Study on Prediction of Energy Conservation and Carbon Reduction in Universities Based on Exponential Smoothing

Abstract

With the continuous development of China’s economy, the phenomenon of energy scarcity has become more and more prominent, for which China has put forward the strategic goal of carbon peak and carbon neutrality (double carbon target). As densely populated areas, the demand for energy is especially tight in universities. In combination with the work of “conservation-oriented colleges” carried out by the Ministry of Education, the accurate monthly electrical and water energy consumption of Kunming University of Science and Technology from 2018–2021 was counted, and the data were plotted into an energy consumption analysis chart to determine its compliance with the prediction range of the smoothing index prediction model. The corresponding smoothing indices were calculated by writing smoothing formulas through Excel, and, finally, the overall energy consumption indexes for 2022 and 2023 were successfully predicted using the exponential smoothing method. The errors between the real and forecasted values of electricity and water consumption in 2021 are reduced to 2.61% and 2.44%. The smoothing index predicts that the baseline discounted electricity energy consumption in 2022 is 5,423,658.235 kgce and in 2023 is 5,758,865.224 kgce; on the other hand, the baseline discounted water energy consumption in 2022 is predicted to be 632,654.321 kgce, while in 2023 it is predicted to be 652,321.238 kgce. The projected values can be used as an early warning line for the energy consumption index, and long-term management approaches and data support for energy conservation and carbon emission reduction can be effectively provided. The mentioned research provides an important reference for the proposal and implementation of efficient management measures, and provides strong theoretical technical support for the implementation of the carbon peak and neutrality in universities.

1. Introduction

Carbon emission reduction is an ongoing global concern and a social challenge that needs to be addressed urgently. Emissions of industrial pollutants, such as carbon dioxide, led to a dramatic increase in the total amount of greenhouse gases, threatening the ecological balance of the earth and the living environment of human beings [1,2]. Against this grim backdrop, most countries within the world have demonstrated plans to cut back greenhouse emissions. Our country has put forward the “Double carbon” strategic goal at this essential moment [3]; that is, to reach the carbon peak by 2030 and achieve carbon neutrality by 2060, which is closely related to the development of the property of the Chinese nation and to the building of a society with a common future for mankind [4,5,6]. According to the national regulations, the carbon peak can be a commitment to prevent the rise in greenhouse gas emissions by 2030, and gradually reduce them after reaching the peak [7,8]; by 2060, in order to reduce greenhouse gas emissions, we have to take a variety of measures [9,10], such as tree planting, energy saving, and emission reduction, to neutralize all pollution hazards [11]. It is vital to promote energy and resource conservation [12], implement a comprehensive conservation and consumption reduction strategy [13], shrink energy consumption and carbon emissions per unit of output continuously [14], and promote a simple, green, and low-carbon healthy lifestyle [15,16].

In the background of comprehensive energy saving and emission reduction, the problem of universities’ high energy consumption becomes very prominent. As indicated by relevant reports, the total annual energy consumption of state agencies, university office buildings, and public buildings accounts for 20% of the overall energy consumption in Chinese urban areas [17,18], and is 10–20 times higher than the ordinary household [19]. As the United States Government Accounting Office and some students noted in their analysis, because of the continual growth of the university scale, the university has, step-by-step, matured into a significant power shopper within the past few decades [20,21,22]. Abdelalim et al. [23] proposed several methods to analyze and visualize building-level water, natural gas, and electricity consumption and the upstream environmental impacts: Sankey diagrams and bar charts that normalize metered values by floor area and occupancy. The methods are then applied to a 45-building Canadian university campus and an array of graphical representations of the data is provided. The resulting analysis and visualization reveal significant variation in consumption between buildings, regardless of building vintage and function. In the research of Zhou et al. [24], a detailed investigation in the form of questionnaire was carried out for the energy consumption of colleges and universities in Guangdong Province, including electricity, water, gas, and cooling energy consumption from 2006 to 2010. The energy meter reading modes, energy conversation investment amount, and energy-efficiency retrofit work were also reported and analyzed. Deng et al. [25] collected the monthly electricity data of 620 rooms within two large dormitory buildings in a university campus located in the hot summer–cold winter climate region for two consecutive years (2018–2019), and administered a questionnaire survey concerning the usage of three common domestic appliances: computer, electric water heating, and air conditioning. Based on a hierarchical multiple regression model, these behaviors accounted for 74.8% and 73.1% of electricity consumption in the years of 2018 and 2019, respectively. Wang et al. [26] investigated the energy consumption at 30 national universities, 9 national universities of science and technology, 17 private universities, and 16 private universities of science and technology in Taiwan from 2015 to 2017. The study could guide the public sector in formulating energy conservation, carbon reduction strategies, and benchmarks for universities.

With the goal of building a “conservation university” and promoting the continuous optimization of the energy cycle of universities, it is necessary to refine the main energy utilization of colleges. The main innovation of this paper is the discussion of the management of and solutions for the university’s recent electricity and water consumption, and the use of exponential smoothing statistical methods at the decision-making level to provide scientific statistics and forecasts of future energy consumption, from which effective energy saving targets can be developed. The exponential smoothing method is used to forecast energy losses and implement measures to control them strictly after clarifying the target values, so as to further improve energy utilization efficiency and achieve the goal of rebalancing energy consumption.

2. Analysis of Energy Consumption

Collectively, the energy use of colleges and universities is mainly concentrated in two aspects: electricity and water consumption. Electricity is mainly composed of teaching electricity, living electricity, scientific research, and infrastructure power consumption. Among them, scientific research and domestic electricity consumption occupy a large proportion [10]. In terms of water consumption, the situation of universities is basically the same as that of electricity consumption, which is integrated by water for research process and domestic use and composition. In general, the energy consumption of universities has the following four distinct characteristics:

(1) Energy use for teaching and living can be primarily affected by holidays and seasonal factors. The holiday factors have great volatility. Scientific research water is also affected by the cold and summer vacation at the same time, and increased additional living water in summer, meaning the total amount of water from universities can change greatly [27,28]. Unlike water energy consumption, the volatility of electric energy consumption is mainly manifested in winter and summer. Due to the needs of refrigeration and heating, the use of electricity is much larger than the usual normal teaching period. During the summer vacation, the demand for electricity is reduced, thus, lowering the consumption of electricity.

(2) Low electricity load rate and high unused consumption rate. Since most universities have a large surplus in the planning of electricity consumption at the early stage of campus construction, the designed electricity load is higher than the actual electricity load, resulting in low utilization rate of equipment in actual use, and the wasted electricity load of much equipment is more than 30% [29]. At the same time, the long cycle of construction of teaching facilities in some colleges and universities caused the phenomenon of high transformer idle rate, especially during winter and summer vacations, when the utilization rate of transformers is only one tenth of the full load.

(3) Poor energy use conditions and inadequate management systems. It is difficult for the relevant school departments to consider the actual situation of students and teachers comprehensively when formulating some management systems for teacher and student groups. Moreover, the school’s logistics managers and relevant executives may make inaccurate judgments about the use and maintenance problems of electrical equipment without a good understanding of the actual situation [30]. When there are problems found with the relevant electrical appliances, the complexity of the application maintenance prevents the relevant maintenance personnel from repairing them in a timely manner, further accelerating the energy loss situation.

(4) Variations in the mainstream orientation of scientific research result in large differences in the total amount of energy used by universities. Therefore, similar to Kunming University of Science and Technology, which is an engineering-oriented institution, the daily research activities such as experiments and tests require a lot of high-powered equipment and instruments, thus, the energy consumption of water and electricity must be greater than that of other comprehensive universities.

3. The Main Reason of Energy Consumption

Energy consumption in colleges and universities has been on the rise in recent years, which is inevitably influenced by the objective factors of market economic development, namely, the increase in electricity and water costs as a result of the increase in market prices of electricity and consumables [31]. The main reasons are discussed in detail as the following specific aspects:

(1) Electricity use management function is not prominent. In the use process, the focus of energy use is placed on ensuring normal teaching, research and living electricity, and water use in the university, resulting in insufficient energy conservation efforts. The serious wastage of water and electricity in the process of living, teaching, and research, such as classrooms with no one turning off the lights, dormitory buildings with bright lights, experimental equipment staying on for a long time, or people leaving the office without turning off the computer, are very common [15]. Some electric facilities are old and consume a lot of electricity, however, many universities are unable to take measures such as timely renewal and maintenance of aging equipment, due to insufficient funds or other reasons, which leads to an overload operation of equipment, and greatly increases safety hazards.

(2) There is a large talent vacancy for specialists responsible for energy management. Universities are a base for talent training, but the technical professionals in universities, especially logistics practitioners, generally do not meet the demand for technical quality and have an insufficient grasp of the details [32]. As a result, the problem of electrical energy waste is not solved scientifically and effectively, and the existing situation cannot be analyzed and improved in time, which is not conducive to the comprehensive promotion of energy conservation.

(3) Students’ ideas of energy saving and emission reduction need to be improved. Most students do not have enough awareness of saving electricity and water, and believe that energy saving is not an essential part of their research and study life. In fact, strengthening energy conservation management and education for students are important goals of management and education, as well as a concrete manifestation of social responsibility [33]. Energy conservation should be fully supported by the majority of students and faculty members, and, at the same time, making energy reduction and conservation an important foundation in the daily study and research process is essential.

As the most important communication and dissemination channel of knowledge and civilization, higher education institutions must play an important role in building a conservation-oriented society. We must mobilize teachers and students to start from their surroundings and, from the smallest steps, increase the efforts of energy saving, carbon emission reduction, and carbon peak and neutrality publicity and education; improve the awareness of energy saving and emission reduction; reach a consensus of energy saving and consumption reduction; and work together to build an energy-saving campus.

4. Exponential Smoothing Methods

More than 60 years have passed since the introduction of exponential smoothing, which is widely used in numerical simulations and experimental statistics [34]. The purpose of the exponential smoothing method is to assume that the trend of past developments between variables will continue into future changes, so that weights with smaller values are assigned to observations of future variables, while larger weights are assigned to observations of present variables. By assigning weights in this way, the desired forecasting results can be obtained, and the results will reflect more up-to-date information and will, likewise, be more consistent with reality, and can be applied to the forecasting of short- and medium-term values [18]. The core computational process can be expressed as responding to the original variables by smoothing them in order to obtain “smoothed values”, and then using the smoothed values to construct a forecasting model for the original variables and, finally, to predict the future values. The main features of the exponential smoothing method are relatively small computational effort, simple operation, strong variables with different time series, and relatively stable prediction results.

Generally speaking, the exponential smoothing model is built once first, and the time series of the forecast object can be set as x (1), x (2), … x (t). S refers to the smooth detailed value of the index; the detailed formula is given in the following equation:

$$S_t^1=\alpha {\chi }_t+\left(1-\alpha \right)S_{t-1}^1, 0<\alpha <1,t=1, 2, 3\dots n$$

(1)

In the above formula, the $\alpha$ refers to the smooth factor, and the smoothness of the outlook data is positively related to the $\alpha$ value. The outlook index of phase t, also known as the smooth value, included in the prediction index is replaced by $S_t^1$. The first term ${\chi }_i$ can be selected as the average of the historical data of the first few periods and set as the initial value of the operation. The recursive formula is expanded:

$$\begin{gathered} S_t^1=\alpha {\chi }_t+\left(1-\alpha \right)S_{t-1}^1 \\ =\alpha {\chi }_t+\left(1-\alpha \right)\left[\alpha {\chi }_{t-1}+\left(1-\alpha \right)S_{t-2}^1\right] \\ =\alpha {\chi }_t+\alpha \left(1-\alpha \right){\chi }_{t-1}+\alpha {\left(1-\alpha \right)}^2{\chi }_{t-2}+\cdots +\alpha {\left(1-\alpha \right)}^{t-1}{\chi }_1+{\left(1-\alpha \right)}^t{S}_0^1 \end{gathered}$$

(2)

From the upper formula, 0 <$\alpha$< 1, so the coefficient $\alpha {\left(1-\alpha \right)}^i$ and i value of ${\chi }_i$ show negative phenomena. It is further derived that the nearest t-period ${\chi }_t$ to period t + 1 has the largest value of the weight $\alpha$, followed by the weight $\alpha \left(1-\alpha \right)$ of ${\chi }_{t-1}$, while the weight of ${\chi }_1$ is the smallest of them. What can be seen is that the recent data have a greater influence on the predicted values of the final data than the distant data, which means that the distant data account for a smaller component of the influence in the overall operation and calculation process. In summary, when the time series values of the target model fluctuate up and down around a constant value, the primary exponential smoothing method can be used for effective forecasting. The exponential smoothing method eliminates the chance variation of variables in the time series, without affecting the prediction effect due to the chance variation of variables, and provides a good protection for the existing recent variable data in the prediction process [34,35].

5. Simulation and Discussion

5.1. Simulation Background

Electric energy and hydraulic energy, as the basic energy sources for daily teaching, scientific research, and daily life in universities, are the basic guarantees for the normal operation of higher education institutions. Nowadays, the lack of standardization of electricity and water use by many university students has led to automatic tripping of electrical energy systems, fires, or leaking pumps [36]. The occurrence of these problems is directly related to the management of energy use in universities [37]. More importantly, universities need to set up a stepped electricity and water use plan and increase the price of utilities beyond the basic index, which can be a good way to save energy. This model can also be used to develop future quantitative management methods for energy-saving indicators, improve the theoretical model of energy saving in universities, and provide long-term management methods and data support for energy saving and emission reduction goals.

In summary, we propose to use the primary exponential smoothing method to simulate and analyze the energy consumption data, hoping to assign larger weights to the data closer to the prediction period without ignoring the background of historical data, and to make the weights form a decreasing relationship from near-to-far according to the exponential law. At the same time, the primary exponential smoothing method is used to count the predicted and actual values according to the current period, and the smoothing coefficient $\alpha$ is substituted into the formula for the calculation of the weighted average, followed by the prediction of the next period’s values. Through the complete prediction and simulation process, we can establish a model based on the existing data of electrical and water energy consumption of Kunming University of Science and Technology, and make a good planning and implementation plan, and then implement effective energy saving measures for the university, which is the most urgent task for the development of university power management.

5.2. Simulation Analysis

First of all, the accurate monthly consumption of electric energy and water energy of Kunming University of Science and Technology from 2018 to 2021 was counted, and the total energy consumption of the two campuses was directly derived from the latest educational administration information management system. The most accurate data are derived from the university data system. Each dormitory, classroom, or lab has an electric and water meter, and the generated figures including maintenance, power outage, and other events are connected to the network to meet the monthly synchronization needs. In order to facilitate the subsequent development of carbon consumption indicators, the two were converted into standard coal consumption. Details are shown in

Table 1. Kunming University of Science and Technology 2018–2021 monthly electric energy and water energy consumption details table.

Year	Month	Power Consumption/kW·h	Water Consumption/t	Electricity Discount Standard Coal Quantity/kgce	Water Discount Standard Coal Quantity/kgce
2018	1	3,001,634	155,592	368,901	40,003
	2	2,537,168	137,649	311,818	35,390
	3	4,019,799	216,354	494,033	55,625
	4	3,992,845	207,886	490,721	53,447
	5	1,447,222	195,529	177,864	50,271
	6	2,096,400	181,992	257,648	46,790
	7	1,375,116	136,440	169,002	35,079
	8	1,683,829	161,850	206,943	41,612
	9	2,632,827	197,090	323,574	50,672
	10	3,792,806	201,190	466,136	51,726
	11	4,100,607	213,536	503,965	54,900
	12	4,493,019	212,707	552,192	54,687
2019	1	3,601,684	159,092	442,647	40,903
	2	2,437,157	134,649	299,527	34,618
	3	4,219,799	226,358	518,613	58,197
	4	3,982,835	227,885	489,490	58,589
	5	3,975,014	242,811	488,529	62,427
	6	3,667,961	209,339	450,792	53,821
	7	2,920,883	156,922	358,977	40,345
	8	3,303,619	186,018	406,015	47,825
	9	3,983,501	219,551	489,572	56,447
	10	4,101,037	202,922	504,017	52,171
	11	4,333,223	216,322	532,553	55,616
	12	5,274,376	222,058	648,221	57,091
2020	1	3,345,474	150,132	411,159	38,599
	2	2,044,380	114,897	251,254	29,540
	3	2,060,428	218,358	253,227	31,040
	4	2,120,329	231,885	260,588	27,197
	5	2,347,227	242,031	288,474	31,944
	6	3,119,261	208,239	383,357	53,401
	7	3,262,524	196,922	400,964	33,437
	8	3,256,254	183,318	400,194	34,516
	9	4,024,604	259,551	494,624	46,789
	10	4,481,205	212,922	550,740	49,652
	11	4,823,719	225,157	592,835	57,888
	12	5,071,119	224,208	623,241	57,644
2021	1	3,057,196	185,413	375,729	47,670
	2	1,397,345	111,643	171,734	28,703
	3	2,360,428	240,588	290,097	61,855
	4	2,320,329	231,860	285,168	59,611
	5	2,457,227	229,289	301,993	58,950
	6	3,100,261	224,259	381,022	57,657
	7	3,692,524	157,217	453,811	40,420
	8	3,956,254	157,591	486,224	40,517
	9	4,824,604	195,906	592,944	50,367
	10	4,966,966	172,490	610,440	44,347
	11	4,823,630	245,168	592,824	63,033
	12	5,171,123	224,233	635,531	57,650

below (the conversion coefficient according to GB/T2589-2020 general rule for comprehensive energy consumption calculation: electricity 0.1229 kgce/(kW·h), water 0.2571 kgce/t).

In this paper, for the purpose of enhancing the intuitiveness of the description, the above data are plotted as energy consumption analysis graphs with the month as the horizontal coordinate unit. As shown in Figure 1, the university’s electricity consumption fluctuates greatly, with the summer and winter holidays being the lowest points and the middle of the semester being the highest. There are many reasons for this situation, and important among them is the presence of two longer holidays in summer and winter. Importantly, in the January–March period of 2020, there is a certain decrease in water and electricity usage compared to previous years, which is evident in the graph, due to the impact of the well-known COVID-19. Thankfully, China’s epidemic prevention policy is relatively effective, and the fact that Yunnan Province is in the southwest border area, which is less affected by the epidemic, ultimately led to normal teaching and living at Kunming University of Science and Technology not being affected too much. As an effective forecasting method, exponential smoothing allows for effective forecasting of short- or medium-term data, and for effective smoothing of data anomalies that do not have a significant impact, allowing the final forecast data to be more accurate. The water consumption shown in Figure 2 is relatively stable throughout the year, with a clear upward trend in summer, and fluctuations are not as pronounced as for electricity consumption. The average annual temperature in Kunming is between 12 ℃ and 22 ℃; Kunming is known as the “Spring City” because of the small temperature difference and the four spring-like seasons. Therefore, there are no air conditioning units or heating equipment installed in the university’s buildings except for research requirements. The graph shows that water energy consumption is less variable and water energy consumption is relatively flat, while electrical energy has large fluctuations due to the presence of electricity for research. Overall, electricity and water consumption show a broad general trend of increasing with the year. The model prediction error value falls within an acceptable range, and the energy consumption in 2022 and 2023 can be further predicted, which provides positive guidance for energy saving and carbon reduction in the coming years.

As observed in Figure 1 and Figure 2, the monthly electricity and water consumption are fluctuating values, and both of them fluctuate up and down in a constant range according to the time series, consistent with the prediction range of the smoothing index prediction model. By using Excel to write the smoothing formula for calculation, and further setting up the minimum mean square deviation as the target with its built-in planning solution function, the best matching smoothing index is calculated. The “planning solution” obtains the desired result from the target cell formula by adjusting the values in the specified changeable cells (variable cells). The “planning solution” can be used to determine the maximum or minimum value of a cell through the change of other cells. The steps are as follows:

• Set the target function, select the cells where the average square error is located, and the minimum of the final equity error is the convergence target;
• Set the variable cell, that is, the α value, which is the value of the smoothing index that best fits the simulation;
• Set the constraint function with 0 < α < 1 as the qualifying condition;
• Selected built-in algorithms, in which the GRG non-linear engine is very suitable for smooth nonlinear programming solution topics.

Based on the above steps, the most consistent smoothing indices are obtained with the minimum mean squared deviation as the standard, and the index values are 0.0727 for electricity consumption and 0.322 for water consumption. The corresponding squared errors of monthly electricity and water energy from 2018 to 2021 were derived sequentially using the two smoothing indices and plotted as error bars attached to Figure 1 and Figure 2 (I-beam line). Using the above smoothing indices, the calculated values of electricity and water consumption in 2019, 2020, and 2021 were calculated within 5% error from the actual values by substituting the smoothing model formulas for the annual energy consumption, with good calculation accuracy. The smoothing index predicts that the baseline discounted electricity energy consumption in 2022 is 5,423,658.235 kgce, the baseline discounted electricity energy consumption in 2023 is 5,758,865.224 kgce; the baseline discounted water energy consumption in 2022 is predicted to be 632,654.321 kgce, while the baseline discounted water energy consumption in 2023 is predicted to be 652,321.238 kgce, as shown in

Table 2. Smoothing index method predicts energy consumption.

Smooth Index	0.0727			0.3220
Project	Electricity consumption			Water consumption
Year	True Value/kgce	Exponential smoothing value/kgce	Absolute value error/%	True Value/kgce	Exponential smoothing value/kgce	Absolute value error/%
2018	4,322,795	4,322,795	/	570,200	570,200	/
2019	5,628,954	5,356,894	4.83%	618,049	590,300	4.49%
2020	4,910,656	5,054,698	2.93%	491,647	509,525	3.64%
2021	5,177,517	5,312,587	2.61%	610,781	625,663	2.44%
2022 (Predictive)	/	5423,658	/	/	632,654	/
2023 (Predictive)	/	5758,865	/	/	652,321	/

.

5.3. The Energy Consumption Forecast and Recommendations

Based on the above simulation results, the overall energy consumption of 2022 and 2023 could be predicted. Since the exponential smoothing method is a weighted smoothing of time series data used to obtain their changing patterns and trends, the combined use of both the smoothed non-linear planning method and the planning method, results in smaller forecast values and more conservative values of energy consumption for the next two years.

Combining the above projections and analysis, some effective recommendations and measures can be used to advance the achievement of the dual carbon goal. In the next step, the projected values can be used as an early warning line for the energy consumption index in 2022 and 2023 (high probability that energy consumption will exceed the warning line). In order to increase the price of related hydropower according to the price of the ladder, the ladder price can eventually have a good energy-saving effect. The operating units of the university are classified into different categories, such as teaching, scientific research, and living. According to the different categories, they will be activated one by one and given power targets to encourage out-of-pocket payments and, consequently, achieve the goal of energy saving and lower consumption. For large electrical installations, monthly or annual power hour quotas can be set to improve the efficiency of usage. Based on this model, we can develop a planning method of a quantitative energy conservation index in the future, which improves the theoretical model of energy saving, and provides long-term management methods and data support for energy conservation and carbon reduction.

6. Conclusions

In order to ensure the safety of energy use and lowering the cost of energy use, it is necessary to make a statistical prediction of energy consumption by means of the exponential smoothing method. In this way, the quantitative management is implemented, and a scientific mechanism of a peak tariff is established to achieve the purpose of intelligent control of electricity consumption in universities. The exponential smoothing method allows for statistical forecasting of electrical energy consumption and the establishment of highly quantifiable management targets based on the forecast results. Then, simple, efficient, and effective carbon emission reduction management measures can be developed to reduce wastage. We advocate that all students, faculty, and staff become “socialist construction of energy saving and emission reduction volunteers”, start from themselves, start from the little bit, and build a conservation-oriented university together!