Non-intrusive monitoring and disaggregation of industrial load based on Petri net theory

In this paper, the non-intrusive monitoring and disaggregation method of industrial load based on Petri net theory is studied. First, the basic concepts of Petri nets are introduced. On the basis of prototype Petri nets, time factor is added. Then, the Petri net with time factor is combined with industrial load production process. Considering the production sequence, start-stop time and production time constraints of industrial load, a non-intrusive monitoring and disaggregation model of industrial load is constructed. Finally, according to the actual production process of industrial load, the electricity consumption situation of each production link is calculated, and the detailed analysis results of industrial user’s electricity consumption behavior are obtained. The research could provide an important reference for realization of friendly interaction between industrial users and power grid.


Introduction
With the development of society and the continuous growth of electricity demand, power saving has direct and critical effects on the sustainable development of the economy and the environment. In the fields of high efficiency, saving and optimizing power consumption, the equipment platforms which carry these technologies in hardware and equipment have broad market demand and development prospects. In recent years, the rapid development of smart grids has also triggered a research boom on "smart power" technology [1][2][3]. Document No. 9 of Electric Power Reform clearly states that demand side management and energy efficiency management should be actively carried out to promote supply-demand balance and energy saving. From this point of view, detailed analysis and in-depth mining of user energy consumption information will provide basic information reference for the above related research. For example, the premise of implementing demand response scheme in smart grid is to collect and analyze load data [4][5][6].
From the perspective of existing demand, load monitoring and energy analysis have received more attention. Electric load monitoring refers to monitoring the operation status and energy consumption of different loads through real-time acquisition of load's electrical parameters. It is a key basic technology for user-side energy management [7]. By understanding the state mode, power consumption period, power consumption level of various loads, it can help the energy users to distribute electricity reasonably and optimize the mode of energy consumption, so that the whole SAMSE 2019 IOP Conf. Series: Materials Science and Engineering 768 (2020) 062111 IOP Publishing doi:10.1088/1757-899X/768/6/062111 2 power system is always in a state of efficient operation, which is of great significance in today's emphasis on energy saving and emission reduction.
The existing research divides load monitoring into intrusive and non-intrusive according to the way the load data is obtained. In terms of non-intrusive monitoring, the literature [8] uses the pattern recognition method to design Fisher's supervisory load identification method, but only classifies the electrical appliances and cannot identify specific electrical appliances. According to the principle of load feature superposition, the literature [9] uses the genetic algorithm to realize the decomposition of the total signal ，which is based on the characteristics of load active power and 3rd harmonic current amplitude. In the literature [10], a load decomposition algorithm for sparse underdetermined solution is proposed, which decomposes the running electrical appliances from the total load signal. However, the algorithm has a large amount of calculation, and it is difficult to realize real-time monitoring when the load is large. In the use of non-intrusive monitoring for power behavior analysis, the current domestic research [8][9][10] mainly stays on the exploration of resident load identification algorithms, and few studies on industrial load identification algorithms.
This paper first introduces the basic concepts of Petri net and defines a mapping from transition set to some time factor set on the basis of prototype Petri net. Then, the Petri net with time factor is combined with the industrial load production process. Considering industrial load production process sequence, start-stop time and production time constraints from the perspective of time, a non-intrusive monitoring and disaggregation model of industrial load is constructed. Finally, according to the actual production process of industrial load, the electricity consumption of each production link is calculated. The research could provide an important reference for realization of friendly interaction between industrial users and power grid.

Petri net theory
Petri net theory is a model for describing distributed systems. It can describe the structure of the system and simulate the operation of the system [11][12][13]. Petri net theory has strict mathematical expressions, intuitive graphical representations, rich system description methods and system behavior analysis techniques. Here we only give a few concepts that are closely related to this article.
Definition 1 A triple ( ) , ; N P T F = that meets the following conditions is called a net: P and T are two disjoint sets, which are called the basic elements of network N, the elements of P are called places, the elements of T are called transitions, and F are the flow relations in the network.
is a marked network and has the following transition firing rules.

Non-intrusive monitoring and disaggregation model of industrial load
Traditional load monitoring takes an intrusive approach, in which sensors are installed on each user's electrical equipment to record their usage. The advantage of this method is that the monitoring data is more comprehensive. The disadvantages are poor practical operability, high implementation cost and high maintenance cost. Generally, multiple data acquisition equipment and sensors are arranged to monitor the load equipment at a fixed point, and the installation work needs to enter the interior of the building. On the other hand, because this method uses more sensors, there are more interference factors and errors, which will also affect the reliability of load operation and data integrity [14,15]. In this paper, the non-intrusive monitoring method is used to install the data acquisition device at the initial end of the industrial load. Its advantages are low economic cost, little disturbance to the production and life of industrial users, low installation cost and convenient maintenance of equipment.

Industrial load production workflow model
As shown in Fig.1, a process is represented by two transitions and a place; where 1 i t is defined that the beginning of the process i , 2 i t is defined that the end of the process i , a mark in the place i s is defined that the process i is in progress. A time value i a is assigned to the place i s , indicating that it takes at least i a unit time from the 1 i t occurrence before 2 i t can occur.

Figure 1. A process is represented by Petri net model
For the whole project, the following steps can be used to construct the time-delay Petri net workflow model.
(1) If the process i is the premise process of the process j , then add a place ij s between 2 i t and and ij s is endowed the time value of 0.
The operation process of ∑ reflects the construction process of the project, when the ∑ runs to the termination mark t M :

Non-intrusive monitoring and disaggregation model based on Petri net theory
The steady-state power of the load can reflect the energy consumption state and energy consumption level of the load, which is the most intuitive feature type among many load characteristics. When different loads are put into use or exited from operation or even switched to the working state, their corresponding power characteristics will be changed, thereby changing the total power consumption or total power consumption. The steady-state power characteristic of load satisfies superposition and is easy to analyze and process. Regardless of noise interference, when the total load power changes, along with the occurrence of load events, the goal of non-intrusive load disaggregation is to monitor and identify the working state of the power load, and complete the total load to the timing. The running track of the load is reproduced.
The target of load identification is the load equipment working sequence combination based on the process operation mark of the Petri net with minimum distance to the measured total load power, which can be expressed as where i T is the operation time of the process i ; ic t is the actual closing time of the process i .
Please keep a second copy of your manuscript in your office. When receiving the paper, we assume that the corresponding authors grant us the copyright to use the paper for the book or journal in question. Should authors use tables or figures from other Publications, they must ask the corresponding publishers to grant them the right to publish this material in their paper.

Simulation example
Setting a production process of the industrial load consists of 12 processes. The connection relationship between each process code, running time and process is shown in Table 2. Based on the industrial load production workflow model proposed in part 3.1, a Petri net with time weights in the place can be constructed. After eliminating some zero-weight places by simplification, the workflow Petri net model is shown in Figure 3.  Figure 4. It can be seen that as the industrial load production process advances, its power consumption curve is constantly changing.  Table 3. The non-intrusive load monitoring technology can conveniently monitor the electricity consumption of each production link, and get accurate electricity usage behavior of industrial users. Based on the Petri net theory, the non-intrusive monitoring and disaggregation model analyzes the power consumption of a production cycle, and the input time, cut-out time and power of each production link are shown in Table 4.

Conclusions
In the non-intrusive monitoring and disaggregation of industrial load, the power feature has a poor recognition effect on the load, and the recognition accuracy is low when multiple loads are simultaneously operated. In order to solve the above problems, this paper introduces the time factor based on the study of Petri net theory, aiming at analyzing the industrial load production process from the time level. The experimental results show that the proposed method has good recognition accuracy and stability, and the accuracy of low-power load that is difficult to identify is significantly improved. Through the research of this paper, it is difficult to identify the industrial users' multi-production process based on a set of monitoring devices. Therefore, based on the research of this paper, it is the next research direction of this paper to continue the load identification and disaggregation of industrial users' multi-production processes.

Acknowledgments
Thanks to research on the diversified trading mechanism of promoting new energy consumption in electricity market of Ningxia Power Exchange Center Company, the design service of power market simulation trading platform of Nanjing SAC Power Grid Automation Co., Ltd and study on the market mechanism of large-scale clean energy consumption across regions of NARI Group Corporation (State Grid Electric Power Research Institute) for supporting this paper.