Mifeng Ren

The inherent randomness, fluctuation, and intermittence of photovoltaic power generation make it difficult to track the scheduling plan. To improve the ability to track the photovoltaic plan to a greater extent, a real-time charge and discharge power control method based on deep reinforcement learning is proposed. Firstly, the photovoltaic and energy storage hybrid system and the mathematical model of the hybrid system are briefly introduced, and the tracking control problem is defined. Then, power generation plans on different days are clustered into four scenarios by the K-means clustering algorithm. The mean, standard deviation, and kurtosis of the power generation plant are used as the features. Based on the clustered results, the state, action, and reward required for reinforcement learning are set. In the constraint conditions of various variables, to increase the accuracy of the hybrid system for tracking the new generation schedule, the proximal policy optimization (PPO) algorithm is used to optimize the charging/discharging power of the energy storage system (ESS). Finally, the proposed control method is applied to a photovoltaic power station. The results of several valid experiments indicate that the average errors of tracking using the Proportion Integral Differential (PID), model predictive control (MPC) method, and the PPO algorithm in the same condition are 0.374 MW, 0.609 MW, and 0.104 MW, respectively, and the computing time is 1.134 s, 2.760 s, and 0.053 s, respectively. The consequence of these indicates that the proposed deep reinforcement learning-based control strategy is more competitive than the traditional methods in terms of generalization and computation time.

Unbalanced data with very few samples for special abnormal conditions frequently occur in actual production processes, which can make accurate monitoring of the process state challenging. This paper proposes a multi-modal few-shot learning method (MMFSL) within a fault diagnosis framework for unbalanced data modelling of industrial bearings. MMFSL can handle two modes of data and therefore contains two data generation channels. The first channel deals with time series data and the second deals with images. The two modes of generated data are then evaluated using the MMFSL evaluation modules, IEI (image evaluation index) and TEI (time series evaluation index), to guarantee the quality of the generated data. In addition, the time series data are analyzed in the time-frequency domain to ensure the consistency of the frequency distribution. Original unbalanced and insufficient data samples are expanded so that the fault diagnosis model can be trained adequately. Through experiments using data from Case Western Reserve University(CWRU) and Pardborn University(PU), generated data by using the MMFSL model can increase fault detection accuracy from 80.9% to 97.5%. At the same time, false alarms rate can be reduced significantly from 4.6% to 0.1%. Moreover, fault classification model accuracy can reach 98.1% compared to 95.4% for the original time series dataset, and 99.3% compared with 96.5% for the image data extension. Various visualizations of comparison results are also provided to show the reliability of the MMFSL framework.

The molecular weight distribution is an important factor that affects the properties of polymers. A control algorithm based on the moment-generating function was proposed to regulate the molecular weight distribution for polymerization processes in this work. The B-spline model was used to approximate the molecular weight distribution, and the weight state space equation of the system was identified by the subspace state space system identification method based on the paired date of B-spline weights and control inputs. Then, a new performance criterion mainly consisting of the moment-generating function was constructed to obtain the optimal control input. The effectiveness of the proposed control method was tested in a styrene polymerization process. The molecular weight distribution of the styrene polymers can be approximated by the B-spline model effectively, and it can also be regulated towards the desired one under the proposed control method.

SPIE Digital Library Proceedings

The emission of NOx in thermal power plants would cause serious pollution to the environment, and effective prediction of NOx has important significance. With the development of deep learning, many methods can effectively solve the prediction problem. In this paper, a long short-term memory (LSTM) neural network based on the generalized correntropy (GC) loss function, which is called the GC-LSTM neural network here, is proposed to predict the NOx emission. In order to make the characteristics of high-dimensional data have the same measurement scale, normalization is used to preprocess data firstly. Then, the processed historical data is used as training data to train the GC-LSTM neural network. After that, the emission of the NOx at each time instant can be predicted according to the online measured relevant data based on the established GC-LSTM neural network. Finally, prediction results under the LSTM neural network with GC, mean square error (MSE) and the mean absolute error (MAE) loss functions are given. And according to the simulation results, it can be seen that the loss function curve of the LSTM network based on GC is more stable and has better convergence. It can be seen that the proposed GC-LSTM prediction is more accurate, by comparing the probability density function (PDF) and root mean square error (RMSE) of the error between the actual values and the prediction values.