by the computing system 13. The method 140 is described in the context of a power system asset that is, but is not limited to, a BESS.
[0209] In step 141, measurements indicative of BESS operating conditions are received. The measurements are processed in steps 142 to 147. A hybrid of data-driven and model-based approaches can be used.
[0210] Parameter estimation is performed in step 142. Measured data from the BESS system is used to identify parameter values of the BESS model that can then be used to determine the system state. As described elsewhere herein, determining parameter values of the BESS model can include determining resistance and capacitance values of a circuit model of the BESS.
[0211] In step 143, the parameter evolution is determined. A mathematical model describing the evolution of the parameter values of the BESS model with respect to external factors (e.g. temperature, time, cycling, load, etc.) can be trained based on the parameters estimated in step 142 and, optionally, additional historical values stored in a database. The mathematical model describing the evolution of the parameter values of the BESS model is also referred to as a parameter evolution model.
[0212] In step 144, model evaluation can be performed. The BESS model and parameter evolution model are utilized to estimate current and future values of battery parameters and battery conditions (e.g., voltage drop including open circuit voltage and current flowing across the battery circuit) based on current and predicted future measurements. Future measurements can be estimated based on various operational scenarios provided by a user or automatically generated by a computing system. Examples of such scenarios include assuming that previous operational trends will continue into the future or defining changes in operation (e.g., temperature control, circuit breaker state, charge / discharge rate, etc.). Once the user selects the desired scenario for execution, the system refers the control unit to the necessary control signals to track the planned scenario.
[0213] In step 145, a descriptive analysis is performed. The descriptive analysis utilizes the current and future states estimated in step 144. Here, the predicted SOH (a measure of remaining capacity) and ohmic resistance (a measure of power fading) can be utilized to estimate the remaining useful life RUL of the battery. The RUL is estimated for all operating scenarios (provided by the user or by the computing system itself). Other KPIs indicative of the BESS performance (e.g. SoC, efficiency, etc.) are also calculated for the current and future states. In step 145, a residual is calculated that compares the output predicted by the BESS model for the current input with the measurements observed in the physical system.
[0214] In step 145, a decision layer procedure is performed. In the decision layer, the outputs from steps 142-145 are processed to calculate warnings, alerts, and / or suggested actions that can be output to a user. Residuals comparing the outputs predicted by the BESS model for the current inputs with the measurements observed in the physical system are used to identify any catastrophic discrepancies between the model and the real system (e.g. due to catastrophic failures). Trends in the evolution of the parameters of the mathematical model of the BESS system are used to establish general degradation in the system (e.g. capacitor or power fading). If the degradation is too great, an alert is flagged to the user, for example based on a threshold comparison. The performance decision layer can be used to calculate the most likely values of key performance indicators (e.g. efficiency) and confidence limits for various operating scenarios. The decision layer for lifetime estimation can be used to calculate the most likely remaining useful life and confidence intervals of the BESS system for various operating scenarios. The output of the decision layer in step 146 is communicated to the end user via a user interface in step 147.
[0215] 14 shows a detailed implementation of the method according to one embodiment. The method may be performed automatically by the computing system 13.
[0216] Measurements indicative of the BESS operating conditions are measured and obtained in step 161. Typical examples of relevant measurements include voltage, current and temperature.
[0217] In step 162, the measurements recorded in step 161 are used to estimate values for the parameters of the BESS model using conventional techniques such as least squares.
[0218] System identification is the starting point, which is achieved by taking the measured voltages and currents as inputs, as described with reference to equations (7) and (8) and FIG. 8. A discrete-time model of a lithium-ion battery can be expressed in the form of a linear difference equation, as described with reference to equation (7). Once the parameters of the vector θ are fitted, the OCV is calculated. The OCV is then used as the input to a second block, where it is used as one of the variables of the linear difference equation, as described with reference to equation (8). Techniques such as those described in X. Tang et al., “Li-ion Battery Parameter Estimation for State of Charge,” in American Control Conference, San Francisco, 2011 and G. Aggoun et al., “SOC Estimation Based on OCV and Online Identification Parameters Of Lithium Ion Batteries with ADALINE,” 8th International Conference on Modelling, Identification and Control, Algiers, 2016, may be used.
[0219] The window of data points (i.e., the vector of sample points for each measured signal) and / or the periodicity of the parameter fitting (i.e., how often parameter estimation is performed) can be adjusted by the user using setting P2.
[0220] In step 163, one or more measured model inputs recorded in step 161 are used as inputs to a model having fitted BESS model parameter values output from step 162 to estimate the current battery state. For example, the measured terminal voltage may be considered as an input to the model. The BESS model then estimates the battery current, open circuit voltage, and voltage drop.
[0221] For example, an unknown input observer can be used to estimate the battery current for known terminal voltages and parameters, as described with reference to Figure 9. The processing technique of DC Cambron et al., "A Lithium-Ion Battery Current Estimation Technique Using an Unknown Input Observer" (IEEE, 2017) can be used.
[0222] One or more measured outputs estimated by the BESS model are compared to corresponding measured values in step 164. For example, referring to Figure 10, the measured current from the BESS can be subtracted from the current estimated by the BESS model to obtain a residual indicative of the difference between the modeled behavior and the true behavior.
[0223] In step 165, the residuals calculated in step 164 are processed to identify catastrophic changes / events in the system that may indicate a catastrophic failure. In an exemplary embodiment, the residuals may be compared to alarm threshold limits set by the user in setting P2. If the residuals exceed either the upper or lower alarm threshold limits, an alarm is passed to the user (step 175). If the residuals are exceeded, a root cause analysis may be performed. This may take the form of an expert system (e.g., if the terminal voltage exceeds the open circuit voltage, the battery may be considered overcharged).
[0224] Steps 161-165 can be considered to represent a procedure for performing a diagnosis of an event occurring on a first, shorter time scale, also referred to as a sudden event.
[0225] The measurements recorded in step 161 and the parameters estimated in step 162 are stored in the historian 159 .
[0226] In step 166, a data-driven model of how the parameters evolve with respect to operating conditions and / or external factors is trained using data from the historian 159. This includes determining a set of second model parameter values for the parameter evolution model. The parameter evolution model can take into account other measurements (e.g., temperature) that are not utilized during parameter identification. Various techniques can be utilized to estimate the evolution of the model parameters with respect to the input parameters, such as neural networks or regression. Parameters of the mathematical model, including calendar and cyclic aging effects, can be fitted using historical data.
[0227] It may be noted that similar to step 162, the window of data points (i.e., the vector of multiple sample points for each measured signal) and / or the periodicity of the parameter fitting (i.e., how often parameter estimation is performed) can be adjusted by the user using setting P2.
[0228] The values and trends of the parameters of the BESS calculated in steps 162 and 166 are used as input to step 167. In step 167, these parameters are processed to identify slowly evolving changes occurring over a second time scale, which is longer than the first time scale. Slowly evolving changes / events in the system may indicate degradation. In an exemplary embodiment, the parameter values of the parameter evolution model (i.e., one or several parameter values included in the set of second model parameter values) may simply be compared with the alarm threshold limits set by an alarm, which is passed to the end user (step 175). For example, capacitor fading can be identified by a decrease in the capacity of the battery to store energy, and an increase in ohmic resistance indicates power fading.
[0229] Steps 161 to 167 can be considered as a procedure for performing a diagnosis of a slowly developing change such as deterioration.
[0230] In step 168, one or more operating scenarios are defined. The default scenario is to assume that current operating trends will continue into the future. Optionally, the user may define other operating scenarios, such as changes in temperature control settings, circuit breaker states, charge and discharge rates, etc. These operating scenarios are passed to the system as part of configuration P2 and converted into model inputs (e.g., voltage operating profiles) that are passed to step 169. The model inputs may also take the form of Monte Carlo simulations (i.e., probabilities of operating scenarios can be considered and simulated multiple times to generate various parameters).
[0231] The operating scenario, converted into model inputs in step 168, along with the parameter evolution model calculated in step 166, are passed to step 169. In step 169, future parameter values of the BESS model are predicted for the operating scenario using the parameter evolution model.
[0232] In step 170, the future state of the BESS system is estimated using the future parameter values of the BESS model estimated in step 169. The estimation can be performed using the model evaluation method described for step 163. These values are passed to step 171 where the future performance of the battery is estimated. This evaluation can take the form of determining key performance indicators (KPIs) such as state of charge, state of health, efficiency, etc., which can be calculated from the estimated future values of current and voltage. These future KPIs are passed to step 172 where the most likely future KPI values for each operating scenario, and the associated variances, can be calculated. This information is passed to the end user (step 175).
[0233] Steps 161-172 can be considered as steps for predicting the future performance of the BESS for various operating scenarios.
[0234] In step 173, the future KPIs calculated in step 171 and the associated future values of the parameters of the BESS model calculated in step 169 are used as input to an RUL estimation step 173. The predicted SOH (an indicator of remaining capacity) and ohmic resistance (an indicator of power fading) can be utilized to estimate the remaining useful life RUL of the battery. The RUL is estimated for all operating scenarios. Furthermore, the probability of failure for a given operating scenario can also be calculated using methods known in the current state of the art (e.g. Bayesian estimation, expert knowledge, etc.). The estimated RUL for all operating scenarios is passed to step 174 where the most likely future RUL value and associated probability for each operating scenario is calculated. This information is passed to the end user (step 175).
[0235] Steps 161-174 can be considered as a procedure for predicting the RUL of the BESS for various operating scenarios.
[0236] In step 175, the user is provided with information about the previous and current performance and health of the BESS for various scenarios, as well as predictions of battery RUL and performance. Using this decision support information, the user can select the desired scenario, which can be translated into control signals or maintenance plans (not shown here). The user can also define various operating scenarios to be tested and feed the scenarios back to step 168.
[0237] Alternatively, or in addition to allowing a user to define operating scenarios, a multi-objective optimization technique can be implemented to automatically select the optimal strategy in terms of maximizing a user-defined goal (e.g., maximizing future availability, maximizing efficiency, etc.).
[0238] As explained with reference to Figures 13 and 14, the processing of the measurements can be performed in different layers. This can include a parameter estimation layer 151, where parameter values of the BESS model are determined, a parameter evolution model layer 152, where a parameter evolution model is trained, a model evaluation layer 153, where future BESS model parameter values and / or future BESS states are predicted, a description layer 154, where residuals between the calculated observables and the measurements are determined (step 164) and / or performance and / or RUL predictions are performed (steps 173 and 170), a decision layer 155, where different decisions can be made and which can include a diagnostic layer 156 for identifying events occurring in a first time scale (e.g. sudden events due to faults) and trends occurring over a second, longer time scale (e.g. degradation processes).
[0239] The method can be used for rechargeable batteries in a BESS that has voltage, current and temperature measurements available from the battery or battery string level. The disclosed method can be implemented for all batteries or all battery strings in the BESS.
[0240] The battery may be a lithium ion battery or another battery type. Results from multiple cells, racks, and batteries in a BESS can be combined with other similar models of other elements in the system. Results from all elements can be combined in a user interface to provide a global overview of the BESS.
[0241] Instead of, or in addition to, sending the parameters and residuals to a user interface, the information can be passed to a cloud infrastructure, allowing for evaluation of parameters and behavior across a fleet of equipment.
[0242] The device and system according to the embodiment can be applied to any BESS, and is not limited thereto, for example, the device and system according to the embodiment can be applied to a BESS having a power of 1 MW or more, for example, 2 MW or more, for example, 5 MW or more.
[0243] The computing system 13 may be integrated into the BESS as shown in FIG.
[0244] 15 shows a BESS system 180. The BESS system 180 includes an enclosure 181. At least one rechargeable battery 11 is provided within the enclosure 181. The enclosure may be a rack, a container, or other enclosure.
[0245] The BESS system 180 includes a power converter 182. The power converter 182 may be disposed within an enclosure 181.
[0246] The BESS system 180 includes a control system 183. The control system 183 may be or may include a battery management system (BMS). The control system 183 may be disposed within the enclosure 181.
[0247] A computing system 13 according to the present invention may be integrated or coupled with a control system 183. The computing system 13 may be disposed within an enclosure 181.
[0248] In a BESS system 180 as shown in FIG. 15, functions such as prediction of short-term events and long-term degradation, diagnostics, etc. can be integrated into a single solution having one or several rechargeable batteries 11, associated converter(s) 182, and a control system 183 housed in a common housing or enclosure.
[0249] The methods and apparatus are applicable to, but not limited to, rechargeable energy storage systems such as rechargeable lithium ion batteries. By way of example, the disclosed techniques can be applied to other power system assets or industrial assets.
[0250] FIG. 16 illustrates a system 190 including an asset 191 and a computing system 13. The computing system 13 is operative to receive measurements indicative of a state of the asset 191 from at least one data acquisition unit 12. The computing system 13 is operative to process the received measurements to determine a sequence of a set of first model parameter values of the power system asset model. Each of the sets of first model parameter values is determined for a different time or time interval. The computing system 13 is operative to use the sequence of the sets of first model parameter values to determine a set of second model parameter values of a parameter evolution model different from the power system asset model. The parameter evolution model describes the evolution of one, some, or all of the first model parameter values of the power system asset model. The computing system 13 is operative to generate and provide an output in response to at least one of the first model parameter value set of the power system asset model and the second model parameter value set of the parameter evolution model.
[0251] The asset 191 may be a transformer. The power system asset model may be a transformer model. Determining the second set of model parameter values may include determining a rate of change of a transformer characteristic associated with insulation, windings, or bushings of the transformer. By way of example, the transformer 191 may have an insulation system having an insulating fluid (e.g., insulating oil). The second set of model parameter values may include a value of a parameter indicative of a rate of change of concentration(s) of gas(es) dissolved in the insulating fluid.
[0252] Although the present invention has been described in detail in the drawings and the foregoing description, such description should be considered as illustrative or exemplary and not restrictive. Modifications to the disclosed embodiments can be understood by those skilled in the art from a study of the drawings, the disclosure, and the appended claims to practice the claimed invention. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite articles "a" or "an" do not exclude a plurality. The mere fact that certain elements or steps are recited in separate claims does not indicate that combinations of these elements or steps cannot be used to advantage, and in particular, any further meaningful combination of claims should be considered as disclosed in addition to the actual claim dependencies.
Claims
1. A method for monitoring and / or operating power system assets (11; 11a-c), the method being performed by a computing system (13), receiving measurements indicative of a condition of said power system assets (11; 11a-c) from at least one data acquisition unit (12); processing the received measurements to determine a sequence of a first set (71-73) of model parameter values for a power system asset model, each of the first sets (71-73) of model parameter values being determined for a different time or time interval; determining a second set of model parameter values (79) of a parameter evolution model different from the power system asset model using the sequence of the first set of model parameter values (71-73), the parameter evolution model describing the evolution of one, some or all of the first model parameter values (71-73) of the power system asset model; generating and providing an output as a function of at least one of the first set of model parameter values (71-73) of the power system asset model and the second set of model parameter values (79) of the parameter evolution model.
2. 2. The method of claim 1, wherein the parameter evolution model describes the evolution of one, some or all first model parameter values (71-73) of the power system asset model as a function of operating and / or ambient conditions.
3. The method of claim 1 or 2, wherein the step of generating an output comprises performing an impairment diagnosis using the parameter evolution model, and the output is generated as a function of a result of the impairment diagnosis.
4. The method of any one of claims 1 to 3, further comprising using the parameter evolution model to generate a forecast of future evolution of the one, some or all of the first model parameter values (71-73) of the power system asset model.
5. 5. The method of claim 4, wherein the forecast is used to generate outputs including at least one of a remaining useful life (RUL) based on the forecasted future developments, a future power system asset state, and a future performance of the power system assets.
6. 6. The method of claim 4 or 5, wherein the parameter evolution model describes the evolution of one, some or all of the first model parameter values (71-73) of the power system asset model as a function of ambient and / or operating conditions, the method comprising using the parameter evolution model to generate a number of forecasts of future evolution of the one, some or all of the first model parameter values (71-73) of the power system asset model, the number of forecasts being established for a number of different future ambient and / or operating conditions.
7. 7. The method of claim 6, further comprising receiving user input specifying at least a subset of the future ambient and / or operating conditions and / or automatically generating at least a subset of the future ambient and / or operating conditions.
8. using the power system asset model to perform a diagnosis of events occurring on a first time scale; and using the parameter evolution model to identify evolutions of the first model parameter values (71-73) occurring at a second time scale that exceeds the first time scale.
9. 9. The method of claim 8, wherein at least one of the first set of model parameter values (71-73) of the power system asset model is used in combination with measurements to calculate a current power system asset state, calculate residuals between the measurements and observables estimated from the current power system asset state, and perform a diagnosis of an event occurring at the first time scale based on the residuals.
10. The output is: Information output via a human machine interface (18) HMI, optionally including an alarm or warning; A control signal is output to a controller, in particular a battery management system BMS or a microgrid controller (41a-c); The method according to any one of claims 1 to 9, comprising one or more of the following:
11. The method of any preceding claim, wherein the output includes power system asset health information.
12. The method of any one of claims 1 to 11, further comprising intermittently adjusting ambient and / or operating conditions of the power system assets (11; 11a-c) to determine the second set of model parameter values for the adjusted ambient and / or operating conditions.
13. The method according to any one of claims 1 to 12, wherein the power system asset is or comprises a rechargeable energy storage system ESS, in particular an electrochemical ESS, in particular a rechargeable battery energy storage system (11; 11a-c) BESS.
14. 14. The method of claim 13, wherein the ESS includes a plurality of batteries or a plurality of battery strings, and the first set of model parameter values (71-73) and the second set of model parameter values (79) of a parameter evolution model are determined separately for each of the plurality of batteries or battery strings.
15. determining the first set of model parameter values (71-73) comprises solving a discrete-time linear differential equation; and / or The method according to any one of the preceding claims, wherein the step of determining the second set of model parameter values (79) comprises at least one of machine learning, fitting, regression analysis, in particular linear regression.
16. 1. A computing system comprising: an interface (17; 31) operative to receive measurements indicative of a condition of the power system assets (11; 11a-c); coupled to said interface (17; 31), processing the received measurements to determine a sequence of a first set (71-73) of model parameter values for a power system asset model, each of the first sets (71-73) of model parameter values being determined for a different time or time interval; determining, using the sequence of the first model parameter values (71-73), a second model parameter value set (79) of a parameter evolution model distinct from the power system asset model, the parameter evolution model describing the evolution of one, some or all of the first model parameter values (71-73) of the power system asset model; and generating and providing an output in response to the first set of model parameter values (71-73) of at least one of the power system asset models and the second set of model parameter values (79) of the parameter evolution model.
17. 1. A system comprising: A battery energy storage system (11; 11a-c) BESS; a data acquisition unit (12) operative to collect measurements indicative of a condition of the BESS; A system comprising: a computing system (13) according to claim 16, coupled to said data acquisition unit (12).
18. The method of claim 10, wherein the control signals control at least one of a charge rate, a discharge rate, an ambient temperature of a power system asset, a depth of discharge, and a protection system.
19. The first set of model parameter values (71-73) self-discharge resistance (Rb), A first RC element (Rh, Ch) for modeling charge / discharge hysteresis; A second RC element (Rp, Cp) for modeling long-term transients; A third RC element (Rc, Cc) for modeling short-term transients The method according to any one of claims 1 to 15 and 18, comprising the parameter values:
Citation Information
Patent Citations
Evaluation device, evaluation method, and evaluation program for power storage system
JP2016092896A
Data processing to characterize the thermal behavior of batteries
JP2019512774A
Method and system for controlling a rechargeable battery
JP2020533772A
Method and system for predicting solar energy production
US20050039787A1
System and Method for Posteriori Adaptation of a State of Charge Model in a Battery
US20140032141A1
