Complex Nonlinear Function Research Articles

Support vector regression model for the prediction of buildings’ maximum seismic response based on real monitoring data

Maximum drift ratio (MDR), one of the engineering demand parameters (EDPs), provides fundamental physical value for predicting building damage. Existing machine learning based prediction models mainly rely on numerical simulation data or structural experiments and are not appropriate for prediction of seismic response of real structures. The New Earthquake Data (NDE1.0) is the most comprehensive publicly available dataset of actual structural seismic response observations. Currently the prediction models using NDE1.0 are mainly based on linear or log-linear regression. In this study, based on the NDE1.0 flatfile, we develop a full-feature support vector regression (SVR) based MDR prediction model (SVR-MDR), treating all the available 41 characteristic parameters including structural information as input feature. To improve the model’s efficiency and practical applicability, we also establish a reduced-feature SVR model (RSVR-MDR) by selecting 10 fundamental parameters based on SHapley Additive exPlanations (SHAP) values and the accessibility of features. Our results demonstrate that SVR-MDR model outperform other machine learning models such as kernel ridge regression and decision tree models, and SVR-MDR and RSVR-MDR models outperform conventional loglinear regression and multinomial models, because SVR can map the complex nonlinear function of multiple variables and consider the available information of buildings especially the fundamental frequency. The proposed RSVR-MDR model have promising potential application for post-event seismic damage assessment and post-event emergency response in near real time.

Machine Learning for Cell Design and Optimization

An essential aspect of battery research and development involves exploring various cell parameters to deliver optimal performance. Battery optimization is challenging due to the huge cost and time required to evaluate different configurations in experiments or simulations. Optimizing the cycling performance is especially costly since battery cycling is extremely time consuming. Adding to this challenge is the expansive parameter space. In this talk, I will present some of our recent works that leverage machine learning and physics-informed machine learning for cell optimization. I will first show an unsupervised machine learning approach that can reduce the battery testing time by about 75% [1]. This framework comprises two main components: a pruner and a sampler. The pruner employs the Asynchronous Successive Halving Algorithm and Hyperband to halt the cycling of unpromising batteries, thereby conserving resources for further exploration. Meanwhile, the sampler, utilizing Tree of Parzen Estimators, predicts promising configurations for future cycles based on query history. Notably, our framework accommodates categorical, discrete, and continuous parameters and can operate asynchronously in parallel to facilitate multiple simultaneous cycling cells. Next, I will show a method to embed physical laws and on-line observation into machine learning [2], so that irrelevant low-cost battery data can be utilized to identify battery parameters by machine learning without knowledge of their ground truth as the training data. We take diffusivity as an example and show that it can be obtained from easily measured sequence of battery voltage over time. Our results show that this method accurately quantifies not only the diffusivities of both positive and negative electrodes, but also as complex non-linear functions of lithium concentration, purely based on the cell voltage data requiring neither diffusivity nor concentration measurement. Notably, it can accurately predict non-monotonic, many-to-one relations such as “w” shape functions. Moreover, this method is immune to measurement noise and capable of simultaneously estimating multiple parameters. Finally, I will report a new approach of Self-directed Online Learning Optimization (SOLO) [3] for cell optimization, which integrates Deep Neural Network (DNN) with Finite Element electrochemical calculations. A DNN learns and substitutes the objective as a function of design variables. A small number of training data is generated dynamically based on the DNN’s prediction of the optimum. The DNN adapts to the new training data and gives better prediction in the region of interest until convergence. The optimum predicted by the DNN is proved to converge to the true global optimum through iterations. Our algorithm reduced the computational time by up to 5 orders of magnitude compared with directly using heuristic methods, and outperformed all state-of-the-art algorithms tested in our experiments. Broadly, we would like to emphasize that integrating knowledge into machine learning [4] can lead to significant advances in cell design. REFERENCES:[1] C. Deng, A. Kim, and W. Lu, “A generic battery-cycling optimization framework with learned sampling and early stopping strategies,” Patterns, 3, 100531 (2022). https://doi.org/10.1016/j.patter.2022.100531[2] B. Wu, B. Zhang, C. Deng, and W. Lu, “Physics-encoded deep learning in identifying battery parameters without direct knowledge of ground truth,” Applied Energy, 321, 119390 (2022). https://doi.org/10.1016/j.apenergy.2022.119390[3] C. Deng, Y. Wang, C. Qin, Y. Fu, and W. Lu, “Self-directed online machine learning for topology optimization,” Nature Communications, 13, 388 (2022). https://doi.org/10.1038/s41467-021-27713-7[4] C. Deng, X. Ji, C. Rainey, J. Zhang, and W. Lu, “Integrating machine learning with human knowledge,” iScience, 23, 101656 (2020). https://doi.org/10.1016/j.isci.2020.101656

A Comprehensive Review of Artificial Intelligence and Machine Learning in Control Theory

Traditional control methods, such as proportional-integral-derivative (PID) controllers and linear-quadratic regulators (LQRs), have proven effective for linear and well-modeled systems. However, these methods often perform poorly in nonlinear, complex and dynamic environments. The paper aims to investigate the modern control systems by integrating artificial intelligence (AI) techniques, such as machine learning (ML), reinforcement learning (RL), deep learning, and fuzzy logic, to enhance their adaptive, robust, and predictive capabilities. And it reviews the literature and analyzes AI integration in control systems. The proposed strategies include supervised learning for trajectory optimization and fault detection, reinforcement learning for optimal control in dynamic environments, neural networks for complex nonlinear function approximation, and fuzzy logic for handling uncertainty and imprecise inputs. AI techniques significantly enhance the ability to tackle nonlinear problems and dynamic changes, demonstrating superior performance in applications like self-driving cars adapting to various road conditions and optimal energy distribution in smart grids. Despite the challenges of computational complexity, scalability, and the safety and reliability in the implementation of interpretable AI models, this paper suggests that hybrid approaches combining traditional control and AI techniques, along with the evolution of interpretable AI and convergence with quantum control, hold great promise for advancing AI-driven control systems.

Predictive modeling of flexible EHD pumps using Kolmogorov–Arnold Networks

We present a novel approach to predicting the pressure and flow rate of flexible electrohydrodynamic pumps using the Kolmogorov–Arnold Network. Inspired by the Kolmogorov–Arnold representation theorem, KAN replaces fixed activation functions with learnable spline-based activation functions, enabling it to approximate complex nonlinear functions more effectively than traditional models like Multi-Layer Perceptron and Random Forest. We evaluated KAN on a dataset of flexible EHD pump parameters and compared its performance against RF, and MLP models. KAN achieved superior predictive accuracy, with Mean Squared Errors of 12.186 and 0.012 for pressure and flow rate predictions, respectively. The symbolic formulas extracted from KAN provided insights into the nonlinear relationships between input parameters and pump performance. These findings demonstrate that KAN offers exceptional accuracy and interpretability, making it a promising alternative for predictive modeling in electrohydrodynamic pumping.

Identification of nonlinear system model and inverse model based on conditional invertible neural network

In applications such as adaptive inverse control, internal model control, and active noise control, the identification accuracy of the system model and the inverse model directly affects the performance. However, it is not easy to identify inverse models for nonlinear systems. Moreover, existing methods require two identification calculations to obtain the system model and the inverse model. Therefore, an identification method of nonlinear system model and inverse model based on conditional invertible neural network (cINN) is proposed. The invertible structure of cINN enables simultaneous approximation of complex nonlinear functions and simultaneous acquisition of the corresponding inverse functions. Consequently, both the nonlinear system model and the inverse model can be identified concurrently through cINN. Moreover, the identification performance of the cINN-based method is validated and applied to disturbance cancellation in a classical nonlinear simulation system. Finally, the inverse model of the actuator is identified by cINN, and the inverse model is applied to the nonlinear compensation of the actuator.

State-dependent multi-agent discrete event simulation for urban rail transit passenger flow

Urban rail transit passenger flow modeling is the foundation of urban rail transit planning, design, and operation. The motivation of this paper is to accurately and efficiently simulate passenger flow dynamics in urban rail transit systems. To this end, we propose a State-dependent Multi-agent Discrete Event Simulation (SdMaDES). The state-dependence (or congestion-dependence) means that the service abilities (or travel times) of bottleneck facilities (e.g. platform screen doors and transfer corridors) are not constant and they interact with the congestion dynamics. Specifically, we first establish a modular Multi-agent Discrete Event Simulation (MaDES), which includes four types of modules (passenger, train, station, and network modules) and three types of agents (passenger, train, and station agents). The logical connections and event-triggering modes between the modules are analyzed and defined, and thirty types of events are designed. The state-dependence is then captured by an Improved Social Force Model (ISFM), which adds an autonomous obstacle avoidance mechanism. The ISFM reproduces passenger movement behavior within bottleneck facilities of urban rail transit systems at the microscopic level. These state-dependent functions or general rules are explicitly formulated by fitting the results of ISFM and are subsequently applied to the proposed MaDES model, resulting in the SdMaDES. This integration aims to enhance the accuracy of the simulation. We conducted a real case from the Chengdu Metro network. Some interesting results are found. (a) The maximum number of boarding passengers in a train carriage is a complex nonlinear function that is dependent on the state (density inside the train carriage). This challenges the linear function commonly utilized in most studies. (b) Compared to the actual data, the proposed SdMaDES model shows a cumulative error of 9.85% after data smoothing, while the conventional MaDES model exhibits a much higher cumulative error of 21.9% after data smoothing. (c) As the overall traffic demand level increases, the gap between the two simulation models’ results is getting wider and wider due to the amplified nonlinear impact of congestion.

Optimizing the LQ45 Stock portfolio using Piecewise Linear Function: A Case Study from an Investor's Point of View

In optimization problem solving, both linear and nonlinear approaches can be used, with nonlinear programming considering constraints or not. One effective method of nonlinear programming is the Piecewise Linear approach, which breaks down complex nonlinear functions into straight-line segments to make their solution easier. This method can be applied in the financial sphere, such as in stock investment. This study discusses the application of piecewise linear function in optimizing investment portfolios in Bank Jago Tbk. (ARTO), Barito Pacific Tbk. (BRPT), and Go To Gojek Tokopedia Tbk. (GOTO) stocks. The purpose of this study is to provide insight that Piecewise Linear can provide optimal solutions in managing investment portfolios by calculating the risks and returns of selected stocks. The results showed a risk with a level as big as 0.01, the expected profit from the investment of the three stocks in the calculation period reached Rp 1,803,109. On the other hand, for investors who are more cautious and have a Level as big as 1, the anticipated profit in the same investment is around Rp 1,275,052.

Blind CFO estimation based on weighted subspace fitting criterion with fuzzy adaptive gravitational search algorithm

AbstractThis paper deals with the blind carrier frequency offset (CFO) estimation based on weighted subspace fitting (WSF) criterion with fuzzy adaptive gravitational search algorithm (GSA) for the interleaved orthogonal frequency-division multiplexing access (OFDMA) uplink system. For the CFO estimation problem, it is well known that the WSF has superior statistical characteristics and better estimation performance. However, the type of CFO estimation must pass through the high-dimensional space problem. Optimizing complex nonlinear multimodal functions requires a large computational load, which is difficult and not easy to maximize or minimize nonlinear cost functions in large parameter spaces. This paper firstly presents swarm intelligence (SI) optimization algorithms such as GSA, particle swarm optimization (PSO), and hybrid PSO and GSA (PSOGSA) to improve estimation accuracy and reduce the computational load of search. At the same time, this paper also integrates a fuzzy inference system to WSF-GSA to dynamically adjust the gravitational constant, which can not only reduce the searching computational load, but also improve the performance of GSA in the global optimization and solution accuracy. Finally, several simulation results are provided for illustrating the effectiveness of the proposed estimator.

Harnessing AI and analytics to enhance cybersecurity and privacy for collective intelligence systems.

Collective intelligence systems like Chat Generative Pre-Trained Transformer (ChatGPT) have emerged. They have brought both promise and peril to cybersecurity and privacy protection. This study introduces novel approaches to harness the power of artificial intelligence (AI) and big data analytics to enhance security and privacy in this new era. Contributions could explore topics such as: leveraging natural language processing (NLP) in ChatGPT-like systems to strengthen information security; evaluating privacy-enhancing technologies to maximize data utility while minimizing personal data exposure; modeling human behavior and agency to build secure and ethical human-centric systems; applying machine learning to detect threats and vulnerabilities in a data-driven manner; using analytics to preserve privacy in large datasets while enabling value creation; crafting AI techniques that operate in a trustworthy and explainable manner. This article advances the state-of-the-art at the intersection of cybersecurity, privacy, human factors, ethics, and cutting-edge AI, providing impactful solutions to emerging challenges. Our research presents a revolutionary approach to malware detection that leverages deep learning (DL) based methodologies to automatically learn features from raw data. Our approach involves constructing a grayscale image from a malware file and extracting features to minimize its size. This process affords us the ability to discern patterns that might remain hidden from other techniques, enabling us to utilize convolutional neural networks (CNNs) to learn from these grayscale images and a stacking ensemble to classify malware. The goal is to model a highly complex nonlinear function with parameters that can be optimized to achieve superior performance. To test our approach, we ran it on over 6,414 malware variants and 2,050 benign files from the MalImg collection, resulting in an impressive 99.86 percent validation accuracy for malware detection. Furthermore, we conducted a classification experiment on 15 malware families and 13 tests with varying parameters to compare our model to other comparable research. Our model outperformed most of the similar research with detection accuracy ranging from 47.07% to 99.81% and a significant increase in detection performance. Our results demonstrate the efficacy of our approach, which unlocks the hidden patterns that underlie complex systems, advancing the frontiers of computational security.

Cellular computation and cognition.

Contemporary neural network models often overlook a central biological fact about neural processing: that single neurons are themselves complex, semi-autonomous computing systems. Both the information processing and information storage abilities of actual biological neurons vastly exceed the simple weighted sum of synaptic inputs computed by the "units" in standard neural network models. Neurons are eukaryotic cells that store information not only in synapses, but also in their dendritic structure and connectivity, as well as genetic "marking" in the epigenome of each individual cell. Each neuron computes a complex nonlinear function of its inputs, roughly equivalent in processing capacity to an entire 1990s-era neural network model. Furthermore, individual cells provide the biological interface between gene expression, ongoing neural processing, and stored long-term memory traces. Neurons in all organisms have these properties, which are thus relevant to all of neuroscience and cognitive biology. Single-cell computation may also play a particular role in explaining some unusual features of human cognition. The recognition of the centrality of cellular computation to "natural computation" in brains, and of the constraints it imposes upon brain evolution, thus has important implications for the evolution of cognition, and how we study it.

Neural Approximation-Based Adaptive Control Using Reinforced Gain for Steering Wheel Torque Tracking of Electric Power Steering System

In practice, steering wheel torque (SWT) control performance for an electric power steering (EPS) system may degrade owing to various reasons, such as model/parameter uncertainties and unpredictable large external disturbances (e.g., reaction force). Furthermore, the desired SWT suddenly varies and its sign also changes because the direction of the steering wheel angle (SWA) often changes according to the driver. In this article, a neural network (NN) approximation-based adaptive nonlinear control (ANC) using reinforced gain (RG) for an EPS system is proposed to resolve the above-mentioned problems using an SWT model. The proposed control method employs an NN approximator, ANC, and RG. The NN approximator is designed to estimate the unknown complex nonlinear function in EPS modeling. The ANC is designed via backstepping to compensate for parametric uncertainties and external disturbances. Further, the RG is designed as a positive nonlinear function of the error to sufficiently suppress the error according to variations in the desired SWT and external disturbance. The RG increases to suppress the error at a rapid transient response, owing to the sudden change in the sign of the desired SWT and the external disturbance. Thus, a high gain (increased RG) is used only when necessary so that an unnecessarily high gain can be avoided.

Multiple Algebraic Function Nonlinear Systems

Abstract In order to obtain a more complex dynamic behaviour for a nonlinear system several methods were used, including higher order design, system parameters variation and connecting analogue and discrete subsystems. The present paper analyses an approach to generate wide bandwidth signals by connecting three elementary algebraic building blocks to obtain a more complex non-linear function to be included in the analogue dynamic system. The resulting non-linear system attractor is presented, its Poincare section is analysed and statistical numerical simulations results are shown to highlight the possible application as noise generator.

Milling tool wear prediction: optimized long short-term memory model based on attention mechanism

To improve the prediction accuracy of milling tool wear, a prediction method based on Attention-LSTM is proposed. In the training phase, first, the data are pre-processed by truncation, downsampling, and the Hampel filtering method, and then features are extracted by the time domain, frequency domain, and time-frequency domain analysis methods. Second, a deep neural network is designed to describe the complex nonlinear function between features and tool wear. Last, aiming at the insufficient prediction accuracy due to the LSTM lacking feature extraction and enhancement, the Attention mechanism is introduced to optimize the model. The results suggest that this prediction method provides an efficient strategy for milling tool wear prediction.

Effectiveness of carbon policies and multi-period delay in payments in a global supply chain under remanufacturing consideration

Carbon emission regulation is one of the most important priorities for the business sector. The rate of carbon generation increases tremendously due to several issues like transportation, repair, or remanufacture. The generation of faulty or imperfect items increases during transportation. Instead of returning them, the retailer can remanufacture/repair those faulty products in their repair shop and sell them to customers as the new product. Carbon dioxide is emitted from this repair shop continuously. In this study, the retailer remanufacturers faulty items under different carbon emission reduction policies and multi-period trade credit financing to control the shortage. Shortages arise due to imperfect products. Three carbon emission reduction policies are reviewed and compared in the context of different credit financing and investments to find the maximum profit. The adaptation of investment is beneficial to reduce carbon emissions and increase the profit of the system. Nine models are built based on long-term credit financing and carbon regulation policies. Analytical approaches are used to find solutions for complex nonlinear profit functions. Numerical illustrations prove that if cycle time is less than the permitted delay period, the system’s profit is 2.58% and 56.76% higher than the other two cases. Limited carbon regulation provides 0.23% and 0.16% better profit than the other two carbon reduction policies. Green investment increases the system profit by 0.063% and reduces emissions by 112 kg per cycle. Adaptation of trade credit policy increases the profit up to 1.55%. Finally, it is noticeable that limited carbon regulation is the best policy when the permissible delay in payment period is greater than the cycle time with remanufacturing under-investment.

Multi-Agent Reinforcement Learning for Traffic Flow Management of Autonomous Vehicles

Intelligent traffic management systems have become one of the main applications of Intelligent Transportation Systems (ITS). There is a growing interest in Reinforcement Learning (RL) based control methods in ITS applications such as autonomous driving and traffic management solutions. Deep learning helps in approximating substantially complex nonlinear functions from complicated data sets and tackling complex control issues. In this paper, we propose an approach based on Multi-Agent Reinforcement Learning (MARL) and smart routing to improve the flow of autonomous vehicles on road networks. We evaluate Multi-Agent Advantage Actor-Critic (MA2C) and Independent Advantage Actor-Critical (IA2C), recently suggested Multi-Agent Reinforcement Learning techniques with smart routing for traffic signal optimization to determine its potential. We investigate the framework offered by non-Markov decision processes, enabling a more in-depth understanding of the algorithms. We conduct a critical analysis to observe the robustness and effectiveness of the method. The method's efficacy and reliability are demonstrated by simulations using SUMO, a software modeling tool for traffic simulations. We used a road network that contains seven intersections. Our findings show that MA2C, when trained on pseudo-random vehicle flows, is a viable methodology that outperforms competing techniques.

Neural networks optimized by genetic algorithms in cosmology

The applications of artificial neural networks in the cosmological field have shone successfully during the past decade, this is due to their great ability of modeling large amounts of datasets and complex nonlinear functions. However, in some cases, their use still remains controversial because their ease of producing inaccurate results when the hyperparameters are not carefully selected. In this paper, to find the optimal combination of hyperparameters to artificial neural networks, we propose to take advantage of the genetic algorithms. As a proof of the concept, we analyze three different cosmological cases to test the performance of the architectures achieved with the genetic algorithms and compare them with the standard process, consisting of a grid with all possible configurations. First, we carry out a model-independent reconstruction of the distance modulus using a type Ia supernovae compilation. Second, the neural networks learn to infer the equation of state for the quintessence model, and finally with the data from a combined redshift catalog the neural networks predict the photometric redshift given six photometric bands (urgizy). We found that the genetic algorithms improve considerably the generation of the neural network architectures, which can ensure more confidence in their physical results because of the better performance in the metrics with respect to the grid method.

Describe, Spot and Explain: Interpretable Representation Learning for Discriminative Visual Reasoning.

Despite the recent success achieved by deep neural networks (DNNs), it remains challenging to disclose/explain the decision-making process from the numerous parameters and complex non-linear functions. To address the problem, explainable AI (XAI) aims to provide explanations corresponding to the learning and prediction processes for deep learning models. In this paper, we propose a novel representation learning framework of Describe, Spot and eXplain (DSX). Based on the architecture of Transformer, our proposed DSX framework is composed of two learning stages, descriptive prototype learning and discriminative prototype discovery. Given an input image, the former stage is designed to derive a set of descriptive representations, while the latter stage further identifies a discriminative subset, offering semantic interpretability for the corresponding classification tasks. While our DSX does not require any ground truth attribute supervision during training, the derived visual representations can be practically associated with physical attributes provided by domain experts. Extensive experiments on fine-grained classification and person re-identification tasks qualitatively and quantitatively verify the use our DSX model for offering semantically practical interpretability with satisfactory recognition performances.

The TSCR method for precision estimation of ill-posed mixed additive and multiplicative random error model

Estimating the precision information of parameter estimation can fully reflect the quality of parameter estimation. In this paper, we first derive the weighted least-square regularization iterative (WLSRI) solution and mean square error (MSE) matrix of the ill-posed mixed additive and multiplicative random error (MAMRE) model. Then, considering that the gradual iterative process of the WLSRI solution will affect the final parameter estimation and precision information and further lead to a complex nonlinear function relationship, the traditional Taylor expansion approximate function method cannot be used to solve. Therefore, this paper introduces the derivative-free third-degree spherical-radial cubature rule (TSCR) method for precision estimation of the ill-posed MAMRE model, which generates a series of samples with the same weight by the fixed sampling strategy and further uses the WLSRI method to calculate. Finally, the experiment research and analysis results illustrate that compared with the existing solutions without considering the ill-posed problem, the WLSRI method is applicable and can obtain reasonable parameter estimation and precision information in solving the ill-posed MAMRE model; while the TSCR method can obtain more accurate parameter estimation and precision information than the WLSRI method, which enriches the theoretical research on the precision estimation problem of ill-posed MAMRE model.

Dynamic economic dispatch using hybrid CSAJAYA algorithm considering ramp rates and diverse wind profiles

• Dynamic economic dispatch performed for 10 and 15 units system. • Ramp rate and valve point effect considered for the study. • Penetration level was compared between three different wind profiles. • Novel hybrid CSAJAYA considered as optimization tool. • Results compared with those available in literature. • Statistical analysis performed for proposed hybrid algorithm. Optimal scheduling of the conventional generating units for two different dynamic test systems are percolated in this paper. This paper compares and contrasts among three types of wind profile formulations, namely linear, quadratic, and cubic, which were used to calculate wind power from hourly wind speed to find the profile with the greatest penetration of wind power. Thereafter, the wind profiles were coupled with the test system to execute dynamic economic dispatch. The optimization tool used for the study was a unique hybrid algorithm modelled by combining the properties of the recently developed crow search algorithm (CSA) and JAYA. Involvement of ramp-rate limits and the valve point effects magnified the practicality of the work thereby assessing the proposed hybrid algorithm in solving complex non-linear functions. Results infer that maximum level of wind penetration was attained by linear wind profile and a fuel cost reduction of 2.92% was realized upon incorporation of the same. Numerical results also claims that proposed hybrid CSAJAYA approach consistently yielded better quality solutions within minimum execution time without being affected by the dimension of the problem, thereby outperforming a long list of algorithms implemented for the study.

Time-Series Deep Learning Models for Reservoir Scheduling Problems Based on LSTM and Wavelet Transformation

In 2022, as a result of the historically exceptional high temperatures that have been observed this summer in several parts of China, particularly in the province of Sichuan, residential demand for energy has increased. Up to 70% of Sichuan’s electricity comes from hydropower, thus creating a sensible and practical reservoir scheduling plan is essential to maximizing reservoir power generating efficiency. However, classical optimization, such as back propagation (BP) neural network, does not take into account the correlation of samples in time while generating reservoir scheduling rules. We proposed a prediction model based on LSTM neural network coupled with wavelet transformation (WT-LSTM) to address the problem. In order to extract the reservoir scheduling rules, this paper first gathers the scheduling operation data from the Xiluodu hydropower station and creates a dataset. Next, it uses the feature of the time-series prediction model with the realization of a complex nonlinear mapping function, time-series learning capability, and high prediction accuracy. The results demonstrate that the time-series deep learning network has high learning capability for reservoir scheduling by comparing evaluation indexes such as root mean square error (RMSE), rank-sum ratio (RSR), and Nash–Sutcliffe efficiency (NSE). The WT-LSTM network model put forward in this research offers better prediction accuracy than conventional recurrent neural networks and serves as a reference base for scheduling decisions by learning previous scheduling data to produce outflow solutions, which has some theoretical and practical benefits.