Loading [MathJax]/jax/element/mml/optable/BasicLatin.js
Research article Special Issues

Distributionally robust parameter estimation for nonlinear fed-batch switched time-delay system with moment constraints of uncertain measured output data

  • In this paper, we investigated a nonlinear continuous-time switched time-delay (NCTSTD) system for glycerol fed-batch bioconversion to 1, 3-propanediol with unknown time-delay and system parameters. The measured output data was uncertain, while the first moment information about its distribution was available. Our goal was to identify these unknown quantities under the environment of uncertain measurement output data. A distributionally robust parameter estimation problem (i.e., a bi-level parameter estimation (BLPE) problem) subject to the NCTSTD system was presented, where the expectation of the discrepancy between the output of the NCTSTD system and the uncertain measured output data with respect to its probability distributions was included in the cost functional. By applying the duality theory, the BLPE problem was transformed into a single-level parameter estimation (SLPE) problem with non-smooth term approximated by a smoothing technique and its error analysis was given. Then, the gradients of the cost function of the SLPE problem were derived. A hybrid optimization algorithm was proposed for solving the SLPE problem. The paper concluded by presenting the simulation results.

    Citation: Sida Lin, Jinlong Yuan, Zichao Liu, Tao Zhou, An Li, Chuanye Gu, Kuikui Gao, Jun Xie. Distributionally robust parameter estimation for nonlinear fed-batch switched time-delay system with moment constraints of uncertain measured output data[J]. Electronic Research Archive, 2024, 32(10): 5889-5913. doi: 10.3934/era.2024272

    Related Papers:

    [1] Yawei Liu, Guangyin Cui, Chen Gao . Event-triggered synchronization control for neural networks against DoS attacks. Electronic Research Archive, 2025, 33(1): 121-141. doi: 10.3934/era.2025007
    [2] Bin Zhen, Ya-Lan Li, Li-Jun Pei, Li-Jun Ouyang . The approximate lag and anticipating synchronization between two unidirectionally coupled Hindmarsh-Rose neurons with uncertain parameters. Electronic Research Archive, 2024, 32(10): 5557-5576. doi: 10.3934/era.2024257
    [3] YongKyung Oh, JiIn Kwak, Sungil Kim . Time delay estimation of traffic congestion propagation due to accidents based on statistical causality. Electronic Research Archive, 2023, 31(2): 691-707. doi: 10.3934/era.2023034
    [4] Duhui Chang, Yan Geng . Distributed data-driven iterative learning control for multi-agent systems with unknown input-output coupled parameters. Electronic Research Archive, 2025, 33(2): 867-889. doi: 10.3934/era.2025039
    [5] Jiping Xing, Yunchi Wu, Di Huang, Xin Liu . Transfer learning for robust urban network-wide traffic volume estimation with uncertain detector deployment scheme. Electronic Research Archive, 2023, 31(1): 207-228. doi: 10.3934/era.2023011
    [6] Anatoliy Martynyuk, Gani Stamov, Ivanka Stamova, Yulya Martynyuk–Chernienko . Regularization scheme for uncertain fuzzy differential equations: Analysis of solutions. Electronic Research Archive, 2023, 31(7): 3832-3847. doi: 10.3934/era.2023195
    [7] Ramalingam Sakthivel, Palanisamy Selvaraj, Oh-Min Kwon, Seong-Gon Choi, Rathinasamy Sakthivel . Robust memory control design for semi-Markovian jump systems with cyber attacks. Electronic Research Archive, 2023, 31(12): 7496-7510. doi: 10.3934/era.2023378
    [8] Ke Liu, Hanzhong Liu . Testing for individual and time effects in unbalanced panel data models with time-invariant regressors. Electronic Research Archive, 2022, 30(12): 4574-4592. doi: 10.3934/era.2022232
    [9] Xudong Hai, Chengxu Chen, Qingyun Wang, Xiaowei Ding, Zhaoying Ye, Yongguang Yu . Effect of time-varying delays' dynamic characteristics on the stability of Hopfield neural networks. Electronic Research Archive, 2025, 33(3): 1207-1230. doi: 10.3934/era.2025054
    [10] Yadan Shi, Yongqin Xie, Ke Li, Zhipiao Tang . Attractors for the nonclassical diffusion equations with the driving delay term in time-dependent spaces. Electronic Research Archive, 2024, 32(12): 6847-6868. doi: 10.3934/era.2024320
  • In this paper, we investigated a nonlinear continuous-time switched time-delay (NCTSTD) system for glycerol fed-batch bioconversion to 1, 3-propanediol with unknown time-delay and system parameters. The measured output data was uncertain, while the first moment information about its distribution was available. Our goal was to identify these unknown quantities under the environment of uncertain measurement output data. A distributionally robust parameter estimation problem (i.e., a bi-level parameter estimation (BLPE) problem) subject to the NCTSTD system was presented, where the expectation of the discrepancy between the output of the NCTSTD system and the uncertain measured output data with respect to its probability distributions was included in the cost functional. By applying the duality theory, the BLPE problem was transformed into a single-level parameter estimation (SLPE) problem with non-smooth term approximated by a smoothing technique and its error analysis was given. Then, the gradients of the cost function of the SLPE problem were derived. A hybrid optimization algorithm was proposed for solving the SLPE problem. The paper concluded by presenting the simulation results.



    Distributionally robust optimization (DRO) is an optimization method designed to address uncertainty, particularly in situations where the underlying probability distribution is not fully known. Unlike traditional robust optimization, DRO assumes that we only have partial information about the uncertain probability distribution, rather than complete certainty. By accounting for the worst-case scenarios, DRO can reduce decision-making risks and enhance the robustness of systems across various real-world applications. It has aroused much interest among researchers to develop algorithms to solve the DRO problems in kernel density estimation [1], data-driven problem [2,3], portfolio optimization problem [4], decision bounding problem [5], and matrix moment constraint [6]. However, no dynamical systems are involved in the aforementioned DRO problems. Recently, the DRO problems involving linear discrete-time systems are discussed in multi-access space-time block coded MIMO system [7], stochastic model predictive control problem [8], modeling of optimal control problem [9], constrained stochastic system [10], partially observable linear stochastic system [11], and output feedback [12]. However, in these literatures, the dynamical systems involved in the DRO problems are discrete-time dynamical (DTD) systems. Continuous-time dynamical (CTD) systems are not involved. There are major difficulties in solving DRO problems governed by CTD systems such as (i) a DRO problem constrained by a CTD system is fundamentally a bi-level optimization problem, which, unlike a DTD system, cannot be simplified to a single-level optimization problem with a finite number of constraints; and (ii) the presence of CTD systems in DRO problems causes the inner-level objective functional to be expressed as the expectation of a non-convex function. In this paper, we examine a DRO problem influenced by a nonlinear continuous-time switched time-delay (NCTSTD) system encountered in fed-batch production of 1, 3-propanediol (1, 3-PD). That is to say, we directly address the DRO problem governed by a CTD system, rather than discretizing the CTD system.

    1, 3-PD is a key chemical raw material for polymer production because of its unique symmetric structure [13]. The methods for the production of 1, 3-PD can be categorized into two types: microbial conversion and chemical synthesis [14]. This paper focuses on the first category due to its eco-friendliness [15]. 1, 3-PD microbial conversion process of glycerol can be divided into three categories [16]: (i) batch culture [17], (ii) continuous culture [18,19], and (iii) fed-batch culture (see Figure 1 in [20]). In fed-batch culture, substrate inhibition is dramatically decreased, allowing for greater consumption of glycerol and alkali, which in turn leads to increased biomass production and higher concentrations of 1, 3-PD. For fed-batch culture, many interesting papers appeared in the existing literature such as those involving state-dependent impulsive system [21], state-dependent switched system [22], multistage optimal control [23], optimal control [24], a novel downstream process [25], and Koopman modeling [26,27]. However, the references mentioned do not consider the presence of time delays.

    Time delays widely exist in many practical problems, including optimal control problem [28], predator-prey model [29], perimeter control problem [30], the survey [31], sequential time scaling transformation [32], non-autonomous time-delayed SIR model [33], multiregion urban traffic network [34], and sparse optimal control [35]. At the beginning of the fed-batch process, all substrates are placed into the fermentator. However, the chemical reaction does not trigger immediately [36], but needs to go through the process in which all substrates need to be well-mixed. Clearly, completing this process will require a certain amount of time. Thus, this process should be characterized by time-delay systems. In the context of fed-batch culture with time-delay, numerous intriguing papers have been published in the existing literature, such as robust parameter identification [37], multi-objective optimization [38], and optimal control [39]. However, the aforementioned papers assumed that the probability distribution of the output measurement data is fully available. This assumption is obviously not realistic. In practical applications, typically only first-order moment information is obtainable. This is the main motivation for this paper.

    This paper examines an NCTSTD system with unknown time-delay and system parameters to characterize the bioconversion process of 1, 3-PD in fed-batch culture of glycerol caused by K. pneumoniae. Furthermore, the uncertain measurement output data is regarded as a stochastic variable and the only first-order moment information of its distribution is obtainable. Our aim is to identify these unknown quantities in a setting with uncertain output data measurement. Considering these factors, we propose a distributionally robust parameter estimation (DRPE) problem governed by the NCTSTD system, where the objective function contains two levels: (i) the inner-level is to maximize the expectation of the discrepancy between the output of the NCTSTD system and the uncertain output data measurement, and (ii) the outer-level is to minimize with respect to the unknown quantities. By the duality theory, the DRPE problem is equivalently translated into a single-level parameter estimation (SLPE) problem with non-smooth term smoothed by a smoothing function. Its convergence analysis is also given. Moreover, the gradients of the objective function are derived. A hybrid algorithm is proposed for solving the SLPE problem. Finally, numerical results show the applicability of the NCTSTD system to characterize the process of fed-batch culture and the availability of the proposed hybrid algorithm for solving the SLPE problem.

    The contributions of the paper are threefold:

    1) Different from the above literature, the optimal solution of the DRPE problem (OSDRPEP) is between the optimal solution of the robust parameter estimation problem (OSRPEP) and the optimal solution of the stochastic parameter estimation problem (OSSPEP). OSDRPEP builds a bridge between OSRPEP and OSSPEP, neither as conservative as OSRPEP nor as strict as OSSPEP when dealing with random probability distribution.

    2) The DRPE problem subject to the CTD system is actually a bi-level parameter estimation problem, which cannot be described as an SLPE problem with finitely many equality constraints as the example of the DTD system. In this paper, a method on the basis of the duality theory is proposed for solving the DRPE problem governed by the CTD system.

    3) The non-smooth term in the objective function is approximated by a smoothing technique. Some noteworthy theorems are required to prove that the smoothing function can overcome the problem that the constraint qualification is not satisfied.

    The remainder of the paper is structured as follows. A DRPE problem is proposed in Section 2. A computational method is proposed for solving the DRPE problem in Section 3. Numerical results are given in Section 4. In Section 5, we draw some concluding remarks and imply some research directions in the future.

    In fed-batch culture, we assume that the conditions given below are satisfied.

    The concentrations of reactants are uniformly distributed in the reactor. Nonuniform space distribution is ignored.

    In fed-batch culture, the reactor is fed exclusively with glycerol and alkali.

    Based on [43], the NCTSTD system governing the fed-batch culture can be described as follows:

    {dx(t)dt=f(x(t),x(th),p)R5,x(τ1+)=x(τ1),t(τ1,τ],I2N+1,x(0)=x0,x(t)=ϕ(t),t0. (2.1)

    For t(τ1,τ],=2i+1,iΛ2,f(x(t),x(th),p) is of the form given by

    f2i+1(x(t),x(th),p):=[f2i+11(x(t),x(th),p),,f2i+15(x(t),x(th),p)]T,

    where its components are

    {f2i+1j(x(t),x(th),p)=qj(t)x1(th),j{1,3,4,5},f2i+12(x(t),x(th),p)=q2(t)x1(th). (2.2)

    On the other hand, for t(τ1,τ],=2i+2,iΛ1,f(x(t),x(th),p) is of the form given by

    f2i+2(x(t),x(th),p):=[f2i+21(x(t),x(th),p),,f2i+25(x(t),x(th),p)]T,

    where its components are

    {f2i+2j(x(t),x(th),p)=qj(t)x1(th)D(t)xj(t),j{1,3,4,5},f2i+22(x(t),x(th),p)=D(t)[Cs01+rx2(t)]q2(t)x1(th), (2.3)

    with

    D(t)=(1+r)νV(t), (2.4)
    V(t)=V0+1j=1(1+r)νj(τjτj1)+(1+r)ν(tτ1), (2.5)
    q1(t)=p1x2(t)x2(t)+k15ȷ=2(1xȷ(t)xȷ), (2.6)
    q2(t)=p2+q1(t)p4+p6x2(t)x2(t)+k2, (2.7)
    q3(t)=p3+q1(t)p5+p7x2(t)x2(t)+k3, (2.8)
    q4(t)=m4+q1(t)Y4+Δ4x2(t)x2(t)+k4, (2.9)
    q5(t)=q2(t)[c1c2+q1(t)x2(t)+p8p9+q1(t)x2(t)]. (2.10)

    The concentrations of substrate and the product of fermentation fluid are restricted to within a set W defined by

    x(t)W:=5j=1[x_j,¯xj]R5+,t[0,tf]. (2.11)

    Based on sensitivity analysis in [43], the time-delay h and the kinetic parameter vector p, the meanings of which are listed in Table 1, are required to be identified. Such analysis allows us to focus our optimization efforts on these parameters and reduce unnecessary complexity. Thus, they are treated as decision variables to be chosen from the sets defined by

    h[0,¯h],p:=(p1,,p9)TP:=9ı=1[p_ı,¯pı].
    Table 1.  Nomenclature.
    Symbol Representation Symbol Representation
    R The set of real numbers R+ The set of nonnegative real numbers
    N+ The set of positive integers tf The terminal time
    Λ1 {0,1,2,...,N1},NN+ Λ2 {0,1,2,...,N},NN+
    x1(t) The concentration of biomass at time t[0,tf] x2(t) The concentration of glycerol at time t[0,tf]
    x3(t) The concentration of 1, 3-PD at time t[0,tf] x4(t) The concentration of acetate at time t[0,tf]
    x5(t) The concentration of ethanol at time t[0,tf] ϕ() A given initial function
    x0 The initial state In The set {1,2,,n},nN+
    AT The transpose of the vector or matrix A r The velocity ratio of feeding alkali and glycerol
    q1(t) The specific growth rate of cells at time t[0,tf] D(t) The dilution rate at time t[0,tf]
    q2(t) The specific consumption rate of substrate at time t[0,tf] q(t),{3,4,5} The specific formation rates of product at time t[0,tf]
    τ2i+1,iΛ1 The moments at each of which glycerol is added and the culture process switches from batch process to continuous process decided a priori in the experiment τ2i+2,iΛ1 The moments at each of which the flow of glycerol is ended and the culture process jumps from continuous process into batch process decided a priori in the experiment
    Cs0 The concentration of initial feed of substrate in the medium V0 The initial volume of solution in the fermentor
    k1 The Monod saturation constant for the substrate (mmol L−1) p2 The maintenance term of substrate consumption (mmol g−1h−1)
    p3 The maintenance term of 1, 3-PD formation (mmol g−1h−1) m4 The maintenance term of acetate formation (mmol g−1h−1)
    p4 The maximum biomass growth yield in (g mmol−1 p5 The maximum extracellular 1, 3-PD yield in (mmol g−1)
    Y4 The maximum acetate yield in (mmol g−1) k2 The saturation constants of substrate in the kinetic equations with excess terms (mmol L−1)
    c1,c1,p8,p9 The parameters for the determination of yield of ethanol on glycerol k3 The saturation constants of extracellular 1, 3-PD in the kinetic equations with excess terms (mmol L−1)
    k4 The saturation constants of EtOH in the kinetic equations with excess terms (mmol L−1) p1 The maximum specific growth rate (h−1)
    p6 The maximum increment of substrate consumption rate (h−1) p7,Δ4 The maximum increments of product formation rates (h−1)
    ν0 The feeding rate of glycerol in (τ1,τ],I2N+1 p_j The lower bound of pj,jI9
    ¯pj The upper bound of pj,jI9 x_j The lower bound of xj(t),t[0,tf],jI5
    ¯xj The upper bound of xj(t),t[0,tf],jI5 ¯h The upper bound of h
    h The time-delay to be optimizated

     | Show Table
    DownLoad: CSV

    This prior information helps narrow down the parameter space, making the optimization process more stable.

    The aim of this section is to present a DRPE problem subject to the NCTSTD system with the unknown time-delay h and system parameter vector p to be optimized.

    Let x(|h,p):=(x1(|h,p),...,x5(|h,p))T be the solution of system (2.1) (i.e., the system output) corresponding to each pair (h,p)[0,ˉh]×P. In [20,37,38,39], assume that the distributions of output data are known precisely is an idealistic assumption. In fact, the distributions of the measured output data are often not available. Thus, the measured output data vector ˜d is regarded as a random vector. Its first-order moment is defined by

    EP[˜d]=˜z, (2.12)

    where EP represents the expected value according to the probability distribution P; ˜d:=((˜d1)T,,(˜dσ)T)T=(˜d11,,˜d13,,˜dσ1,,˜dσ3)TR3σ;˜dυ refers to the output vector recorded at specific sample times tυ,υIσ subject to 0t1<<tσtf; and ˜z:=(˜z11,,˜z13,,˜zσ1,,˜zσ3)TR3σ.

    The support set ˜D of ˜d is defined as

    ˜D={˜dR3σ:d_υι˜dυι¯dυι,υIσ,ιI3}, (2.13)

    where d_υι and ¯dυι are, respectively, the lower and the upper limits of ˜dυι with d_υι<¯dυι.

    The ambiguity set ˜P of probability measures of ˜d is defined as

    ˜P={P:P(˜d˜D)=1,EP[˜d]=˜z}. (2.14)

    Definition 1. For a given (h,p)[0,ˉh]×P, the discrepancy between the system output xι(tυ|h,p),tυT,ιI3,υIσ, and the measured output data ˜dυι,ιI3,υIσ, is defined by

    Re(h,p,˜d):=13σ3ι=1συ=1{xι(tυ|h,p)˜dυι¯dυι}2, (2.15)

    where T represents the set of specified time points in fed-batch culture, and |T|=σ represents the number of elements in the set T.

    It is assumed that P˜P. Therefore, the worst-case scenario for Re(h,p,˜d) is defined as maxP˜PEP[Re(h,p,˜d)]. Considering this, our DRPE problem is stated below:

    ProblemA:min(h,p)[0,ˉh]×P{maxP˜PEP[Re(h,p,˜d)]}s.t.P(˜d˜D)=1,EP[˜d]=˜z,gj(x(t|h,p))0,t[0,tf],jI10,

    where gj(x(t|h,p))=xj(t|h,p),gj+5(x(t|h,p))=xj(t|h,p)ˉxj,jI5.

    This section presents a numerical approach for solving Problem A. Figure 1 summarizes the approximation and transformation process of solving Problem A. A thorough discussion is given in Section 3.1. Based on Theorems 3 and 4, the aim of Theorem 3 is to calculate the gradients of JHρ,ϵ(h,p,ψ,ζ) in Problem H. Theorems 1 and 2 are applied to carry out the error analysis. Algorithm 1 is employed to solve Problem H.

    Figure 1.  Solution process for Problem A.

    Problem A is a bi-level parameter estimation problem subject to the NCTSTD system. In this subsection, Problem A is converted into a deterministic single-level parameter estimation problem.

    The inner-level problem of Problem A can be expressed as given below.

    Problem B:maxP˜PEP[Re(h,p,˜d)]s.t.P(˜d˜D)=1,EP[˜d]=˜z.

    For ease of use, the expectation of the measured output data vector ˜d is expressed using the Lebesgue integral. Then, Problem B is reformulated as:

    Problem C:maxP˜P˜DRe(h,p,˜d)dPs.t.˜D1(˜d˜D)dP=1,˜D˜ddP=˜z.

    Problem C is difficult to solve since it is a parameter estimation problem with respect to P˜P on ˜D. Considering this, in view of the duality theory [42], Problem C can be converted into its dual problem (i.e., Problem D), which can be solved easier than Problem C. Through the application of the duality theory [42], we obtain the duality of Problem C as follows:

    Problem D:minζR3σ,ψRψ+˜zTζs.t.ψ+dTζRe(h,p,˜d),d˜D, (3.1)

    where ζ:=(ζ11,,ζ13,,ζσ1,,ζσ3)TR3σ,ψ is the dual variable, and Re(h,p,˜d) is as defined in (2.15).

    Based on the definition of Re(h,p,˜d) in (2.15) and max, we have

    (3.1)3ι=1συ=1(dυι¯dυι)23ι=1συ=1(2xι(tυ|h,p)¯dυι+ζυι)dυι¯dυιψ+3ι=1συ=1(xι(tυ|h,p)¯dυι)20,d˜Dmaxd˜D{3ι=1συ=1[(dυι¯dυι)2(2xι(tυ|h,p)¯dυι+ζυι)dυι¯dυι+(xι(tυ|h,p)¯dυι)2]}ψ0. (3.2)

    Based on the separability of the variables dυι,ιI3,υIσ, constraint (3.2) is equivalent to constraint (3.3).

    (3.2)3ι=1συ=1maxdυι[d_υι,¯dυι]{[(dυι¯dυι)2(2xι(tυ|h,p)¯dυι+ζυι)dυι¯dυι+(xι(tυ|h,p)¯dυι)2]}ψ0. (3.3)

    From the properties of quadratic functions, it follows that constraint (3.3) constraint (3.4).

    (3.3)3ι=1συ=1max{(d_υι¯dυι)2(2xι(tυ|h,p)¯dυι+ζυι)d_υι¯dυι+(xι(tυ|h,p)¯dυι)2,1(2xι(tυ|h,p)¯dυι+ζυι)+(xι(tυ|h,p)¯dυι)2}ψ0. (3.4)

    Since max{a1,a2}=a1+a22+|a1a2|2, constraint (3.4) is equivalent to constraint (3.5).

    (3.4)123ι=1συ=1{(d_υι¯dυι)2+1(2xι(tυ|h,p)¯dυι+ζυι)(d_υι¯dυι+1)+2(xι(tυ|h,p)¯dυι)2+|(d_υι¯dυι1)(d_υι¯dυι+12xι(tυ|h,p)¯dυιζυι)|}ψ0. (3.5)

    Letting αυι(x(tυ|h,p),ζ):=(d_υι¯dυι)2+1(2xι(tυ|h,p)¯dυι+ζυι)(d_υι¯dυι+1)+2(xι(tυ|h,p)¯dυι)2;βυι(x(tυ|h,p),ζ):=(d_υι¯dυι1)(d_υι¯dυι+12xι(tυ|h,p)¯dυιζυι). Based on max{a1,a2}=a1+a22+|a1a2|2, we have

    (3.5)123ι=1συ=1{αυι(x(tυ|h,p),ζ)+2max{βυι(x(tυ|h,p),ζ),0}βυι(x(tυ|h,p),ζ)}ψ0max{Λ(h,p,ζ,ψ),0}=0, (3.6)

    where Λ(h,p,ζ,ψ):=123ι=1συ=1{αυι(x(tυ|h,p),ζ)+2max{βυι(x(tυ|h,p),ζ),0}βυι(x(tυ|h,p),ζ)}ψ.

    Problem A is equivalently transformed into the following SLPE problem:

    Problem E:minζR3σ,ψR,h[0,ˉh],pPJE(h,p,ψ,ζ):=ψ+˜zTζs.t.gj(x(t|h,p))0,t[0,tf],jI10,Λ(h,p,ζ,ψ)0. (3.7)

    Problem E involves an infinite number of constraints (i.e., constraint (3.7) that restricts the state variables at each point in [0,tf]). Therefore, Problem E can be regarded as a semi-infinite programming problem [46]. Based on the constraint transformation method involving the definition of max{,} [46], Problem E is equivalently transformed into Problem Ec.

    Problem Ec:minζR3σ,ψR,h[0,ˉh],pPJE(h,p,ψ,ζ):=ψ+˜zTζs.t.10j=1tf0max{gj(x(t|h,p)),0}=0,max{Λ(h,p,ζ,ψ),0}=0. (3.8)

    In view of the penalty function method [45,46], Problem Ec can be approximated by Problem F.

    Problem F:minζR3σ,ψR,h[0,ˉh],pPJFρ(h,p,ψ,ζ):=JE(h,p,ψ,ζ)+ρ10j=1tf0max{gj(x(t|h,p)),0}dt+ρmax{Λ(h,p,ζ,ψ),0}. (3.9)

    Remark 1. It can be demonstrated that if the penalty parameter ρ is sufficiently increased, then any minimizer of Problem F within the region ζR3σ,ψR,h[0,ˉh],pP will be feasible for Problem Ec. Therefore, a solution to Problem Ec can be achieved by minimizing Problem F with appropriately chosen values for the penalty parameter ρ.

    One of the main hardships to solve Problem F numerically is that the second and the third terms in (3.9) are non-smooth, with respect to h and p, due to the existence of the function max{,0}. To surmount this hardship, the function max{,0} is approximated by the smoothing technique [44] as follows.

    γϵ,ρ,κ[η]={0,if ηϵκρ,κρ2ϵη2+η+ϵ2κρ,if ϵκρη0,η+ϵ2κρ,if η>0, (3.10)

    where η is a continuous inequality constraint; ρ is a penalty factor, κ is the number of continuous inequality constraints, and ϵ>0 is called the smoothing parameter.

    Approximate the non-smooth terms in Problem F using the function (3.10), and we get

    Problem H:minζR3σ,ψR,h[0,ˉh],pPJHρ,ϵ(h,p,ψ,ζ):=JE(h,p,ψ,ζ)+ρ10j=1tf0γϵ,ρ,10[gj(x(t|h,p))]dt+ργϵ,ρ,1[Λϵ(h,p,ζ,ψ)],

    where Λϵ(h,p,ζ,ψ):=123ι=1συ=1{αυι(x(tυ|h,p),ζ)+2γϵ,ρ,3σ[βυι(x(tυ|h,p),ζ)]βυι(x(tυ|h,p),ζ)}ψ.

    Since the functions involved with Problem H are smooth, it can be addressed using the gradient-based nonlinear programming (NLP) algorithm. It is evident that the difference between the solution of Problem H and the solution of Problem E is of importance. This section conducts an error analysis comparing Problem H, Problem F, and Problem E. With this in mind, Lemmas 1 and 2 are needed in the process of error analysis.

    Lemma 1. For any ϵ>0, the function γϵ,ρ,κ[η] has the following properties:

    0γϵ,ρ,10[gj(x(t|h,p))]max{gj(x(t|h,p)),0}ϵ20ρ, (3.11)
    0γϵ,ρ,3σ[βυι(x(tυ|h,p),ζ)]max{βυι(x(tυ|h,p),ζ),0}ϵ6σρ, (3.12)
    0γϵ,ρ,1[Λϵ(h,p,ζ,ψ)]max{Λ(h,p,ζ,ψ),0}ϵρ. (3.13)

    Proof. In view of Theorem 3.2.1 in [44], it is easy to prove the validity of (3.11) and (3.12) and

    0γϵ,ρ,1[Λϵ(h,p,ζ,ψ)]max{Λϵ(h,p,ζ,ψ),0}ϵ2ρ. (3.14)

    Now, we will prove (3.13). Based on (3.12), we have

    0Λϵ(h,p,ζ,ψ)Λ(h,p,ζ,ψ)=3ι=1συ=1{γϵ,ρ,3σ[β(xι(tυ|h,p),ζυι)]max{β(xι(tυ|h,p),ζυι),0}}ϵ2ρ, (3.15)

    which implies

    0max{Λϵ(h,p,ζ,ψ),0}max{Λ(h,p,ζ,ψ),0}Λϵ(h,p,ζ,ψ)Λ(h,p,ζ,ψ)ϵ2ρ. (3.16)

    According to (3.14) and (3.16), it follows that

    0γϵ,ρ,1[Λϵ(h,p,ζ,ψ)]max{Λ(h,p,ζ,ψ),0}=γϵ,ρ,1[Λϵ(h,p,ζ,ψ)]max{Λϵ(h,p,ζ,ψ),0}+max{Λϵ(h,p,ζ,ψ),0}max{Λ(h,p,ζ,ψ),0}ϵρ, (3.17)

    which completes the proof. ■

    Lemma 2. For ρ>0 and ϵ>0, it holds that

    0JHρ,ϵ(h,p,ψ,ζ)JFρ(h,p,ψ,ζ)ϵ(1+tf2).

    Proof. Based on Lemma 1, we have

    0JHρ,ϵ(h,p,ψ,ζ)JFρ(h,p,ψ,ζ)=ρ10j=1tf0{γϵ,ρ,10[gj(x(t|h,p))]max{gj(x(t|h,p)),0}}dt+ρ{γϵ,ρ,1[Λϵ(h,p,ζ,ψ)]max{Λ(h,p,ζ,ψ),0}}ϵ(1+tf2). (3.18)

    The proof is complete. ■

    Using Lemmas 1 and 2 as a foundation, we now proceed with the error analysis comparing Problem H, Problem F, and Problem E.

    Theorem 1. Let (h,p,ψ,ζ) and (h,p,ζ,ψ) be the solutions of Problems F and H, respectively. Then,

    0JHρ,ϵ(h,p,ζ,ψ)JFρ(h,p,ψ,ζ)ϵ(1+tf2).

    Proof. In view of Lemma 2,

    0JHρ,ϵ(h,p,ζ,ψ)JFρ(h,p,ζ,ψ)ϵ(1+tf2), (3.19)
    0JHρ,ϵ(h,p,ψ,ζ)JFρ(h,p,ψ,ζ)ϵ(1+tf2). (3.20)

    Because (h,p,ψ,ζ) is the solution of Problem F, we have

    JFρ(h,p,ζ,ψ)JFρ(h,p,ψ,ζ),

    which is equivalent to

    JHρ,ϵ(h,p,ζ,ψ)JFρ(h,p,ζ,ψ)JHρ,ϵ(h,p,ζ,ψ)JFρ(h,p,ζ,ψ). (3.21)

    As (τ,p,θ,η) is the solution of Problem SLI2, it is clear that

    JHρ,ϵ(h,p,ζ,ψ)JHρ,ϵ(h,p,ζ,ψ),

    which is equivalent to

    JHρ,ϵ(h,p,ζ,ψ)JFρ(h,p,ζ,ψ)JHρ,ϵ(h,p,ζ,ψ)JFρ(h,p,ζ,ψ). (3.22)

    Thus, it follows from (3.19)–(3.22) that

    0JHρ,ϵ(h,p,ζ,ψ)JFρ(h,p,ζ,ψ)JHρ,ϵ(h,p,ζ,ψ)JFρ(h,p,ζ,ψ)JHρ,ϵ(h,p,ζ,ψ)JFρ(h,p,ζ,ψ)ϵ(1+tf2).

    The proof is complete. ■

    Definition 2. If gj(x(t|hϵ,pϵ))ϵ,jI10;Λ(hϵ,pϵ,ψϵ,ζϵ)ϵ, then the set of decision variables (hϵ,pϵ,ψϵ,ζϵ) is called ϵfeasible to Problem E.

    Theorem 2. Let (h,p,ψ,ζ) and (h,p,ζ,ψ) be the solution of Problems F and H, respectively. Furthermore, suppose that (h,p,ψ,ζ) is feasible to Problem E and that (h,p,ζ,ψ) is ϵfeasible to Problem E. Then,

    ϵ[20ρtf+tf+2ρ+2]2JE(h,p,ψ,ζ)JE(h,p,ψ,ζ)ϵ(1+tf2).

    Proof. Due to the feasibility of (h,p,ζ,ψ) in Problem E, we get

    max{Λ(h,p,ζ,ψ),0}=0,max{gj(x(t|h,p)),0}dt=0,jI10. (3.23)

    From the ϵfeasibility of (h,p,ζ,ψ) in Problem E, it follows that

    gj(x(t|h,p))ϵ,jI10, (3.24)
    Λ(h,p,ζ,ψ)ϵ. (3.25)

    Based on (3.15) and (3.25), we obtain

    Λϵ(h,p,ζ,ψ)Λ(h,p,ζ,ψ)+ϵ2ρϵ+ϵ2ρ. (3.26)

    From (3.24), (3.26), and the monotonicity of the γϵ,ρ,κ[η] in (3.10), it follows that

    0γϵ,ρ,10[gj(x(t|h,p))]ϵ+ϵ20ρ,jI10;0γϵ,ρ,1[Λϵ(h,p,ζ,ψ)]ϵ+ϵρ,

    which implies

    0ρ10j=1tf0γϵ,ρ,10[gj(x(t|h,p))]dt+ργϵ,ρ,1[Λϵ(h,p,ζ,ψ)]ϵ[20ρtf+tf+2ρ+2]2. (3.27)

    This can be rearranged to yield

    ϵ[20ρtf+tf+2ρ+2]2ρ10j=1tf0γϵ,ρ,10[gj(x(t|h,p))]dtργϵ,ρ,1[Λϵ(h,p,ζ,ψ)]0. (3.28)

    It is clear from Theorem 1 that

    0JE(h,p,ψ,ζ)+ρ10j=1tf0γϵ,ρ,10[gj(x(t|h,p))]dt+ργϵ,ρ,1[Λϵ(h,p,ψ,ζ)]JE(h,p,ζ,ψ)ρ10j=1tf0max{gj(x(t|h,p)),0}dtρmax{Λ(h,p,ζ,ψ),0}ϵ(1+tf2).

    Combining (3.23) and (3.27) gives

    ϵ[20ρtf+tf+2ρ+2]2JE(h,p,ψ,ζ)JE(h,p,ψ,ζ)ϵ(1+tf2).

    The proof is complete. ■

    In the process of the numerical optimization, the gradient formulae of JHρ,ϵ(h,p,ψ,ζ) of Problem H with respect to h,p,ψ, and ζ are required. With these gradient formulae, a gradient-based NLP algorithm can be employed for solving Problem H. Therefore, this section is dedicated to deriving the gradient formulas.

    The partial derivative of x(t|,p) with respect to h is known as the state variation with respect to h.

    Lemma 3. For each pair (h,p)[0,ˉh]×P, it holds that

    x(t|h,p)h=ϖ(t|h,p)=(ϖ1(t|h,p),,ϖ5(t|h,p))T,t(τ1,τ],I2N+1, (3.29)

    where ϖ(t|h,p)R5 is the solution of the following variational system:

    {˙ϖ(t)=f(x(t),x(th),p)x(t)ϖ(t)+f(x(t),x(th),p)x(th)ϖ(th)+f(x(t),x(th),p)x(th)χ(th),t(τ1,τ],I2N+1,ϖ(τ1+)=ϖ(τ),I2N+1,ϖ(t)=0,t0,χ(t)={dϕ(t)dt,t0,f(x(t),x(th),p),t(τ1,τ],I2N+1. (3.30)

    Proof. This theorem can be proven in a manner similar to that used for Theorem 5 in [43]. ■

    To solve Problem , we also require the information of the state variation with respect to .

    Lemma 4. For each pair it holds that

    (3.31)

    where is the solution to the following variational system:

    (3.32)

    Proof. This theorem can be proven in a manner similar to that used for Theorem 5 in [43]. ■

    The state variation with respect to and can be computed through solving system (2.1) together with systems (3.30) and (3.32) forward in time. Then, the gradient formulae of of Problem with respect to and can be easily obtained. By Lemmas 3 and 4 and the chain rule, we have

    Theorem 3. The gradients of the cost function with respect to are given by

    where

    Remark 2. There exist two steps to calculate the gradients of the cost function with respect to : (i) solve systems (2.1), (3.30), and (3.32) using a numerical differential scheme, progressing forward in time to obtain their respective solutions; (ii) compute the gradients based on Theorem 3. When these gradients are derived, a gradient-based NLP method can be applied to solve Problem H.

    The numerical techniques for solving Problem can be divided into two main categories: deterministic methods and intelligent methods [46]. Deterministic methods rely on known information, such as gradients, to solve Problem , and are recognized for their fast convergence. In contrast, intelligent methods often employ probabilistic strategies that are more effective for global optimization. While stochastic methods excel at exploring the global search space, they depend on random search processes, which can be time-consuming.

    Typically, gradient-based methods [46] are applied to local optimization problems, as they are well-suited for refining solutions near the initial starting point. To address issues like getting stuck in local optima and premature convergence, a particle swarm optimization (PSO) algorithm, an example of an intelligent method, has been introduced [41].

    To achieve a balance between local refinement and global exploration, a hybrid optimization algorithm, referred to as Algorithm 1, has been developed to effectively address Problem . In this algorithm, the modified PSO method [41] is employed for the exploration phase, while gradient-based techniques are used for local exploitation, as noted in Line 3 of Algorithm 1. This hybrid approach combines the advantages of the PSO algorithm with the efficiency of gradient-based methods, which leverage the gradient information of the objective function in Problem .

    Let be a sufficiently large number. Let be a predefined parameter used to ensure the termination of the penalty approach. Let and be the initial values of and , respectively. Let be a predefined parameter used to ensure the termination of Algorithm 1. Let be the convergence tolerance of the cost function. Let Let be the initial value of .

    Algorithm 1 Solving Problem H.
    1: Choose initial values of
    2: Set
    3: Solving Problem H with the cost function using the hybrid algorithm designed through the combination of PSO [41] and the gradient-based method [46] developed based on Theorem 3 to obtain the optimal solution, denoted by
    4: Set
    5: if then
    6:  return ;
    7: else
    8:  
    9:  if then
    10:    Set and go to 3;
    11:  else
    12:    
    13:    if then
    14:      Set and go to 3;
    15:    else
    16:      return
    17:    end if
    18:  end if
    19: end if

    Remark 3. Essentially, the goals of Algorithm 1 are for decreasing the value of to approximate the function and increase the value of to ensure that is satisfied.

    The convergence of Algorithm 1 is assured as stated in the following. (i) Theorem 1 guarantees that if the smoothing parameter is small enough, the solution of Problem is an approximate solution of Problem (ii) by virtue of the fundamental principles of the penalty function, when is large enough, any solution to Problem is a feasible solution to Problem [40,45]. Note that in Section 3.1, Problem A can be converted into Problem .

    The selection of hyperparameters is often based on a combination of experience, experimentation, and evaluation metrics. For example, inertia weight (w): typically chosen between 0.4 and 0.9; acceleration factors (): generally set between 1.5 and 2; population size: usually ranges from 20 to 50.

    All computations are performed using MATLAB R2016a on a computer equipped with a 3.70 GHz Intel Core i9-10900K CPU and 32.0 GB of RAM. In the numerical computation, the cultivation conditions have been studied in [16]. During the fed-batch culture, alkali, which is intermittently put in the fermentor to maintain its pH value at approximately 7.0, can cause a chemical reaction with acetic acid. This operation significantly affects the correctness of the extracellular concentrations of acetic acid and ethanol. Therefore, in this paper, the focus is on the discrepancy between the measured output data and the system output for the first three substances. The fed-batch culture evolves starting from the initial states

    The feeding process starts at The feeding time starts at and stops at . They were determined through the experiment reported in [43]. To obtain the value of the cost function (3.9) in Problem F, system (2.1) needs to be solved numerically. Thus, the computational load is rather large. Therefore, for saving the calculation time, the process of the fed-batch culture reported in [47] is grouped into ten stages (i.e., Batch Stage and Stages II–X). In each of Stages II–X, the time duration for each batch stage is 100s minus the duration of the feeding process. The feeding processes in Stages II–X last for 5, 7, 8, 7, 6, 4, 3, 2, 1, 2, and 1 seconds out of every 100 seconds, leaving 95, 93, 92, 93, 96, 97, 98, 99, 98, and 99 seconds for batch processes, respectively.

    Based on [43], the experiment parameters used in the calculation of the solution of system (2.1) are, respectively,

    Euler method is used to solve system (2.1) with a step size of 1/3600. In Algorithm 1, the value of are, respectively, In addition, the value of is selected in its feasible region. The initial function is obtained through the use of the cubic spline method to interpolate the experimental data [40].

    Remark 4. Practitioners of numerical approximation are most concerned with truncation error, which in a numerical method is an error that is caused by using simple approximations to represent exact mathematical formulas. We seek information about errors on both a local and global scale. Local truncation error is the amount of truncation error that occurs in one step of a numerical approximation. Global truncation error is the amount of truncation error that occurs in the use of a numerical approximation to solve a problem.

    We obtain the global truncation error at step n in any numerical approximation of (2.1) by

    The local truncation error for Euler's method is . The local truncation error has two factors of , and we say that it is Based on [48], we have

    where

    We can see from the above description that the error vanishes as . Therefore, it is appropriate to use the Euler method for solving differential equations with which is small enough for the requirement of microbial fermentation simulation.

    By virtue of Algorithm 1, the obtained optimal values of time-delay and system parameters are, respectively, and .

    The optimal time delay in literature [43] was obtained using precise experimental data. In contrast, the optimal time delay presented in this paper is derived under the worst-case distribution of experimental data. Therefore, the optimal time delays obtained in the two cases differ.

    The relative errors between the computed values and the experimental data obtained with and without (denoted by "C" and "B", respectively) and taking into consideration of the worst-case probability distribution of the experimental data (WCPDED) are listed in Table 2, where the relative errors are defined by

    Table 2.  Relative errors between B and C.
    Relative error
    B 10.7042322473258 14.8532582883951 5.64026160583546
    C

     | Show Table
    DownLoad: CSV

    Here, denotes the solution of system (2.1) corresponding to the obtained optimal time-delay and system parameters with the initial concentration and is the experiment data at the time point

    Table 2 reveals that the relative error of 'C' exceeds that of 'B'. This discrepancy can be attributed to the fact that the relative error for 'C' is calculated based on the WCPDED. This WCPDED reflects greater variability and uncertainty of the distribution of the experimental data in the measurements for 'C', which naturally leads to a higher relative error. In contrast, surrounding 'B' without taking into consideration of the WCPDED contributes to its comparatively lower relative error. Understanding these distinctions is crucial for accurately interpreting the reliability of our findings.

    Under the obtained and , the curve of the concentration of the first three substances are drawn in Figures 24 (the horizontal axis stands for time; the vertical axis stands for the concentrations of biomass, glycerol, and 1, 3-PD, respectively; A: experimental data; B: the concentration change curve of biomass, glycerol, and 1, 3-PD without considering the WCPDED, respectively; C: the concentration change curve of biomass, glycerol, and 1, 3-PD that takes into account the WCPDED, respectively). The analysis presented in Figures 24 illustrates a reasonable alignment between the computed values and the experimental data under the WCPDED framework. To facilitate comparison, we determined the relative errors between the computed concentrations of biomass, glycerol, and 1, 3-PD against the experimental results, utilizing the optimal time-delay and system parameters outlined in [43]. It is important to note that, unlike the optimization problem discussed in [43], our approach to parameter estimation relies solely on the first-order moment of the experimental data, thereby simplifying the process. Moreover, Algorithm 1 is designed to operate effectively without requiring precise knowledge of the distribution of the experimental data. The entire calculation process took approximately 15,653 seconds.

    Figure 2.  Comparison of experimental data and computed values of biomass concentrations under the value of and .
    Figure 3.  Comparison of experimental data and computed values of glycerol concentrations under the value of and .
    Figure 4.  Comparison of experimental data and computed values of 1, 3-PD concentrations under the the value of and .

    Figures 24 demonstrate that the NCTSTD system, utilizing the parameters and , effectively captures the dynamics of the fed-batch culture process, even within the constraints of the WCPDED. This suggests that our model not only aligns well with observed behaviors but also provides a robust framework for understanding the intricacies of the cultivation process. The ability of the NCTSTD system to maintain accuracy under these conditions highlights its potential for applications in optimizing bioprocesses.

    Note that our method is under the assumption that the first-order moment information of the experimental data can be accurately obtained. Obviously, it will be practically important to extend this method to cases where moment information is also uncertain. This will be an interesting further research direction.

    This paper studies an NCTSTD with unknown quantities, in which the measured output data is uncertain. The focus is to identify these unknown quantities. For this, we formulate the problem as a DRPE problem subject to the NCTSTD system. A hybrid algorithm is presented to solve the DRPE problem. Finally, numerical results are obtained, which demonstrate the validity of the designed algorithm.

    On the other hand, it is important to note that the designed algorithm is based on the assumption that the first-order moment information of the measured output data with respect to its probability distribution is available, highlighting the need to extend the algorithm to cases where this moment information is uncertain. This would be an interesting further research area.

    In summary, the designed algorithm provides a promising method to optimally identify time-delay and system parameters in the fed-batch culture process of glycerol bioconversion to 1, 3-PD under the environment of uncertain measured output data. The potential extension of this algorithm to more general cases would enhance its scope of applications.

    The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.

    This work was supported in part by the National Key Research and Development Program of China under Grant 2022YFB3304600; in part by the Nature Science Foundation of Liaoning Province of China under Grant 2024-MS-015; in part by the Fundamental Research Funds for the Central Universities under Grant 3132024196, and Grant DUT22LAB305; in part by the National Natural Science Foundation of China under Grant 12271307, and Grant 12161076; in part by the China Postdoctoral Science Foundation under Grant 2019M661073; in part by the Xinghai Project of Dalian Maritime University.

    The authors declare there is no conflicts of interest.



    [1] W. Liu, L. Yang, B. Yu, Kernel density estimation based distributionally robust mean-CVaR portfolio optimization, J. Glob. Optim., 84 (2022), 1053–1077. https://doi.org/10.1007/s10898-022-01177-5 doi: 10.1007/s10898-022-01177-5
    [2] E. Delage, Y. Ye, Distributionally robust optimization under moment uncertainty with application to data-driven problems, Oper. Res., 58 (2010), 595–612. https://doi.org/10.1287/opre.1090.0741 doi: 10.1287/opre.1090.0741
    [3] C. Peng, E. Delage, Data-driven optimization with distributionally robust second order stochastic dominance constraints, Oper. Res., 72 (2024), 1298–1316. https://doi.org/10.1287/opre.2022.2387 doi: 10.1287/opre.2022.2387
    [4] S. Wang, L. Pang, H. Guo, H. Zhang, Distributionally robust optimization with multivariate second-order stochastic dominance constraints with applications in portfolio optimization, Optimization, 72 (2023), 1839–1862. https://doi.org/10.1080/02331934.2022.2048382 doi: 10.1080/02331934.2022.2048382
    [5] X. Tong, M. Li, H. Sun, Decision bounding problems for two-stage distributionally robust stochastic bilevel optimization, J. Glob. Optim., 87 (2023), 679–707. https://doi.org/10.1007/s10898-022-01227-y doi: 10.1007/s10898-022-01227-y
    [6] H. Xu, Y. Liu, H. Sun, Distributionally robust optimization with matrix moment constraints: Lagrange duality and cutting plane methods, Math. Program., 169 (2018), 489–529. https://doi.org/10.1007/s10107-017-1143-6 doi: 10.1007/s10107-017-1143-6
    [7] B. Li, Y. Rong, J. Sun, K. L. Teo, A distributionally robust linear receiver design for multi-access space-time block coded MIMO systems, IEEE Trans. Wireless Commun., 16 (2017), 464–474. https://doi.org/10.1109/TWC.2016.2625246 doi: 10.1109/TWC.2016.2625246
    [8] B. Li, Y. Tan, A. Wu, G. Duan, A distributionally robust optimization based method for stochastic model predictive control, IEEE Trans. Autom. Control, 67 (2022), 5762–5776. https://doi.org/10.1109/TAC.2021.3124750 doi: 10.1109/TAC.2021.3124750
    [9] A. Shapiro, Distributionally robust modeling of optimal control, Oper. Res. Lett., 50 (2022), 561–567. https://doi.org/10.1016/j.orl.2022.08.002 doi: 10.1016/j.orl.2022.08.002
    [10] B. P. Van Parys, D. Kuhn, P. J. Goulart, M. Morari, Distributionally robust control of constrained stochastic systems, IEEE Trans. Autom. Control, 61 (2015), 430–442. https://doi.org/10.1109/TAC.2015.2444134 doi: 10.1109/TAC.2015.2444134
    [11] A. Hakobyan, I. Yang, Wasserstein distributionally robust control of partially observable linear stochastic systems, IEEE Trans. Autom. Control, 69 (2024), 6121–6136. https://doi.org/10.1109/TAC.2024.3394348 doi: 10.1109/TAC.2024.3394348
    [12] B. Li, T. Guan, L. Dai, G. R. Duan. Distributionally robust model predictive control with output feedback. IEEE Trans. Autom. Control, 69 (2024), 3270–3277. https://doi.org/10.1109/TAC.2023.3321375
    [13] L. Huang, X. Zhou, N. Wu, Y. Sun, Z. Xiu, Selective extraction of 1, 3-propanediol by phenylboronic acid-based ternary extraction system, J. Chem. Technol. Biotechnol., 99 (2024), 1530–1540. https://doi.org/10.1002/jctb.7647 doi: 10.1002/jctb.7647
    [14] C. Groeger, W. Sabra, A. Zeng, Simultaneous production of 1, 3-propanediol and n-butanol by Clostridium pasteurianum: In situ gas stripping and cellular metabolism, Eng. Life Sci., 16 (2016), 664–674. https://doi.org/10.1002/elsc.201600058 doi: 10.1002/elsc.201600058
    [15] J. Zhou, J. Shen, X. Wang, Y. Sun, Z. Xiu, Metabolism, morphology and transcriptome analysis of oscillatory behavior of Clostridium butyricum during long-term continuous fermentation for 1, 3-propanediol production, Biotechnol. Biofuels, 13 (2020), 191. https://doi.org/10.1186/s13068-020-01831-8 doi: 10.1186/s13068-020-01831-8
    [16] S. Duan, Z. Zhang, X. Wang, Y. Sun, Y. Dong, L. Ren, et al., Co-production of 1, 3-propanediol and phage phiKpS2 from the glycerol fermentation by Klebsiella pneumoniae, Bioresour. Bioprocess., 11 (2024), 44. https://doi.org/10.1186/s40643-024-00760-w doi: 10.1186/s40643-024-00760-w
    [17] L. Wang, J. Yuan, C. Wu, X. Wang, Practical algorithm for stochastic optimal control problem about microbial fermentation in batch culture, Optim. Lett., 13 (2019), 527–541. https://doi.org/10.1007/s11590-017-1220-z doi: 10.1007/s11590-017-1220-z
    [18] J. Wang, X. Zhang, J. Ye, J. Wang, E. Feng, Optimizing design for continuous conversion of glycerol to 1, 3-propanediol using discrete-valued optimal control, J. Process Control, 104 (2021), 126–134. https://doi.org/10.1016/j.jprocont.2021.06.010 doi: 10.1016/j.jprocont.2021.06.010
    [19] J. Yuan, S. Lin, S. Zhang, C. Liu, Distributionally robust system identification for continuous fermentation nonlinear switched system under moment uncertainty of experimental data, Appl. Math. Modell., 127 (2024), 679–695. https://doi.org/10.1016/j.apm.2023.12.023 doi: 10.1016/j.apm.2023.12.023
    [20] J. Yuan, C. Wu, C. Liu, K. L. Teo, J. Xie, Robust suboptimal feedback control for a fed-batch nonlinear time-delayed switched system, J. Franklin Inst., 360 (2023), 1835–1869. https://doi.org/10.1016/j.jfranklin.2022.12.027 doi: 10.1016/j.jfranklin.2022.12.027
    [21] C. Liu, G. Shi, G. Liu, D. Hu, Optimal control of a nonlinear state-dependent impulsive system in fed-batch process, Int. J. Biomath., 16 (2023), 2350001. https://doi.org/10.1142/S1793524523500018 doi: 10.1142/S1793524523500018
    [22] T. Niu, J. Zhai, H. Yin, E. Feng, C. Liu, The uncoupled microbial fed-batch fermentation optimization based on state-dependent switched system, Int. J. Biomath., 14 (2021), 2150025. https://doi.org/10.1142/S179352452150025X doi: 10.1142/S179352452150025X
    [23] X. Li, M. Sun, Z. Gong, E. Feng, Multistage optimal control for microbial fed-batch fermentation process, J. Ind. Manage. Optim., 18 (2022), 1709–1721. https://doi.org/10.3934/jimo.2021040 doi: 10.3934/jimo.2021040
    [24] T. Niu, J. Zhai, H. Yin, E. Feng, Optimal control of nonlinear switched system in an uncoupled microbial fed-batch fermentation process, J. Franklin Inst., 355 (2018), 6169–6190. https://doi.org/10.1016/j.jfranklin.2018.05.012 doi: 10.1016/j.jfranklin.2018.05.012
    [25] C. Zhang, S. Sharma, W. Wang, A. Zeng, A novel downstream process for highly pure 1, 3-propanediol from an efficient fed-batch fermentation of raw glycerol by Clostridium pasteurianum, Eng. Life Sci., 21 (2021), 351–363. https://doi.org/10.1002/elsc.202100012 doi: 10.1002/elsc.202100012
    [26] J. Yuan, S. Zhao, D. Yang, C. Liu, C. Wu, T. Zhou, et al., Koopman modeling and optimal control for microbial fed-batch fermentation with switching operators, Nonlinear Anal. Hybrid Syst., 52 (2024), 101461. https://doi.org/10.1016/j.nahs.2023.101461 doi: 10.1016/j.nahs.2023.101461
    [27] J. Yuan, C. Wu, Z. Liu, S. Zhao, C. Yu, K. Teo, et al., Koopman modeling for optimal control of the perimeter of multi-region urban traffic networks, Appl. Math. Modell., 138 (2025), 115742. https://doi.org/10.1016/j.apm.2024.115742 doi: 10.1016/j.apm.2024.115742
    [28] D. Wu, Y. Bai, C. Yu, A new computational approach for optimal control problems with multiple time-delay, Automatica, 101 (2019), 388–395. https://doi.org/10.1016/j.automatica.2018.12.036 doi: 10.1016/j.automatica.2018.12.036
    [29] M. Wang, N. Liu, Qualitative analysis and traveling wave solutions of a predator-prey model with time delay and stage structure, Electron. Res. Arch., 32 (2024), 2665–2698. https://doi.org/10.3934/era.2024121 doi: 10.3934/era.2024121
    [30] J. Yuan, C. Wu, K. Teo, S. Zhao, L. Meng, Perimeter control with state-dependent delays: optimal control model and computational method, IEEE Trans. Intell. Transp. Syst., 23 (2022), 20614–20627. https://doi.org/10.1109/TITS.2022.3179729 doi: 10.1109/TITS.2022.3179729
    [31] D. Wu, Y. Chen, C. Yu, Y. Bai, K. Teo, Control parameterization approach to time-delay optimal control problems: A survey, J. Ind. Manage. Optim., 19 (2023), 3750–3783. https://doi.org/10.3934/jimo.2022108 doi: 10.3934/jimo.2022108
    [32] Y. Chen, X. Zhu, C. Yu, K. Teo, Sequential time scaling transformation technique for time-delay optimal control problem, Commun. Nonlinear Sci. Numer. Simul., 133 (2024), 107988. https://doi.org/10.1016/j.cnsns.2024.107988 doi: 10.1016/j.cnsns.2024.107988
    [33] Z. Li, L. Pei, G. Duan, S. Chen, A non-autonomous time-delayed SIR model for COVID-19 epidemics prediction in China during the transmission of Omicron variant. Electron. Res. Arch., 32 (2024), 2203–2228. https://doi.org/10.3934/era.2024100
    [34] J. Yuan, C. Wu, K. Teo, J. Xie, S. Wang, Computational method for feedback perimeter control of multiregion urban traffic networks with state-dependent delays, Transp. Res. Part C Emerging Technol., 153 (2023), 104231. https://doi.org/10.1016/j.trc.2023.104231 doi: 10.1016/j.trc.2023.104231
    [35] J. Yuan, D. Yang, D. Xun, K. Teo, C. Wu, A. Li, et al., Sparse optimal control of cyber-physical systems via PQA approach, Pac. J. Optim., Accepted, 2024.
    [36] C. Liu, R. Ryan, Q. Lin, K. Teo, Dynamic optimization for switched time-delay systems with state-dependent switching conditions, SIAM J. Control Optim., 56 (2018), 3499–3523. https://doi.org/10.1137/16M1070530 doi: 10.1137/16M1070530
    [37] C. Liu, C. Sun, Robust parameter identification of a nonlinear impulsive time-delay system in microbial fed-batch process, Appl. Math. Modell., 111 (2022), 160–175. https://doi.org/10.1016/j.apm.2022.06.032 doi: 10.1016/j.apm.2022.06.032
    [38] X. Gao, J. Zhai, E. Feng, Multi-objective optimization of a nonlinear switched time-delay system in microbial fed-batch process, J. Franklin Inst., 357 (2020), 12609–12639. https://doi.org/10.1016/j.jfranklin.2020.07.036 doi: 10.1016/j.jfranklin.2020.07.036
    [39] C. Liu, M. Han, Time-delay optimal control of a fed-batch production involving multiple feeds, Discrete Contin. Dyn. Syst. - S, 13 (2020), 1697–1709. https://doi.org/10.3934/dcdss.2020099 doi: 10.3934/dcdss.2020099
    [40] J. Nocedal, S. J. Wright, Numerical Optimization, Springer, New York, 2006. https://doi.org/10.1007/978-0-387-40065-5
    [41] A. Banerjee, I. Abu-Mahfouz, A comparative analysis of particle swarm optimization and differential evolution algorithms for parameter estimation in nonlinear dynamic systems, Chaos, Solitons Fractals, 58 (2014), 65–83. https://doi.org/10.1016/j.chaos.2013.11.004 doi: 10.1016/j.chaos.2013.11.004
    [42] E. Anderson, P. Nash, Linear Programming in Infinite-Dimensional Spaces: Theory and Applications, Wiley, Chichester, United Kingdom, 1987. https://cir.nii.ac.jp/crid/1130000796325253632
    [43] C. Liu, Sensitivity analysis and parameter identification for a nonlinear time-delay system in microbial fed-batch process, Appl. Math. Modell., 38 (2014), 1449–1463. https://doi.org/10.1016/j.apm.2013.07.039 doi: 10.1016/j.apm.2013.07.039
    [44] P. Liu, X. Li, X. Liu, Y. Hu, An improved smoothing technique-based control vector parameterization method for optimal control problems with inequality path constraints, Optim. Control. Appl. Methods, 38 (2017), 586–600. https://doi.org/10.1002/oca.2273 doi: 10.1002/oca.2273
    [45] X. Wu, J. Lin, K. Zhang, M. Cheng, A penalty function-based random search algorithm for optimal control of switched systems with stochastic constraints and its application in automobile test-driving with gear shifts, Nonlinear Anal. Hybrid Syst., 45 (2022), 101218. https://doi.org/10.1016/j.nahs.2022.101218 doi: 10.1016/j.nahs.2022.101218
    [46] K. Teo, B. Li, C. Yu, V. Rehbock, Applied and Computational Optimal Control: A Control Parametrization Approach, Springer, 2021. https://doi.org/10.1007/978-3-030-69913-0
    [47] X. Chen, D. J. Zhang, W. Qi, S. Gao, Z. Xiu, P. Xu, Microbial fed-batch production of 1, 3-propanediol by klebsiella pneumoniae under microaerobic conditions, Appl. Microbiol. Biotechnol., 63 (2003), 143–146. https://doi.org/10.1007/s00253-003-1369-5 doi: 10.1007/s00253-003-1369-5
    [48] K. Atkinson, W. Han, D. Stewart, Numerical Solution of Ordinary Differential Equations, John Wiley and Sons, 2009. http://homepage.math.uiowa.edu/atkinson/papers/NAODE_Book.pdf
  • Reader Comments
  • © 2024 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(532) PDF downloads(34) Cited by(0)

Figures and Tables

Figures(4)  /  Tables(2)

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog