Simple Black-box Model Research Articles

Aging Study of In-Use Lithium-Ion Battery Packs to Predict End of Life Using Black Box Model

In order to study the state of health (SOH) of unbalanced battery packs in real life, a thorough analysis is carried out using only data available and standard charging material. The possible relationships between the different parameters and how they affect aging are studied, leading to the identification of five key parameters to indicate aging, as well as parameters influencing aging. Based on the measurement results, a simple black box model using evolutionary genetic algorithm is presented, which is used as end-of-life prediction model of the battery pack, successfully providing an approximate estimation of aging. This approach might thus be used for the supervision of battery systems during real-life use.

Hydrological process knowledge in catchment modelling – Lessons and perspectives from 60 years development

AbstractHydrological process knowledge has advanced significantly during the past six decades. During the same period catchment models have undergone major developments including simple black box models, lumped conceptual models, hydrological response unit models, spatially distributed process‐based models and, recently, the emergence of machine learning hybrid models. This development has been enabled by improved understanding of hydrological processes together with ever increasing computer power and improved availability and accessibility of data. During the first couple of decades, a key assumption motivating the development towards increasing complexity of model codes was that more detailed process description would lead to more accurate model simulations and enable prediction of impacts from human activities that previous models were not able to provide. Subsequently, scientific tests showed that this is very often not the case, leading towards a recognition of the importance of careful model evaluation accounting for key uncertainties in data, model parameters and model conceptual understanding. We have reviewed 54 model studies from the past 60 years and characterized them with respect to model type, spatial discretization and model evaluation techniques. This showed clear development trends and different strategies for enhancing hydrological process knowledge in models. In addition, we present a case study, where we use two models for the same catchment. The models are identical except for the spatial discretization of 100 m and 500 m, respectively. The two models have an apparent equal performance measured against standard calibration metrics, but nevertheless show large differences when considering detailed process information such as partitioning of streamflow components and water table depth patterns, that was not considered during the model calibration process. The paper discusses perspectives for enhancing hydrological process knowledge in future catchment modelling concluding that the emergence of big data is likely to become a major game changer.

Deep Reinforcement Learning for room temperature control: a black-box pipeline from data to policies

Deep Reinforcement Learning (DRL) recently emerged as a possibility to control complex systems without the need to model them. However, since weeks long experiments are needed to assess the performance of a building controller, people still have to rely on accurate simulation environments to train and tune DRL agents in tractable amounts of time before deploying them, shifting the burden back to the original issue of designing complex models. In this work, we show that it is possible to learn control policies on simple black-box linear room temperature models, thereby alleviating the heavy engineering usually required to build accurate surrogates. We develop a black-box pipeline, where historical data is taken as input to produce room temperature control policies. The trained DRL agents are capable of beating industrial rule-based controllers both in terms of energy consumption and comfort satisfaction, using novel penalties to introduce expert knowledge, i.e. to incentivize agents to follow expected behaviors, in the reward function. Moreover, one of the best agents was deployed on a real building for one week and was able to save energy while maintaining adequate comfort levels, indicating that low-complexity models might be enough to learn control policies that perform well on real buildings.

USING META-MODELS IN SIMULATION-BASED INVESTMENT ANALYSIS � STUDYING THE FINANCING MIX OF METAL MINING INVESTMENTS

This paper is the first documented research effort on how simple meta-models can be used in simulation-based investment analysis. Modern computers allow the construction and simulation of near real-world emulating models, often referred to as “digital twins”, that offer requisite variety to real world phenomena, such as an industrial investment. These models can be extremely complex and computationally demanding which reduces the scope of their practical applications. This is where meta-models can help. Meta-models are simple black-box models that are fitted with the input-output -combinations from more complex models to be able to approximate complex model behavior. As the simple meta-models are very fast to solve they may be used to explore much larger solution spaces with considerably higher speed and less computing power needed than the original models. We demonstrate how the meta-modeling approach can be used in the context of metal mining investment analysis that is originally conducted with a dynamic system model constructed based on a real-world metal mining investment. We show how two simple meta-models, a linear regression model and a regression-tree model, can be used in gaining insight about a suitable financing-mix for the said metal mining investment

Determining the Time Constant of Arcs at Arbitrary Current Levels

For the development and optimization of gas circuit breakers and switchgear, a detailed understanding of the arc related processes is of great importance. Ideally, analytical or numerical models with predicitive capability can be found and used during the design process preceding costly and time-consuming experiments.In the present contribution, we report on a novel measurement and evaluation technique to determine the thermal arc time constant ("thermal inertia") that is commonly used in simple black-box models to describe the arc's dynamical properties. The method is introduced and applied to example arcs under varying blow gas conditions in air.

Modern Optimization Methods in Water Resources Planning, Engineering and Management

Mathematical (analytical, numerical and optimization) models are employed in many disciplines including the water resources planning, engineering and management. These models can vary from a simple black-box model to a sophisticated distributed physics-based model. Recently, development and employment of modern optimization methods (MOMs) have become popular in the area of mathematical modeling. This paper overviews the MOMs based on the evolutionary search which were developed over mostly the last 30 years. These methods have wide application in practice from finance to engineering and this paper focuses mostly on the applications in the area of water resources planning, engineering and management. Although there are numerous optimization algorithms, the paper outlines the ones that have been widely employed especially in the last three decades; such as the Genetic Algorithm (GA), Ant Colony (AC), Differential Evolution (DE), Particle Swarm (PS), Harmony Search (HS), Genetic Programming (GP), and Gene Expression Programming (GEP). The paper briefly introduces theoretical background of each algorithm and its applications and discusses the merits and, if any, shortcomings. The wide spectrum of applications include, but not limited to, flood control and mitigation, reservoir operation, irrigation, flood routing, river training, flow velocity, rainfall-runoff processes, sediment transport, groundwater management, water quality, hydropower, dispersion, and aquifers.

Predicting operating room case load: An aid to resource allocation

Hospital patient bed utilization can reach 100% with an impact on elective surgery schedules. Analysis of the demand for beds created by elective surgical operations is desirable to manage overall resources under these conditions. For planning and allocating operating rooms, staff, beds and equipment on any given day, hospital administrators would benefit from an accurate prediction of the number of surgical cases that will be completed. Current scheduling techniques do not predict, for a given day in the future, the number of cases that will actually be performed. A study was performed at a 247 bed hospital with 10 operating rooms. The operating rooms were available for reservation more than two weeks in advance. Both block scheduling and open time were available. Using reservation data with a simple Black Box model allows the prediction of the total number of cases to be performed up to two weeks in advance with 90% accuracy. The resultant predictive demand should allow for better resource planning for the Operating Suite as well as required post-op hospital patient beds.

Predicting physical clogging of porous and permeable pavements

Summary Porous pavements are easily retrofitted, and effective in improving water quality and hydrology, but prone to clogging. Despite being a major determinant in the lifespan of porous pavements, there is limited information on the physical clogging processes through these systems. The aim of this study was to understand the main physical processes that govern physical clogging and develop a simple black-box model that predicts physical clogging. The key variables that were hypothesised to influence clogging were pavement design and climate characteristics. Two compressed time scale laboratory experiments were conducted over 3 years on three common porous pavement types; monolithic porous asphalt, modular Hydrapave and monolithic Permapave. Pavement design was found to be an important role in clogging. Permapave did not clog even after 26 years of operation in simulated sub-tropical Brisbane (Australia) climate while porous asphalt and Hydrapave clogged after just 12 years, from surface clogging and geotextile clogging, respectively. Each system was tested using two different dosing patterns: (1) continual wetting with no dry periods and (2) variable inflow rates with drying periods (i.e. representing more natural conditions). The latter dosing method approximately doubled the lifespan of all systems suggesting the influence of climate conditions on clogging. Clogging was found to be highly correlated with cumulative volume and flow rate. A simple black-box regression model that predicts physical clogging was developed as a function of cumulative volume and Brisbane climatic conditions. However it is very likely that the shape of this regression is general, and that it could be calibrated for different climates in the future.

Bifurcation theory of ac electric arcing

The performance of alternating current (ac) electric arcing devices is related to arc extinction or its re-ignition at zero crossings of the current (so-called ‘current zero’, CZ). Theoretical investigations thus usually focus on the transient behaviour of arcs near CZ, e.g. by solving the modelling differential equations in the vicinity of CZ. This paper proposes as an alternative approach to investigate global mathematical properties of the underlying periodically driven dynamic system describing the electric circuit containing the arcing device. For instance, the uniqueness of the trivial solution associated with the insulating state indicates the extinction of any arc. The existence of non-trivial attractors (typically a time-periodic state) points to a re-ignition of certain arcs. The performance regions of arcing devices, such as circuit breakers and arc torches, can thus be identified with the regions of absence and existence, respectively, of non-trivial attractors. Most important for applications, the boundary of a performance region in the model parameter space is then associated with the bifurcation of the non-trivial attractors. The concept is illustrated for simple black-box arc models, such as the Mayr and the Cassie model, by calculating for various cases the performance boundaries associated with the bifurcation of ac arcs.

Modelling and Performance Evaluation of Solid Oxide Fuel Cell for Building Integrated Co‐ and Polygeneration

AbstractModels of fuel cell based combined heat and power systems, used in building energy performance simulation codes, are often based on simple black or grey box models. To model a specific device, input data from experiments are often required for calibration. This paper presents an approach for the theoretical derivation of such data. A generic solid oxide fuel cell (SOFC) system model is described that is specifically developed for the evaluation of building integrated co‐ or polygeneration. First, a detailed computational cell model is developed for a planar SOFC and validated with available numerical and experimental data for intermediate and high temperature SOFCs with internal reforming (IT‐DIR and HT‐DIR). Results of sensitivity analyses on fuel utilisation and air excess ratio are given. Second, the cell model is extended to the stack model, considering stack pressure losses and the radiative heat transfer effect from the stack to the air flow. Third, two system designs based on the IT‐DIR and HT‐DIR SOFCs are modelled. Electric and CHP efficiencies are given for the two systems, as well as performance characteristics, to be used in simulations of building integrated co‐ and polygeneration systems.

Modelling monthly streamflows in two Australian dryland rivers: Matching model complexity to spatial scale and data availability

Summary There is new pressure to develop some of Australia’s dryland river basins for irrigation as development on highly regulated rivers has halted with increasing ecological degradation. Sparse populations, limited infrastructure, and minimal water resource development in Australian dryland river basins mean streamflow records are few and typically short in duration. The Paroo and Warrego basins (74,003 and 62,908 km 2 , respectively; largely undeveloped) in the northwest of the Murray-Darling Basin of Australia, each have four streamflow gauges with more than 15 years of daily record, and four and eight years, respectively, of coincident record. These short records do not fully characterise the highly variable flow regimes. Improved characterisation of the flow regime and determination of the potential impacts of water resource development can be based on streamflow simulation using climate data. Numerous floodplain lakes and wetlands are important ecological assets in these basins. For these environments impacts will be a result of changes in the higher (above mean) flows, which can be reasonably assessed by changes in monthly streamflows. However, high spatial and temporal variability of runoff generation, and the delicate hydrologic balance (runoff is a small fraction of rainfall) in dryland regions means the relationship between rainfall and runoff is complex, and a daily time step model was therefore required to adequately simulate monthly streamflow. It was neither necessary, nor (because of limited data) possible to employ a complex spatially distributed model. Use of a highly parameterised, lumped-parameter conceptual model was also deemed unnecessarily complex, however, it was recognised that a simple black box model was unlikely to be successful. As a compromise, we implemented a simple conceptualisation of the daily water balance in PC spreadsheet software. The model was calibrated and validated using monthly streamflow volumes. The model accounts for 76–88% and 60–81% of the variance in monthly streamflows in the Paroo (four stations) and Warrego (three stations) basins, respectively. We use the model to explore the relative importance of local and catchment rainfall in flood generation, and to predict the timing and magnitude of large floods in these basins over the last 100 years.

Application of linear and nonlinear techniques in river flow forecasting in the Kilombero River basin, Tanzania / Application de techniques linéaires et non-linéaires à la prévision des débits dans le bassin de la Rivière Kilombero, en Tanzanie

Modelling of the rainfall–runoff transformation process and routing of river flows in the Kilombero River basin and its five sub-catchments within the Rufiji River basin in Tanzania was undertaken using three system (black-box) models—a simple linear model, a linear perturbation model and a linear varying gain factor model—in their linear transfer function forms. A lumped conceptual model—the soil moisture accounting and routing model—was also applied to the sub-catchments and the basin. The HEC-HMS model, which is a distributed model, was applied only to the entire Kilombero River basin. River discharge, rainfall and potential evaporation data were used as inputs to the appropriate models and it was observed that sometimes the system models performed better than complex hydrological models, especially in large catchments, illustrating the usefulness of using simple black-box models in datascarce situations.

Model Study of Short-Term Dynamics of Secondary Treatment Reed Beds at Saxby (Leicestershire, UK)

Relatively simple black-box models, such as the well-known k-C* model, are commonly applied to design horizontal sub-surface flow constructed treatment wetlands. Important shortcomings of this model are the oversimplification of reality on the one hand, and the inability to predict short-term effluent dynamics on the other. A possible solution for these drawbacks could be the application of dynamic compartmental models. This article reports on the calibration requirements and the simulation results of such a dynamic model. A quantitative sensitivity analysis was used to identify the most sensitive parameters after which model predictions were optimized by adjusting those parameter values. Model fits were acceptable but missed some of the short-term dynamics observed in reality. At this point, it might therefore still be unwise to use the model as a design tool. Further model adjustments and calibration efforts are needed to enhance its reliability.

Robust control of separated shear flows in simulation and experiment

Closed-loop flow control is gaining more and more interest in the last few years. Whereas most of the published results are based on simulation studies, this work explores the synthesis of closed-loop controllers for separated wall-bounded shear flows in experiments. A methodology and first results are presented for the robust control of a flow over a well established benchmark system, namely a backward-facing step. This system can be seen as a simple representation of the situation in a burner or behind a flame holder. Based on numerical solutions of the Navier–Stokes equations and on experimental wind tunnel results, possible candidates for an online determination of an appropriate output variable are proposed. For our purpose, a substitute for the reattachment length is taken as the control variable. Simple black-box models are derived to describe the behaviour of the system in the vicinity of set-points. Based on these, different robust controllers are synthesised. As expected, these controllers show superior behaviour over the open-loop strategies mainly proposed in the flow control literature.

System identification of electronic nose data from cyanobacteria experiments

Linear black-box modeling techniques are applied to both steady state and dynamic data gathered from two electronic nose experiments involving cyanobacteria cultures. Analysis of the data from a strain identification experiment shows that very simple low order MISO black box model structures are able to produce very high success rates (up to 100%) when identifying the toxic strain of cyanobacteria. This is comparable with the best success rates for steady state data reported elsewhere using artificial neural networks. Analysis of data from a growth phase identification experiment using MIMO black-box models produces success rates of 82.3% for steady state data and 76.6% for dynamic data. This compares poorly with the best performing nonlinear artificial neural networks, which obtained a 95.1% success rate on the same data. This demonstrates the limitations of these linear techniques when applied to more difficult problems.

A simple black box model for erbium-doped fiber amplifiers

We demonstrate a simple black box model to evaluate both saturated gain and corresponding noise figure. This approach is simple and very suitable for system modeling with satisfied accuracy.

Simple black box models predicting potential control parameters during disturbances to a fluidised bed anaerobic reactor

Models of the anaerobic digestion process which predict digester behaviour sufficiently accurately could be used in process control. Although the process is generally considered to be non-linear, it could possibly be represented by an adaptive linear model, where the model adapts rapidly enough to represent the process at differing operating conditions and times in its operating life. Simple linear black box models of low order were investigated, predicting over a limited horizon and relying on current and recent data values to refine the prediction. Independent black box ARX models were identified for gas production rate, % CO2, bicarbonate alkalinity and Total Organic Carbon using on-line data from a fluidised bed reactor at varying organic load. Model predictions looked ahead one sample step (30 minutes) and when validated using data obtained in a different time period (separated by 4-8 weeks) gave significant predictions in each case. All the models consisted of only second or third order polynomials. The non-linear nature of the process was found to have little effect over the operating conditions investigated. Also the variation of the process within a 4-8 week period was not sufficient to cause the models to predict badly.

Simple black box models predicting potential control parameters during disturbances to a fluidised bed anaerobic reactor

Models of the anaerobic digestion process which predict digester behaviour sufficiently accurately could be used in process control. Although the process is generally considered to be non-linear, it could possibly be represented by an adaptive linear model, where the model adapts rapidly enough to represent the process at differing operating conditions and times in its operating life. Simple linear black box models of low order were investigated, predicting over a limited horizon and relying on current and recent data values to refine the prediction. Independent black box ARX models were identified for gas production rate, % CO2, bicarbonate alkalinity and Total Organic Carbon using on-line data from a fluidised bed reactor at varying organic load. Model predictions looked ahead one sample step (30 minutes) and when validated using data obtained in a different time period (separated by 4-8 weeks) gave significant predictions in each case. All the models consisted of only second or third order polynomials. The non-linear nature of the process was found to have little effect over the operating conditions investigated. Also the variation of the process within a 4-8 week period was not sufficient to cause the models to predict badly.