**1. Introduction**

For some time, the global energy issue has been a major concern, hindering the development of human society [1,2]. The 2019 BP World Energy Statistical Yearbook shows that, in 2018, global primary energy demand increased 2.9% and carbon emissions increased 2.0%. This was the fastest growth year since 2010. In 2019, affected by the new coronavirus epidemic, the growth rate of global primary energy consumption slowed to 1.3% as compared to 2018, but carbon emissions caused by energy consumption increased significantly, by 2.0% [3]. China accounts for more than three-quarters of the net increase in global energy consumption and has become its largest driving force. For the sustainable development of both the economy and society, the Chinese government has put energy conservation and emissions reduction front and center [4]. At the 2021 China Energy Work Conference, there was a call for the strict implementation of a "dual control" system involving both total energy consumption and intensity, with total energy consumption to be limited to within five billion tons of standard coal at an average annual growth rate of less than 3%.

The industrial sector is the largest consumer of energy. According to data from the United Nations Industrial Development Organization, in developing countries and countries with economies in transition, the growth rate of industrial energy use will

**Citation:** Liu, J.; Qian, Y.; Yang, Y.; Yang, Z. Can Artificial Intelligence Improve the Energy Efficiency of Manufacturing Companies? Evidence from China. *Int. J. Environ. Res. Public Health* **2022**, *19*, 2091. https://doi.org/10.3390/ ijerph19042091

Academic Editor: Nir Krakauer

Received: 12 December 2021 Accepted: 11 February 2022 Published: 13 February 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

be 1.8–3.1% per year, with 50% of energy to be supplied to industrial systems. At the same time, the contradiction between economic development and limited energy supply has become increasingly prominent. Therefore, how to manage the energy demand of Chinese manufacturing enterprises and improve their energy efficiency is very important for achieving regional and global reductions in greenhouse gas emissions and reducing corporate energy intensity [5]. Studies to date have pointed out that technology can indeed improve energy efficiency and reduce energy consumption [6,7], but current industrial energy efficiency is far below the best technically feasible levels.

With the rise of a new global scientific and technological revolution, AI has developed rapidly around the world and has now become an important developmental trend in global manufacturing. The use of industrial robots is an important manifestation of the application of AI in the manufacturing sector [8,9]. From the perspective of the application of industrial robots in China (see Figure 1), although the number of industrial robots put into use is increasing year by year, its growth rate is far lower than that for the number of industrial robots purchased in China. That is, companies have purchased artificial intelligence (AI) equipment, but the proportion of production applications is not high. There is a practical problem: the operation of artificial intelligence requires a lot of energy. After it is put into production, can artificial intelligence improve manufacturing enterprises' energy efficiency? Manufacturing enterprises may have more stringent technical conditions for using AI and may face higher investment costs, resulting in the number of purchases of AI being greater than the number of applications for their use. Will this lead to a waste of resources for the company, thereby inhibiting energy efficiency? The questions to be studied in this paper are as follows:


**Figure 1.** Installation and use of industrial robots in China.

To answer the above questions, it is necessary to conduct empirical research on the basis of looking at related theories and combining them with real world data in China.

Most of the existing research on artificial intelligence and industrial robots has focused on the labor market, economic growth, carbon emissions, etc. [10–12]. However, empirical research between AI and energy efficiency is relatively rare. We aim to supplement this research herein. The purpose of this paper is to analyze the impact mechanism of artificial intelligence on the energy efficiency of manufacturing enterprises and to answer the question of how artificial intelligence affects energy efficiency. This paper uses the data of manufacturing enterprises to construct a DEA-Malmquist model and multiple fixed effect model for empirical testing. We further analyze the role of firm age and firm performance in moderating the impact of AI on energy efficiency. This paper provides micro-evidence for the impact of artificial intelligence on energy efficiency, and expands the research on artificial intelligence and enterprise energy efficiency.

The structure of this article is as follows: This first part reviews the related research on the factors affecting energy efficiency. Through the method of literature review, it analyzes how artificial intelligence has an impact on energy efficiency and proposes research hypotheses. The second part is the literature review and research hypothesis. The third part is model design, and mainly deals with model design, variable selection and descriptive statistics of data. The main goal is to build an econometric model that empirically tests the impact of AI on energy efficiency. A data envelopment model is established to measure the energy efficiency of manufacturing enterprises. Finally, select the relevant control variables and describe the data. The fourth part is the empirical test. This part mainly analyzes the empirical results and uses the instrumental variable method (IV) to alleviate the endogeneity problem of the model and to discuss the heterogeneity of enterprises. The fifth part is discussion. This part will use the literature comparison method to compare the conclusions of this paper with the published literature, answer the research questions of this paper and summarize its contributions. The last part is the conclusion and policy recommendations.

#### **2. Literature Review and Research Hypothesis**

With the increasingly prominent energy problem, discovering the factors that affect energy efficiency has become the focus of scholars' research, including urbanization level [13], energy cost [14], environmental regulation [15,16], and resource endowment [17], etc.

Relevant research on the impact of technological progress on corporate energy efficiency can be divided into two categories. One view is that technological advancement can improve corporate energy efficiency and thereby reduce energy consumption [18–20] Popp [21] used patent data to estimate the impact of technological progress on energy consumption. The results of the study proved that technological advancement can save enterprises more energy in the long run. Welsch and Ochsen [22] demonstrated, through an empirical study in the Federal Republic of Germany, that technological progress can improve corporate energy efficiency, while factor substitution and biased technological progress are important factors for fluctuations in energy intensity. Technological progress can effectively narrow the energy efficiency gap between European companies. In the future, the EU should support European manufacturing companies in introducing and using both sustainable processes and product innovation to narrow the energy efficiency gap [23]. Wang and Wang [24], using the number of patents granted to measure technological innovation, found that technological innovation in China has significantly improved urban energy efficiency. Sun et al. [25], based on their research on 24 innovative countries, found that there is a significant positive relationship between technological innovation and energy efficiency.

Other scholars believe that innovations in artificial intelligence, information and communication technology have led to a decrease in the unit cost of energy. This will stimulate enterprises to expand production, bring on a "rebound effect" to energy consumption and thus lead to a more complex kind of energy efficiency for manufacturing enterprises [26,27]. Currently, academia has widely accepted the existence of the rebound effect [28–30], although it is still controversial as to whether the rebound effect will completely offset increases in energy efficiency brought about by technological progress (that is, the rebound effect is greater than 100%). A study by Jin [31], based on the electricity consumption data of 3500 households in South Korea, gave empirical calculations showing that the energy rebound effect was about 30%. Vélez-Henao et al. [32] believed that every 1% drop in energy prices in Colombia would increase the rebound effect by 38.56%. Adha et al. [33] found that the short-term and long-term rebound effects in Indonesia were 87.2% and −45.5%, respectively, indicating that technological improvements can improve energy efficiency in the long-term. Therefore, technological progress has become an important factor affecting energy efficiency.

Artificial intelligence is considered to be a general technology that can support other innovations [34]. While enterprises use AI to achieve technological progress, the use of AI further promotes new technological innovations, thus forming a new virtuous circle [35]. Therefore, considering that related research on the impact of AI on corporate energy efficiency is still relatively rare, we look first at research showing how technological progress impacts corporate energy efficiency. This will significantly affect energy efficiency [34].

First, artificial intelligence can accelerate knowledge spillover and creation and promote technological progress of enterprises in energy saving and cleaner production, and thereby improve energy efficiency [25]. The stronger the ability of an enterprise to learn and absorb, the higher its ability to innovate [36]. Through deep learning and computer vision technology, artificial intelligence can screen out a large amount of effective information and create new knowledge and new computing solutions more efficiently than ever before, thereby accelerating the process of knowledge reorganization [37]. The acceleration of knowledge reorganization can promote the re-creation of knowledge and information [38]. At the same time, artificial intelligence breaks the boundaries of knowledge dissemination within and between enterprises and can accelerate knowledge spillover and information sharing, thereby promoting technological innovation [39]. With the improvement in the level of artificial intelligence, this information contribution ability has been further strengthened. The learning and absorptive capacity of the employees of the enterprise is also continuously improved, thereby promoting the absorption and creation of knowledge within the enterprise [34]. In turn, this promotes technological innovation of enterprises, results in more optimized equipment and energy use decisions, and improves energy efficiency [35].

Second, artificial intelligence promotes technological progress and improves energy efficiency by increasing investment in R&D and talent. The development of artificial intelligence will bring more intelligent devices such as industrial robots, thereby producing a labor substitution effect [40]. The shortage of high-skilled labor caused by this complementary substitution of labor will further force manufacturing enterprises to increase investment in talents and R&D [41]. Talent and R&D investment can further promote technological progress [42]. At the same time, with the increasing global trend of using industrial robots, companies are actively improving their production processes and manufacturing skills. Among them, the use of artificial intelligence technology to improve product processes has become an important way to gain competitive advantage [43]. In turn, through technological progress, the production process is optimized, thereby improving energy efficiency [44]. In the waste management sector, the application of neural networks and machine learning can predict the amount of waste generated, promote waste reuse and improve energy efficiency [45].

In addition, artificial intelligence can improve energy efficiency by increasing technological efficiency. That is, AI can also shorten the gap between businesses and optimal energy efficiency by improving technological efficiency. On the one hand, artificial intelligence can improve production efficiency. Manufacturing companies can use industrial robots to replace low-skilled production workers [46]. Using intelligent technology for production can effectively improve product quality and reduce energy consumption caused

by repeated production due to substandard products [47]. With the help of artificial intelligence technology, such as machine learning, deep learning, etc., enterprises can complete the design, production and sales of products faster [48,49]. On the other hand, artificial intelligence can improve the efficiency of resource allocation. Enterprises use advanced intelligent equipment to make equipment self-perceive, self-analyze, and self-decide. This results in real-time feedback and optimization of production information, reduces equipment response time, reduces energy waste and significantly improves resource allocation efficiency and energy efficiency [50,51].

Artificial intelligence has been widely used in various sectors to improve energy efficiency [52,53]. For example, in the construction sector, the combination of artificial intelligence and big data can improve the energy efficiency of buildings and the comfort of houses [54]. In the energy supply sector, the application of smart meter data can help to accurately predict the consumption of electricity and natural gas so as to better plan and operate the energy supply system [55,56]. Huang and Koroteev [45] believe that AI technologies such as neural networks and machine learning are more successful in energy and waste management, which can then be used to improve the efficiency of electricity, heat and gas in the future. Chen et al. [57] believe that artificial intelligence can optimize equipment scheduling and operation, and their proposed AIEM model can effectively improve energy efficiency and promote the use of renewable energy.

Based on the above analysis, we propose the following hypothesis:

**Hypothesis 1:** *AI can improve the energy efficiency of manufacturing companies.*

**Hypothesis 2:** *AI can improve corporate energy efficiency by promoting technological progress and technical efficiency.*
