Jiandong Wang’s research while affiliated with Shandong University of Science and Technology and other places

In gas-to-methanol processes, optimizing multi-energy systems is a critical challenge toward efficient energy allocation. This paper proposes an entropy-based stochastic optimization method for a multi-energy system in a gas-to-methanol process, aiming to achieve optimal allocation of gas, steam, and electricity to ensure executability under modeling uncertainties. First, mechanistic models are developed for major chemical equipments, including the desulfurization, steam boilers, air separation, and syngas compressors. Structural errors in these models under varying operating conditions result in noticeable model uncertainties. Second, Bayesian estimation theory and the Markov Chain Monte Carlo approach are employed to analyze the differences between historical data and model predictions under varying operating conditions, thereby quantifying modeling uncertainties. Finally, subject to constraints in the model uncertainties, equipment capacities, and energy balance, a multi-objective stochastic optimization model is formulated to minimize gas loss, steam loss, and operating costs. The entropy weight approach is then applied to filter the Pareto front solution set, selecting a final optimal solution with minimal subjectivity and preferences. Case studies using Aspen Hysys-based simulations show that optimization solutions considering model uncertainties outperform the counterparts from a standard deterministic optimization in terms of executability.

Continuous-time dynamic models are indispensable for many disciplines of science and engineering. This paper proposes a new approach for estimating unknown parameters of continuous-time dynamic models. The proposed approach is composed of three main steps: a searching grid of model parameters is formulated in certain step sizes; objective functions of all grid points are calculated; optimal model parameters are found as the ones corresponding to the minimum value of the objective function. Compared to existing identification approaches, the proposed approach has one new feature that model uncertainties are measured based on the fuzzy set theory by a number of companion model parameters that are associated with objective functions close to the minimum one. The proposed approach does not require a restrictive assumption for existing approaches that the true unknown system must be enclosed by the model set being considered, and provides model uncertainties in the presence of model structure errors. The main obstacle of the grid search is a high computation cost in calculating objective functions for all grid points. This obstacle is overcome to an acceptable level via parallel computation using a number of CPUs in multiple computers. The proposed approach is validated and compared with the existing approach through numerical and experimental examples.

Nonlinear flow characteristics of main steam regulating valves play an important role on the performance of primary frequency control in thermal power generation units. Manual testing is a traditional method to capture nonlinear flow characteristics by maintaining constant steam pressures, changing control valve openings, and observing power output changes; however, such a method disturbs normal operations of power generation units. This article proposes a new method to identify nonlinear flow characteristics of main steam regulating valves by exploiting special segments hidden in historical operating data. The special segments refer to steady-state segments with constant amplitudes and slope-response segments with large amplitude changes, both of which can be automatically extracted from historical operating data. Relationships between the special segments and a nonlinear model of flow characteristics are theoretically established, and unknown model parameters are estimated by a linear dynamic programming algorithm. These special segments can separate the nonlinear flow characteristics of regulating valves from dynamic effects of steam turbines and generators at different operating points. The necessity of such a separation is demonstrated by a comparison with the Hammerstein model identification method and the sparse identification method. The identified model can be visually verified by comparing measured outputs in the extracted special segments with simulated model outputs. Industrial case studies demonstrate the effectiveness of the proposed method.

Intermittent heating refers to heating buildings only during occupied periods. Owing to the thermal inertia of buildings, a preheating time is necessary to bring the indoor temperature into a comfortable range at the start of occupied periods. This paper proposes a new method to estimate the preheating time in intermittent heating with hot-water radiators by considering model uncertainties. First, a dynamic physical model is established based on the thermal equilibrium of buildings, and is approximated as a first-order continuous-time model by ignoring the influence of ambient conditions on the indoor temperature. The ignored part is represented as model uncertainties. Second, the continuous-time model is identified from collected data of indoor temperature and supply water temperature, and the model uncertainty is obtained based on the identified model. Third, the preheating time is determined with a credible probability to ensure that the indoor temperature meet thermal comfortable requirements. The proposed method has two novel features: (i) it can be applied to buildings with unknown physical parameters and unmeasurable ambient conditions, and (ii) it can yield reliable estimates of preheating time based on the model uncertainties to ensure the success of preheating under a given probability. Experimental examples demonstrate the effectiveness of the proposed method.

Delay timers are widely used in practice to remove nuisance alarms. This paper proposes a method to design delay timers based on probability mass functions (PMFs) of alarm durations. The main rationale is that all nuisance alarms with alarm durations less than m samples are removed by a m-sample delay timer. The proposed method is composed of three steps. First, distribution change points of alarm durations are detected to separate consecutive samples of alarm durations into segments. Second, a reliable PMF estimate of alarm durations is obtained via the Bayesian estimation for each segment. Third, an optimal value of m is designed based on the estimated PMFs as the smallest one to make a designing index, being formulated as the percentage of false alarm occurrences to be unremoved, less than an upper bound. Contemporary methods are confined to analog process variables, which are usually assumed to be independently and identically distributed (IID) and whose probability density functions are exploited to design delay timers. In contrast, the proposed method does not require the IID assumption, and is equally applicable to both analog and digital process variables. Numerical and industrial examples are provided to support the proposed method.