Physics-informed neural networks(PINNs)have become an attractive machine learning framework for obtaining solutions to partial differential equations(PDEs).PINNs embed initial,boundary,and PDE constraints into the los...Physics-informed neural networks(PINNs)have become an attractive machine learning framework for obtaining solutions to partial differential equations(PDEs).PINNs embed initial,boundary,and PDE constraints into the loss function.The performance of PINNs is generally affected by both training and sampling.Specifically,training methods focus on how to overcome the training difficulties caused by the special PDE residual loss of PINNs,and sampling methods are concerned with the location and distribution of the sampling points upon which evaluations of PDE residual loss are accomplished.However,a common problem among these original PINNs is that they omit special temporal information utilization during the training or sampling stages when dealing with an important PDE category,namely,time-dependent PDEs,where temporal information plays a key role in the algorithms used.There is one method,called Causal PINN,that considers temporal causality at the training level but not special temporal utilization at the sampling level.Incorporating temporal knowledge into sampling remains to be studied.To fill this gap,we propose a novel temporal causality-based adaptive sampling method that dynamically determines the sampling ratio according to both PDE residual and temporal causality.By designing a sampling ratio determined by both residual loss and temporal causality to control the number and location of sampled points in each temporal sub-domain,we provide a practical solution by incorporating temporal information into sampling.Numerical experiments of several nonlinear time-dependent PDEs,including the Cahn–Hilliard,Korteweg–de Vries,Allen–Cahn and wave equations,show that our proposed sampling method can improve the performance.We demonstrate that using such a relatively simple sampling method can improve prediction performance by up to two orders of magnitude compared with the results from other methods,especially when points are limited.展开更多
In this paper,we establish a new multivariate Hermite sampling series involving samples from the function itself and its mixed and non-mixed partial derivatives of arbitrary order.This multivariate form of Hermite sam...In this paper,we establish a new multivariate Hermite sampling series involving samples from the function itself and its mixed and non-mixed partial derivatives of arbitrary order.This multivariate form of Hermite sampling will be valid for some classes of multivariate entire functions,satisfying certain growth conditions.We will show that many known results included in Commun Korean Math Soc,2002,17:731-740,Turk J Math,2017,41:387-403 and Filomat,2020,34:3339-3347 are special cases of our results.Moreover,we estimate the truncation error of this sampling based on localized sampling without decay assumption.Illustrative examples are also presented.展开更多
With its generality and practicality, the combination of partial charging curves and machine learning(ML) for battery capacity estimation has attracted widespread attention. However, a clear classification,fair compar...With its generality and practicality, the combination of partial charging curves and machine learning(ML) for battery capacity estimation has attracted widespread attention. However, a clear classification,fair comparison, and performance rationalization of these methods are lacking, due to the scattered existing studies. To address these issues, we develop 20 capacity estimation methods from three perspectives:charging sequence construction, input forms, and ML models. 22,582 charging curves are generated from 44 cells with different battery chemistry and operating conditions to validate the performance. Through comprehensive and unbiased comparison, the long short-term memory(LSTM) based neural network exhibits the best accuracy and robustness. Across all 6503 tested samples, the mean absolute percentage error(MAPE) for capacity estimation using LSTM is 0.61%, with a maximum error of only 3.94%. Even with the addition of 3 m V voltage noise or the extension of sampling intervals to 60 s, the average MAPE remains below 2%. Furthermore, the charging sequences are provided with physical explanations related to battery degradation to enhance confidence in their application. Recommendations for using other competitive methods are also presented. This work provides valuable insights and guidance for estimating battery capacity based on partial charging curves.展开更多
In this work,multivariate detection limits(MDL)estimator was obtained based on the microelectro-mechanical systems–near infrared(MEMS–NIR)technology coupled with two sampling accessories to assess the detection capa...In this work,multivariate detection limits(MDL)estimator was obtained based on the microelectro-mechanical systems–near infrared(MEMS–NIR)technology coupled with two sampling accessories to assess the detection capability of four quality parameters(glycyrrhizic acid,liquiritin,liquiritigenin and isoliquiritin)in licorice from di®erent geographical regions.112 licorice samples were divided into two parts(calibration set and prediction set)using Kennard–Stone(KS)method.Four quality parameters were measured using high-performance liquid chromatography(HPLC)method according to Chinese pharmacopoeia and previous studies.The MEMS–NIR spectra were acquired from¯ber optic probe(FOP)and integrating sphere,then the partial least squares(PLS)model was obtained using the optimum processing method.Chemometrics indicators have been utilized to assess the PLS model performance.Model assessment using chemometrics indicators is based on relative mean prediction error of all concentration levels,which indicated relatively low sensitivity for low-content analytes(below 1000 parts per million(ppm)).Therefore,MDL estimator was introduced with alpha error and beta error based on good prediction characteristic of low concentration levels.The result suggested that MEMS–NIR technology coupled with fiber optic probe(FOP)and integrating sphere was able to detect minor analytes.The result further demonstrated that integrating sphere mode(i.e.,MDL0:05;0:05,0.22%)was more robust than FOP mode(i.e.,MDL0:05;0:05,0.48%).In conclusion,this research proposed that MDL method was helpful to determine the detection capabilities of low-content analytes using MEMS–NIR technology and successful to compare two sampling accessories.展开更多
Joint chance constraints(JCCs)can ensure the consistency and correlation of stochastic variables when participating in decision-making.Sample average approximation(SAA)is the most popular method for solving JCCs in un...Joint chance constraints(JCCs)can ensure the consistency and correlation of stochastic variables when participating in decision-making.Sample average approximation(SAA)is the most popular method for solving JCCs in unit commitment(UC)problems.However,the typical SAA requires large Monte Carlo(MC)samples to ensure the solution accuracy,which results in large-scale mixed-integer programming(MIP)problems.To address this problem,this paper presents the partial sample average approximation(PSAA)to deal with JCCs in UC problems in multi-area power systems with wind power.PSAA partitions the stochastic variables and historical dataset,and the historical dataset is then partitioned into non-sampled and sampled sets.When approximating the expectation of stochastic variables,PSAA replaces the big-M formulation with the cumulative distribution function of the non-sampled set,thus preventing binary variables from being introduced.Finally,PSAA can transform the chance constraints to deterministic constraints with only continuous variables,avoiding the large-scale MIP problem caused by SAA.Simulation results demonstrate that PSAA has significant advantages in solution accuracy and efficiency compared with other existing methods including traditional SAA,SAA with improved big-M,SAA with Latin hypercube sampling(LHS),and the multi-stage robust optimization methods.展开更多
Physics-informed neural networks(PINNs)have shown remarkable prospects in solving the forward and inverse problems involving partial differential equations(PDEs).The method embeds PDEs into the neural network by calcu...Physics-informed neural networks(PINNs)have shown remarkable prospects in solving the forward and inverse problems involving partial differential equations(PDEs).The method embeds PDEs into the neural network by calculating the PDE loss at a set of collocation points,providing advantages such as meshfree and more convenient adaptive sampling.However,when solving PDEs using nonuniform collocation points,PINNs still face challenge regarding inefficient convergence of PDE residuals or even failure.In this work,we first analyze the ill-conditioning of the PDE loss in PINNs under nonuniform collocation points.To address the issue,we define volume weighting residual and propose volume weighting physics-informed neural networks(VW-PINNs).Through weighting the PDE residuals by the volume that the collocation points occupy within the computational domain,we embed explicitly the distribution characteristics of collocation points in the loss evaluation.The fast and sufficient convergence of the PDE residuals for the problems involving nonuniform collocation points is guaranteed.Considering the meshfree characteristics of VW-PINNs,we also develop a volume approximation algorithm based on kernel density estimation to calculate the volume of the collocation points.We validate the universality of VW-PINNs by solving the forward problems involving flow over a circular cylinder and flow over the NACA0012 airfoil under different inflow conditions,where conventional PINNs fail.By solving the Burgers’equation,we verify that VW-PINNs can enhance the efficiency of existing the adaptive sampling method in solving the forward problem by three times,and can reduce the relative L 2 error of conventional PINNs in solving the inverse problem by more than one order of magnitude.展开更多
基金Project supported by the Key National Natural Science Foundation of China(Grant No.62136005)the National Natural Science Foundation of China(Grant Nos.61922087,61906201,and 62006238)。
文摘Physics-informed neural networks(PINNs)have become an attractive machine learning framework for obtaining solutions to partial differential equations(PDEs).PINNs embed initial,boundary,and PDE constraints into the loss function.The performance of PINNs is generally affected by both training and sampling.Specifically,training methods focus on how to overcome the training difficulties caused by the special PDE residual loss of PINNs,and sampling methods are concerned with the location and distribution of the sampling points upon which evaluations of PDE residual loss are accomplished.However,a common problem among these original PINNs is that they omit special temporal information utilization during the training or sampling stages when dealing with an important PDE category,namely,time-dependent PDEs,where temporal information plays a key role in the algorithms used.There is one method,called Causal PINN,that considers temporal causality at the training level but not special temporal utilization at the sampling level.Incorporating temporal knowledge into sampling remains to be studied.To fill this gap,we propose a novel temporal causality-based adaptive sampling method that dynamically determines the sampling ratio according to both PDE residual and temporal causality.By designing a sampling ratio determined by both residual loss and temporal causality to control the number and location of sampled points in each temporal sub-domain,we provide a practical solution by incorporating temporal information into sampling.Numerical experiments of several nonlinear time-dependent PDEs,including the Cahn–Hilliard,Korteweg–de Vries,Allen–Cahn and wave equations,show that our proposed sampling method can improve the performance.We demonstrate that using such a relatively simple sampling method can improve prediction performance by up to two orders of magnitude compared with the results from other methods,especially when points are limited.
文摘In this paper,we establish a new multivariate Hermite sampling series involving samples from the function itself and its mixed and non-mixed partial derivatives of arbitrary order.This multivariate form of Hermite sampling will be valid for some classes of multivariate entire functions,satisfying certain growth conditions.We will show that many known results included in Commun Korean Math Soc,2002,17:731-740,Turk J Math,2017,41:387-403 and Filomat,2020,34:3339-3347 are special cases of our results.Moreover,we estimate the truncation error of this sampling based on localized sampling without decay assumption.Illustrative examples are also presented.
基金supported by the National Natural Science Foundation of China (52075420)the National Key Research and Development Program of China (2020YFB1708400)。
文摘With its generality and practicality, the combination of partial charging curves and machine learning(ML) for battery capacity estimation has attracted widespread attention. However, a clear classification,fair comparison, and performance rationalization of these methods are lacking, due to the scattered existing studies. To address these issues, we develop 20 capacity estimation methods from three perspectives:charging sequence construction, input forms, and ML models. 22,582 charging curves are generated from 44 cells with different battery chemistry and operating conditions to validate the performance. Through comprehensive and unbiased comparison, the long short-term memory(LSTM) based neural network exhibits the best accuracy and robustness. Across all 6503 tested samples, the mean absolute percentage error(MAPE) for capacity estimation using LSTM is 0.61%, with a maximum error of only 3.94%. Even with the addition of 3 m V voltage noise or the extension of sampling intervals to 60 s, the average MAPE remains below 2%. Furthermore, the charging sequences are provided with physical explanations related to battery degradation to enhance confidence in their application. Recommendations for using other competitive methods are also presented. This work provides valuable insights and guidance for estimating battery capacity based on partial charging curves.
基金This work was financially supported fromthe National Natural Science Foundation of China(81303218)Doctoral Fund of China (20130013120006)Special Fund of Outstanding Young Teachers and Innovation Team.
文摘In this work,multivariate detection limits(MDL)estimator was obtained based on the microelectro-mechanical systems–near infrared(MEMS–NIR)technology coupled with two sampling accessories to assess the detection capability of four quality parameters(glycyrrhizic acid,liquiritin,liquiritigenin and isoliquiritin)in licorice from di®erent geographical regions.112 licorice samples were divided into two parts(calibration set and prediction set)using Kennard–Stone(KS)method.Four quality parameters were measured using high-performance liquid chromatography(HPLC)method according to Chinese pharmacopoeia and previous studies.The MEMS–NIR spectra were acquired from¯ber optic probe(FOP)and integrating sphere,then the partial least squares(PLS)model was obtained using the optimum processing method.Chemometrics indicators have been utilized to assess the PLS model performance.Model assessment using chemometrics indicators is based on relative mean prediction error of all concentration levels,which indicated relatively low sensitivity for low-content analytes(below 1000 parts per million(ppm)).Therefore,MDL estimator was introduced with alpha error and beta error based on good prediction characteristic of low concentration levels.The result suggested that MEMS–NIR technology coupled with fiber optic probe(FOP)and integrating sphere was able to detect minor analytes.The result further demonstrated that integrating sphere mode(i.e.,MDL0:05;0:05,0.22%)was more robust than FOP mode(i.e.,MDL0:05;0:05,0.48%).In conclusion,this research proposed that MDL method was helpful to determine the detection capabilities of low-content analytes using MEMS–NIR technology and successful to compare two sampling accessories.
基金supported by the National Natural Science Foundation of China(No.51977042)。
文摘Joint chance constraints(JCCs)can ensure the consistency and correlation of stochastic variables when participating in decision-making.Sample average approximation(SAA)is the most popular method for solving JCCs in unit commitment(UC)problems.However,the typical SAA requires large Monte Carlo(MC)samples to ensure the solution accuracy,which results in large-scale mixed-integer programming(MIP)problems.To address this problem,this paper presents the partial sample average approximation(PSAA)to deal with JCCs in UC problems in multi-area power systems with wind power.PSAA partitions the stochastic variables and historical dataset,and the historical dataset is then partitioned into non-sampled and sampled sets.When approximating the expectation of stochastic variables,PSAA replaces the big-M formulation with the cumulative distribution function of the non-sampled set,thus preventing binary variables from being introduced.Finally,PSAA can transform the chance constraints to deterministic constraints with only continuous variables,avoiding the large-scale MIP problem caused by SAA.Simulation results demonstrate that PSAA has significant advantages in solution accuracy and efficiency compared with other existing methods including traditional SAA,SAA with improved big-M,SAA with Latin hypercube sampling(LHS),and the multi-stage robust optimization methods.
基金supported by the National Natural Science Foundation of China(Grant No.92152301)the National Key Research and Development Program of China(Grant No.2022YFB4300200)the Shaanxi Provincial Key Research and Development Program(Grant No.2023-ZDLGY-27).
文摘Physics-informed neural networks(PINNs)have shown remarkable prospects in solving the forward and inverse problems involving partial differential equations(PDEs).The method embeds PDEs into the neural network by calculating the PDE loss at a set of collocation points,providing advantages such as meshfree and more convenient adaptive sampling.However,when solving PDEs using nonuniform collocation points,PINNs still face challenge regarding inefficient convergence of PDE residuals or even failure.In this work,we first analyze the ill-conditioning of the PDE loss in PINNs under nonuniform collocation points.To address the issue,we define volume weighting residual and propose volume weighting physics-informed neural networks(VW-PINNs).Through weighting the PDE residuals by the volume that the collocation points occupy within the computational domain,we embed explicitly the distribution characteristics of collocation points in the loss evaluation.The fast and sufficient convergence of the PDE residuals for the problems involving nonuniform collocation points is guaranteed.Considering the meshfree characteristics of VW-PINNs,we also develop a volume approximation algorithm based on kernel density estimation to calculate the volume of the collocation points.We validate the universality of VW-PINNs by solving the forward problems involving flow over a circular cylinder and flow over the NACA0012 airfoil under different inflow conditions,where conventional PINNs fail.By solving the Burgers’equation,we verify that VW-PINNs can enhance the efficiency of existing the adaptive sampling method in solving the forward problem by three times,and can reduce the relative L 2 error of conventional PINNs in solving the inverse problem by more than one order of magnitude.
