Mathematical Problems in Engineering

Volume 2018, Article ID 4671850, 11 pages

https://doi.org/10.1155/2018/4671850

## Forecasting the Electricity Demand and Market Shares in Retail Electricity Market Based on System Dynamics and Markov Chain

Correspondence should be addressed to Chao Qin; moc.361@80oahcniq

Received 29 August 2017; Revised 11 January 2018; Accepted 23 January 2018; Published 20 February 2018

Academic Editor: Quanxin Zhu

Copyright © 2018 Qingyou Yan et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

#### Abstract

Due to the deregulation of retail electricity market, consumers can choose retail electric suppliers freely, and market entities are facing fierce competition because of the increasing number of new entrants. Under these circumstances, forecasting the changes in all market entities, when market share stabilized, is important for suppliers making marketing decisions. In this paper, a market share forecasting model was established based on Markov chain, and a system dynamics model was constructed to forecast the electricity consumption based on the analysis of five factors which are economic development, policy factors, environmental factors, power energy substitution, and power grid development. For a real application, the retail electricity market of Guangdong province in China was selected. The total, industrial, and commercial electricity consumption in Guangdong from 2016 to 2020 were predicted under different scenarios, and the market shares of the main market entities were analyzed using Markov chain model. Results indicated that the direct trading electricity would account for 70% to 90% of the total electricity consumption in the future. This provided valuable reference for the decision-making of suppliers and the development of electricity industry.

#### 1. Introduction

The new round of electricity market reform (hereafter referred to as the “new electricity reform”), which was launched in 2015, included the reform of transmission and distribution price, the orderly deregulation of electricity generation and consumption plans, and the construction of electricity market. Among them, the deregulation of retail electricity market will bring about important welfare gains [1]. It will change the previous monopoly situation of power grid enterprises and provide opportunities for power generation companies, energy-saving service companies, industrial parks, and so on to create electricity sales companies and then join the market competition. Consequently, the cost of suppliers will be reduced and welfare gains will be brought to industrial enterprises and residential consumers [1]. In a word, the electricity market opening will have positive impact on economic growth and sustainability of power industries [2]. However, the reform of electricity market will inevitably change the current market structure and bring fierce competitions among suppliers. In this situation, it is of great significance to choose an appropriate method to predict the electricity market; an accurate prediction is conducive to the decision-making of suppliers, the development of power industry, and even the regional economic progress.

Nowadays, numerous scholars have carried out researches on various subjects relevant to the retail electricity market such as business model [3], pricing strategy [4], risk management [5, 6] and electricity price prediction [7–11]. In the matter of energy demand, various forecasting models have been applied to predict the electricity consumption. Traditional methods such as regression analysis (RA), time series analysis, nonparametric method, and small-sample adaptive hybrid model as well as soft computing techniques such as fuzzy logic, genetic algorithm, and neural networks are being extensively used [12–17]. Support vector regression, ant colony, and particle swarm optimization are new techniques being adopted for energy demand forecasting. Bottom up models such as MARKAL and LEAP are also being used at the national and regional level for energy demand management [18]. However, most of them are not appropriate for the forecast of deregulated electricity market of China because of their complicated input factors and the lack of available observation [19].

Markov chain is a stochastic process model with discreteness [20–22] and presents a quantitative analysis of the status of a system from one state to another. It has been widely used in the energy market, such as biogas production [23], wind power forecasting [24], crude oil import [25], and energy supply and demand [26], and has provided strong theoretical support for the price and demand forecast. Also, Markov chain can be used for macroeconomic analysis, for example, dynamic forecasting of industrial structure [27]; these analyses provide reference for China’s industrial policy and other macro-development strategies. Moreover, Markov chain has a wide range of applications in the field of electricity. It has been used to analyze the reliability of power system, energy consumption structure, and the combination of electric load [28–32]. For instance, Sun and Xu presented an improved Grey-Markov chain model based on the wavelet transform to predict the energy production and consumption of China [26], Hong et al. optimized the size of renewable energy generations in a community microgrid using Markov model [33]. In conclusion, Markov chain is an effective method to predict the market volatility and describe the unobservable states. Since the reform and opening up, the economy in China has been in a state of rapid development. Therefore, this paper constructs a market share forecasting model based on Markov chain.

In order to forecast the market share of every market entity, deregulated electricity which means the part of electricity allowed to be sold by various suppliers in retail market needs to be figured out. According to the regulation of Chinese government, deregulated electricity is determined by industrial and commercial electricity consumption. Therefore, this paper uses system dynamics to predict the electricity consumption after the deregulation of retail market.

System dynamics is an interdisciplinary approach which combines the feedback and control of information, the decision-making theory, and computer technology. It is especially suitable for simulating the behavioral characteristics of nonlinear, high order, and complex time-varying systems [34] and is widely used in many fields. Energy systems are complex dynamic systems that are often associated with uncertain system behavior which is influenced by dynamic uncertainties, nonlinear relationships between system variables, interactive feedback loop, and so on [35]. The widespread deregulation leads to an ever-increasing size and complexity of energy systems. Under this climate, system dynamics as a system analysis approach is useful for organizing knowledge of the energy system in an efficient way. Recent studies applying system dynamics in energy system have focused on the alternative energy. For instance, Sisodia et al. (2015) explored solar energy as an alternative source of energy for the fulfillment of energy demand in India through a system dynamics approach [36]. Saavedra M. et al. (2018) identified the latest system dynamics contributions and trends related to the supply chain of renewable energy [37]. However, the influence factors of electricity consumption are complex and highly linked to each other. Due to the unique advantages of system dynamics in multivariate quantitative analysis and causal feedback, this paper constructed a system dynamics-based model to predict the electricity consumption.

To the best of our knowledge, system dynamics and Markov chain based methodology for the electricity demand and market shares, in Chinese context, is scarce in the energy literature. Therefore, this paper contributes to the existing literature by proposing the system dynamics and Markov chain based methodology that would have decision-making and policy implication for both the suppliers and government.

The rest of the paper is structured as follows: Section 2 determines the transition probability matrix by learning from relevant literature [38] and establishes the market share forecasting model based on the Markov chain; Section 3 analyzes current state and trading mode of the retail electricity market, constructs the system dynamics-based electricity consumption model by selecting five factors which are policy, environment, power energy substitution, macroeconomics, and power grid development, and sets six scenarios representing different economic development and industrial restructuring rates; Section 4 predicts electricity consumption and market shares in Guangdong province during 2016–2020 under different scenarios; Section 5 concludes the paper.

#### 2. Market Forecast Based on Markov Chain

The reform of retail electricity market severely impacts the monopoly status of power grid enterprises resulting in a huge change in the structure of retail market. Using Markov chain model to predict and analyze the retail market is conducive to the identification and evasion of market risks facing the power grid enterprises. Consequently, it offers some grounds for power grid enterprises making investment decision about whether to establish their own electricity sales companies.

##### 2.1. Markov Chain Method

A Markov chain is a sequence of random variables at different states, where each state value depends on the previous finite state. The range of these variables, that is, the set of all their possible values, is called the “state space.” In every step of the Markov chain, the system can change from one state to another according to the probability distribution, and it can also maintain the current state. The change of state is called transition, and the probability related to state changes is called the transition probability.

Set random variable sequence as , and state space, , is an countable or finite set, for any positive integer, , ; if and satisfy (1), the is called Markov chain:

Suppose that the conditional probability on the right side of the equal sign is independent of ; then (1) can de denoted aswhere is the transition probability from state to state after steps.

The transition probability, , is only related to the number of steps but is not related to the initial time. If the number of steps , the one step transition matrix of the Markov chain is expressed as follows:

##### 2.2. Steps of the Retail Electricity Market Forecast

###### 2.2.1. Construction of Initial Transition Probability Matrix

In order to construct the initial transition probability matrix, the market shares of all suppliers need to be obtained and the electricity consumers switching from one supplier to another during the next period should be investigated. These could be acquired by compiling the electricity sales in a region. This paper assumes that the switching of electricity consumers will not change over time in the future.

Suppose is the transition probability that the consumers of the th supplier continue to purchase electricity from that supplier. indicates the probability from the th supplier to the th supplier. Then the transition probability matrix can be obtained:

The relation between the -step state transition probability matrix and the transition probability matrix is given by

In the early days of the establishment of electricity sales company, the work investigating customer switching in the retail market is heavy and inefficient. An estimate method for initial transition probability matrix can be expressed as follows:

Through (6)–(10), the initial transition probability matrix can be obtained, and .

###### 2.2.2. Construction of Market Share Forecasting Model

Given the initial market share obtained by market investigation, the initial state transition probability matrix can be constructed:

Set the as the market share of each supplier during the th period; then the market share forecasting model during the th period can be obtained as follows:

###### 2.2.3. Calculation of the Market Share in Equilibrium

When the market is in equilibrium, it is known that according to (7); by combining it with the equation , the unique solution to the market share of each supplier can be obtained.

#### 3. Market Analysis and Electricity Consumption Forecasting Model

##### 3.1. Market Analysis

###### 3.1.1. Current Situation

On March 1, 2016, two power trading centers in Beijing and Guangzhou were set up and marked the reform of electricity market. These two national-level power trading centers are no longer the wholly owned subsidiaries of power grid enterprises; they provide nonprofit and normative services publicly and transparently for both sides of power trading under the government regulation.

According to a series of market access qualifications established by National Development and Reform Commission (NDRC) for suppliers, the total asset of a supplier corresponds to its electricity sales. For example, for a supplier whose registered capital is between 100 million and 200 million CNY, its annual electricity sales are between 3 and 6 billion kWh. The main retail electric entities include power generation enterprises, power grid enterprises, industrial parks, and energy-saving service companies. The diversification of market entities means that the retail electricity market is getting severely competitive. Although new suppliers continually register or apply for registration, most of them have not started selling electricity, and the electricity sales are far from the upper limit.

###### 3.1.2. Trading Modes

In 2016, the National Energy Bureau specified the types of trading modes in the context of the deregulation of retail electricity market. The first type is direct power-purchase; that is, large electricity consumers purchase power from the power generation enterprises directly. In the second type, entities purchase electricity using the suppliers as agents; for instance, large consumers or industrial parks choose a supplier to help purchase electricity. The third is that electricity consumers who do not participate in the trading market continue to purchase electricity from power grid enterprises. Power users can choose among these three ways. The emergence of suppliers as power purchasing agents and the expansion of direct power-purchase will affect the market scale of power grid enterprises. Figure 1 displays the trading modes of all types of consumers before and after the reform of retail market, respectively.