A unified model for diagnosing energy usage abnormalities in regional integrated energy service systems

1 Introduction

With the rapid socioeconomic development of China, the danger of an energy crisis continues to intensify.According to current statistics, the per capita resource ownership in China is less than half of the world average, with 70% of these resources being unrenewable fossil fuels.Furthermore, the energy consumption per unit of gross domestic product in China is 2.5 times that of the global average [1-3].Therefore, China is vigorously developing its renewable energy industry (e.g., wind, solar) while promoting energy savings and emissions reduction nationwide.The aim of these efforts is to improve energy efficiency and solve issues like energy-resource shortages, irrational allocation of energy structures, climate deterioration, and low energy and capacity utilization rates [3-4].

Integrated energy systems (IESs) are a new type of energy service whose purpose is to fulfill the various energy generation and consumption needs of end users.In an IES, multiple energy sources (e.g., electricity, gas, cooling, and heating) are horizontally coordinated so that they complement each other and operate at an optimal level.An IES, therefore, enables high-efficiency operations and source-grid-load-storage coordination and control in the production, transmission, conversion, storage, and consumption of vertically coordinated and controlled energy sources [4-7].

Regional integrated energy management systems are used to manage the various energy facilities and resources of a region (e.g., power distribution, gas supply, heat pipeline, and water supply networks) in a unified manner.The regions being managed by these systems include large industrial parks, large development zones, and new cities, which may, in turn, contain large industrial and mining enterprises, multiple commercial buildings, and residential suburbs.Regional energy service companies collect the energy consumption data of end users, including electricity, water, gas, heating, and cooling consumption data to provide real-time and historical data analytics and comparisons.The energy usage efficiency may then be assessed and analyzed via various indices, whereas energy efficiency diagnoses can be used to discover problems in the processes and structures of end-user consumption [8-11].

Some energy-intensive enterprises in China have independently developed real-time energy consumption monitoring and management systems to save energy, reduce emissions, and lower costs; however, these systems generally cannot analyze energy-flow directions in real time and are incapable of detecting energy usage abnormalities.Additionally, these systems are also unable to analyze and assess energy consumption, resulting in a lack of supporting data for the assessment of energy savings.There is also a lack of effective and accurate energy consumption data from production sites, which contributes to high industrial energy consumption rates and difficulties in achieving energy savings/emission reduction objectives [12-15].To address these issues, the Xuji Group developed a regional IES that supports four major functions: energy monitoring and dispatching, energy efficiency analyses and diagnoses, energy service maintenance, and energy trading.Energy efficiency and diagnosis refers to the use of data mining and artificial intelligence techniques to discover sudden changes in user energy consumption and energy wastage and to provide diagnostic recommendations and energy optimization strategies for the optimization of user energy operations and energy usage structures.This improves the efficiency of extant power generation equipment and thereby helps save energy, improve efficiency, and facilitate high-efficiency energy usage.

Electricity/water/gas/cooling/heating data are typically decentralized, multivariate, and spatiotemporally complex by nature.Big-data techniques can be used to model, store, and manage these multi-source heterogeneous data in a unified manner to enable the standardization, management, and analysis of energy information.However, different standards and units are used in the daily reports of energyuse data acquisition systems and those produced by smart meters.This causes difficulties in energy-use diagnosis and energy wastage detection.To address this, we herein propose a unified model for the diagnosis of energy usage abnormalities.This model can be used to detect sudden changes in energy usage and energy wastage and can thereby locate faulty acquisition devices, prevent energy leakages, and/or provide data for locating energy theft.

2 Energy efficiency analytics and diagnostic systems for IESs

Regional IESs are usually used in ordinary residential suburbs, by commercial users/public institutions, and by large industrial clients.For residential users, our IES provides various basic IES efficiency analysis services, such as monitoring and display of electricity, gas, cooling, and heating usage.These energy data are summarized and analyzed by the IES for the calculation of statistical indicators and provision of management tools.The IES also provides historical data for year-on-year or seasonon-season comparisons, which can be used to highlight periods in which sudden changes in energy usage occur, thus drawing attention to possible energy wastage.For commercial users or public institutes, our IES provides a more comprehensive set of energy efficiency analytics and diagnostics in addition to basic energy efficiency analytics available to residential users, such as the ability to perform comprehensive comparisons among zones, units, periods,and devices.These analytics include historical energy-use data comparisons (i.e., year-on-year, season-on-season, or peak-versus-dip comparisons) on a daily, weekly, monthly, or yearly basis.Benchmark analyses are also included.The yearly energy consumption indices of a subject are compared to its benchmark energy consumption indices, and energy consumption reports are included.Energy usage is output in daily, monthly, and yearly reports, which can be used to highlight sudden changes or identify periods and locations in which energy wastage may be occurring.The most comprehensive energy efficiency analytics and diagnostic systems are given to large industrial clients.In addition to the functions available to the other two categories of users, comprehensive energy efficiency evaluations are also provided to large industrial clients, including quota analysis.This function mathematically compares actual energy usage to an energy usage quota defined by the system, and energy usage cost analyses are included.Maximum and minimum energy usage, prices, and overall costs are calculated and displayed by this function.Comprehensive energy efficiency analysis uses the user’s conversion efficiency, usage efficiency, energy consumption per unit of production, energy consumption per unit area, and energy consumption per person to discover problematic processes and energy usage problems, thus providing an effective tool for resolving energy wastage [16-17].

3 Data processing procedures for the diagnosis of energy usage abnormalities

Ninety-six energy load datapoints are acquired daily from smart meters at our regional IESs.Thus, one datapoint is acquired every 15 min.Energy load data include information about electric power usage (i.e., electric energy, voltage, current, power, active power, and reactive power), water supply usage (i.e., cumulative flow, instantaneous flow rate, and pressure), gas usage (i.e., temperature, pressure, instantaneous flow rate, and cumulative standardized flow), and cooling/heating usage (i.e., temperature, pressure, cumulative flow, and instantaneous flow rate).The energy data are first converted into a standard data format by a report processor within the forward server before they are imported into a distributed real-time database.Then, the data are stored in a distributed historical database.The functions of the IES are realized using a big-data analytics engine to utilize the information contained by real-time and historical databases to perform real-time energy usage monitoring and energy efficiency analytics and diagnostics.The procedure by which the integrated energy data are processed is shown in Fig.1.Energy usage abnormality diagnostics can be divided into two parts: sudden changes in energy usage and energy wastage.

The diagnosis of sudden changes in energy usage is performed by comparing current energy-use data with a benchmark value, determined using the historical data of a user over the past 30 days.The sudden-change datapoints are then recorded in the historical database and displayed on a graphical user interface (GUI).Big-data mining analysis is then performed to identify the cause of the sudden changes in usage.Energy wastage diagnoses are performed by comparing current usage levels to some pre-defined energy wastage thresholds at midday and nighttime and during non-working day periods.This will allow a user to identify energy wastage points.The procedures used for the diagnosis of energy usage abnormalities and energy wastage are shown in Fig.2.

Fig.1 Block diagram of the IES data processing procedure

Fig.2 Unified model for diagnosing energy usage abnormalities

4 Unified model for diagnosing energy usage abnormalities

The load and usage characteristics of the water, electricity, gas, and heat supply systems managed by a regional IES company are intrinsically different.Therefore, the design of a unified management model must be based on the distribution of actual energy-use data [18-19].

(1) In our model, different criteria are used for each target user.For example, the energy usage of office buildings is expected to be much higher on working days than on non-working days, whereas the opposite is true for shopping malls.

(2) Clustering calculations are performed on the historical data from the prior 30 days, with working days and non-working days grouped into separate clusters.

(3) The distances of the current-day energy usage from the centers of the working day and non-working day clusters are compared.The energy usage of the current day is then grouped into the closer cluster.

(4) The datapoints are weighted per their temporal distance from the current date.

(5) Normalization is performed, and the benchmark for the current-day energy usage is calculated.

(6) The actual current-day energy usage is compared to the benchmark value, and the determination calculations are performed.

The determination criteria for sudden changes in energy usage and energy wastage are as follows:

Where Etoday is the current-day energy usage, Estd is the predicted benchmark value, ε1 is the threshold for sudden changes in energy usage (relative value), Esum1 is the cumulative sum of energy usage within the given interval, and ε2 is the threshold for energy wastage.The periods analyzed for the judgment of energy wastage are as follows: midday (12:00-14:00 LST), nighttime (18:00-08:00 LST), and nonworking days (Saturday + Sunday + public holidays).

4.1 Data preprocessing

The influence of noise data on the calculation results needs to be considered.Data preprocessing requires removing some of the noise beforehand.

The average value is:

The variance is:

The noise judgment is:

4.2 Sample data classification

The sample data are classified by analyzing daily energy-use values.Our vertical recognition scheme is as follows: N1 is the number of nodes in the working days cluster; N2 is the number of nodes in the non-working days cluster; and xi represents a datapoint in the historical list of daily energy usage over the previous 30 days.

The center of the working days cluster is:

The density of the working days cluster is:

The center of the non-working days cluster is:

The density of the non-working days cluster is:

Fig.3 The energy usage abnormalities display screen

4.3 Noise reduction

Noise reduction is performed according to cluster density:

The distances of the current values (i.e., the actual current-day energy usage) from the centers of the working day and non-working day clusters are calculated.The current value is then grouped into the closer cluster.

4.4 Weighting calculations

Weighting is performed according to the date of the datapoint: the closer the date, the greater the weighting.The normalized weighting formulae are:

4.5 Benchmark value calculations

Historical data are then weighted to obtain the current predicted benchmark value.The predicted benchmark value, Estd, is given below, with N being the number of cluster nodes in the given category:

5 Analysis of abnormal IES data

After the IES data are judged for sudden changes and energy wastage determinations, the results are then displayed as plots on the GUI.This can be used to search for the cause of sudden changes in energy usage, which could be caused by normal or abnormal changes in energy usage.Normal changes refer to those based on the operational conditions of the users themselves, such as the sudden addition of production tasks or overtime work at night or during midday periods.Abnormal changes refer to faults in metering devices or user-induced energy wastage.The energy usage abnormalities screen is shown in Fig.3.The red balloons indicate instances of abnormal energy usage.These plots therefore allow the times and magnitudes of all energy usage abnormalities to be identified with ease.

Normal changes in energy usage can be resolved by system management personnel, whereas abnormal changes in energy usage can be imported into data mining analysis to search for the possible causes of the changes.Abnormal changes in energy usage are generally divided into three categories: data acquisition device abnormalities, metering abnormalities, and energy usage abnormalities [5-6].

In intelligent terminals, data acquisition device abnormalities include communication faults, warnings, incorrect times, abnormal parameters, and an excessive number of resets.

Metering abnormalities include energy theft by a user, phase loss, inconsistencies among smart meter totals, and the sum of all rates, smart meter measurement anomalies, reverse active power values greater than 0, reversals in smart meter counts, stopped meters, incorrect times on meters, the covers of electricity meters being opened, meter opening and closing, phase sequence abnormalities, and meter-rate device abnormalities.For example, the condition for determining phase loss for users of specialized 10-kV transformers is the accumulation of more than three datapoints in which the voltage of the A phase or C phase is lower than 80% pu in a single day in a three-phase, three-wire system.

Energy usage abnormalities include anomalous voltages/currents/temperatures beyond specified limits, current imbalances greater than the threshold value, overcurrent anomalies, electric meter demands greater than the capacity of the transformer, loads greater than the capacity of the transformer, power factor anomalies, voltage imbalances greater than the threshold value, loads continuously below the lower limit of expected loads, anomalous power differentials, power-flow reversals, and power outages.For example, the conditions for determining that the current imbalance threshold has been exceeded are the (maximum current - minimum current) / maximum current ＞ 15% and the accumulation of more than 10 such datapoints in a single day.

6 Conclusion

China is currently facing a battery of environmental and energy issues, including a resource crisis, low energy and facilities utilization rates, wastage of renewable energy resources, environmental pollution, and climate change.IESs provide an effective means to coordinate and optimize the use of multiple energy sources, promoting advanced energy technologies and reducing energy costs.These systems have already become popular in recent years [20].In this study, we proposed a unified model that enables the processing of multiple energy sources (e.g., electricity, gas, water, cooling, and heating) and the diagnoses of energy usage abnormalities based on IES energy efficiency analytics and diagnostics.This system enables the detection of abnormalities and alerts users to these abnormalities.Furthermore, our model also uses big-data mining and analysis techniques to search for the root cause of data anomalies.The algorithms and data processing procedures described in this article have already been applied in regional IESs developed by the Xuji Group.This system has been applied with excellent results in an IES pilot project of a business center in a suitable region.

Acknowledgements

This work was supported by National Key Research and Development Program of China (No.2017YFB903304) and the State Grid Science and Technology Program (Hybrid Simnlation Key Technology for Integrated Energy System and Platform Construction).