arXiv subjects

Mingzhi Zhang

Publications and source records attributed to Mingzhi Zhang.

11 recordsLinked to original sources

arXiv2026

GLM-4.5V and GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning

We present GLM-4.1V-Thinking, GLM-4.5V, and GLM-4.6V, a family of vision-language models (VLMs) designed to advance general-purpose multimodal understanding and reasoning. In this report, we share our key findings in the development of the reasoning-centric training framework. We first develop a capable vision foundation model with significant potential through large-scale pre-training, which arguably sets the upper bound for the final performance. We then propose Reinforcement Learning with Curriculum Sampling (RLCS) to unlock the full potential of the model, leading to comprehensive capability enhancement across a diverse range of tasks, including STEM problem solving, video understanding, content recognition, coding, grounding, GUI-based agents, and long document interpretation. In a comprehensive evaluation across 42 public benchmarks, GLM-4.5V achieves state-of-the-art performance on nearly all tasks among open-source models of similar size, and demonstrates competitive or even superior results compared to closed-source models such as Gemini-2.5-Flash on challenging tasks including Coding and GUI Agents. Meanwhile, the smaller GLM-4.1V-9B-Thinking remains highly competitive-achieving superior results to the much larger Qwen2.5-VL-72B on 29 benchmarks. We open-source both GLM-4.1V-9B-Thinking and GLM-4.5V. We further introduce the GLM-4.6V series, open-source multimodal models with native tool use and a 128K context window. A brief overview is available at https://z.ai/blog/glm-4.6v. Code, models and more information are released at https://github.com/zai-org/GLM-V.

cs.CV↗

arXiv2025

Exploring the Use of Autonomous Unmanned Vehicles for Supporting Power Grid Operations

This paper explores the use of autonomous unmanned vehicles to support power grid operations. With built-in batteries and the capability to carry additional battery energy storage, the rising number of autonomous vehicles can represent a substantial amount of capacity that is currently underutilized in the power grid. Unlike traditional electric vehicles that require drivers, the operations of autonomous vehicles can be performed without human intervention. To guide idle vehicles to autonomously support power grids, we propose a tractable optimization-based method to effectively integrate these "mobile batteries" into grid operations. During real-time operations, the vehicles are strategically routed to target locations to maintain power balance and reduce operating costs. Numerical studies have confirmed both the validity and the scalability of the proposed algorithm to efficiently integrate autonomous vehicles into routine power system operations.

eess.SY↗

arXiv2023

Uncertainty-inspired Open Set Learning for Retinal Anomaly Identification

Failure to recognize samples from the classes unseen during training is a major limitation of artificial intelligence in the real-world implementation for recognition and classification of retinal anomalies. We established an uncertainty-inspired open-set (UIOS) model, which was trained with fundus images of 9 retinal conditions. Besides assessing the probability of each category, UIOS also calculated an uncertainty score to express its confidence. Our UIOS model with thresholding strategy achieved an F1 score of 99.55%, 97.01% and 91.91% for the internal testing set, external target categories (TC)-JSIEC dataset and TC-unseen testing set, respectively, compared to the F1 score of 92.20%, 80.69% and 64.74% by the standard AI model. Furthermore, UIOS correctly predicted high uncertainty scores, which would prompt the need for a manual check in the datasets of non-target categories retinal diseases, low-quality fundus images, and non-fundus images. UIOS provides a robust method for real-world screening of retinal anomalies.

cs.LG↗

arXiv2022

Grid Impact Analysis and Mitigation of En-Route Charging Stations for Heavy-Duty Electric Vehicles

This paper presents a consolidated grid impact analysis design and corresponding mitigation strategies for heavy-duty electric vehicle (EV) charging stations. The charging load of heavy-duty charging station can reach several megawatts, which could induce adverse impacts on the distribution grid if not effectively mitigated. To analyze the impacts and provide corresponding solutions, we select four representative distribution systems - including both single-feeder cases and a multi-feeder case - and design thorough test metrics for the impact analysis. The charging load profiles used in the analysis are derived from realistic conventional heavy-duty vehicle travel data. Based on the analysis results, charging stations are placed at three different representative locations in each distribution system: best, good, and worst locations. Mitigation strategies using a combination of smart charger functionality, on-site photovoltaic (PV) generation, and on-site energy storage (ES) are proposed and tested. A sizing method is also proposed to find the optimal PV-ES-charger capacity that minimizes the capital cost.

eess.SY↗

arXiv2021

Reinforcement Learning for Volt-Var Control: A Novel Two-stage Progressive Training Strategy

This paper develops a reinforcement learning (RL)approach to solve a cooperative, multi-agent Volt-Var Control (VVC) problem for high solar penetration distribution systems. The ingenuity of our RL method lies in a novel two-stage progressive training strategy that can effectively improve training speed and convergence of the machine learning algorithm. In Stage 1(individual training), while holding all the other agents inactive, we separately train each agent to obtain its own optimal VVC actions in the action space: {consume, generate, do-nothing}. In Stage 2 (cooperative training), all agents are trained again coordinatively to share VVC responsibility. Rewards and costs in our RL scheme include (i) a system-level reward (for taking an action), (ii) an agent-level reward (for doing-nothing), and(iii) an agent-level action cost function. This new framework allows rewards to be dynamically allocated to each agent based on their contribution while accounting for the trade-off between control effectiveness and action cost. The proposed methodology is tested and validated in a modified IEEE 123-bus system using realistic PV and load profiles. Simulation results confirm that the proposed approach is robust and computationally efficient; and it achieves desirable volt-var control performance under a wide range of operation conditions.

eess.SY↗

arXiv2021

A Novel Data Segmentation Method for Data-driven Phase Identification

This paper presents a smart meter phase identification algorithm for two cases: meter-phase-label-known and meter-phase-label-unknown. To improve the identification accuracy, a data segmentation method is proposed to exclude data segments that are collected when the voltage correlation between smart meters on the same phase are weakened. Then, using the selected data segments, a hierarchical clustering method is used to calculate the correlation distances and cluster the smart meters. If the phase labels are unknown, a Connected-Triple-based Similarity (CTS) method is adapted to further improve the phase identification accuracy of the ensemble clustering method. The methods are developed and tested on both synthetic and real feeder data sets. Simulation results show that the proposed phase identification algorithm outperforms the state-of-the-art methods in both accuracy and robustness.

eess.SY↗

arXiv2021

A Data-driven Probabilistic-based Flexibility Region Estimation Method for Aggregated Distributed Energy Resources

This paper presents a data-driven, distributionally robust chance-constrained optimization method for estimating the real and reactive power controllability of aggregated distributed energy resources (DER). At the DER-level, a two-dimensional flexibility region can be formed based on the real and reactive power regulating limits of each DER considering forecast uncertainty. At the feeder-level, an aggregated flexibility region is computed via a multi-directional search method. In each search direction, extend the real and reactive power of each controllable DER towards its operational limits until: i) all DERs' maximum operational limits are reached or, ii) one or more of the distribution network operational limits are violated. The method enables three key features for operating aggregated DER resources: controllability estimation, visualization, and risk quantification. Simulation results demonstrated the effectiveness of the algorithm and quantified the impact of different parameter settings on the flexibility region boundary changes. The proposed algorithm is robust and computationally efficient and can meet real-time computing needs.

eess.SY↗

arXiv2021

A Novel Interactive Two-stage Joint Retail Electricity Market for Multiple Microgrids

To accommodate the advent of microgrids (MG) managing distributed energy resources (DER) in distribution systems, an interactive two-stage joint retail electricity market mechanism is proposed to provide an effective platform for these prosumers to proactively join in retail transactions. Day-ahead stochastic energy trading between the distribution system operator (DSO) and MGs is conducted in the first stage of a centralized retail market, where a chance-constrained uncertainty distribution locational marginal price (CC-UDLMP) containing the cost of uncertainty precautions is used to settle transactions. In the second stage, a novel intra-day peer-to-peer-based (P2P) flexibility transaction pattern is implemented between MGs in local flexibility markets under the regulation of DSO to eliminate power imbalances caused by rolling-based estimates whilst considering systematic operations. A fully distributed iterative algorithm is presented to find the equilibrium solution of this two-stage sequential game framework. Moreover, in order to enhance the versatility of this algorithm, an improved Lp-box alternating direction methods of multipliers (ADMM) algorithm is used to efficiently resolve the first-stage stochastic economic dispatch problem with a mixed-integer second-order cone structure. It is verified that the proposed market mechanism can effectively improve the overall market efficiency under uncertainties.

eess.SY↗

arXiv2020

A Transactive Retail Market Mechanism for Active Distribution Network Integrated with Large-scale Distributed Energy Resources

The burgeoning integration of distributed energy resources (DER) poses new challenges for the economic and safe operation of the electricity system. The current distribution-side policy is largely based on mandatory regulations and incentives, rather than the design of a competitive market mechanism to arouse DERs to freely compete in the retail market. To address this issue, we proposed a transactive retail market mechanism to attract the active participation of profit-driven DER retailers in a deregulated way. Considering the limited competitive property of the retail market, a bi-level DSO-dominated framework is constructed to simulate the virtual game between distribution system operator (DSO) and DER retailers. Specifically, the flexible interval pricing of retailers is modeled as a series of binary revenue constraints to influence the DSO's centralized economic dispatch decisions, where the time-varying distribution locational marginal price is adopted to settle energy activities at different locations over the time horizon. Due to the uncertain power flow path during the coordinated decision-making process, we proposed a general undirected second-order cone-based AC radial power flow model and provided sufficient conditions to ensure its exactness. Also, we applied a series of approximation and relaxation techniques to transform the bi-level mixed-integer quadratic framework with highly discrete induced domains into a solvable mixed-integer semidefinite programming problem. It is demonstrated in the case study that the proposed mechanism can not only improve market efficiency but also eliminate market power and the resulting market failures.

eess.SY↗

arXiv2020

An Networked HIL Simulation System for Modeling Large-scale Power Systems

This paper presents a network hardware-in-the-loop (HIL) simulation system for modeling large-scale power systems. Researchers have developed many HIL test systems for power systems in recent years. Those test systems can model both microsecond-level dynamic responses of power electronic systems and millisecond-level transients of transmission and distribution grids. By integrating individual HIL test systems into a network of HIL test systems, we can create large-scale power grid digital twins with flexible structures at required modeling resolution that fits for a wide range of system operating conditions. This will not only significantly reduce the need for field tests when developing new technologies but also greatly shorten the model development cycle. In this paper, we present a networked OPAL-RT based HIL test system for developing transmission-distribution coordinative Volt-VAR regulation technologies as an example to illustrate system setups, communication requirements among different HIL simulation systems, and system connection mechanisms. Impacts of communication delays, information exchange cycles, and computing delays are illustrated. Simulation results show that the performance of a networked HIL test system is satisfactory.

eess.SY↗

arXiv2017

On the Enumeration and Congruences for m-ary Partitions

Let $m\ge 2$ be a fixed positive integer. Suppose that $m^j \leq n< m^{j+1}$ is a positive integer for some $j\ge 0$. Denote $b_{m}(n)$ the number of $m$-ary partitions of $n$, where each part of the partition is a power of $m$. In this paper, we show that $b_m(n)$ can be represented as a $j$-fold summation by constructing a one-to-one correspondence between the $m$-ary partitions and a special class of integer sequences rely only on the base $m$ representation of $n$. It directly reduces to Andrews, Fraenkel and Sellers' characterization of the values $b_{m}(mn)$ modulo $m$. Moreover, denote $c_{m}(n)$ the number of $m$-ary partitions of $n$ without gaps, wherein if $m^i$ is the largest part, then $m^k$ for each $0\leq k<i$ also appears as a part. We also obtain an enumeration formula for $c_m(n)$ which leads to an alternative representation for the congruences of $c_m(mn)$ due to Andrews, Fraenkel, and Sellers.

math.CO↗