As a new form of smart grid, the energy transmission mode of the Energy Internet (EI) has changed from one direction to the interconnected form. Centralized scheduling of traditional power grids has the problems of low communication efficiency and low system resilience, which do not contribute to long-term development in the future. Owing to the fact that it is difficult to achieve an optimal operation for centralized control, we propose a decentralized energy flow control framework for regional Energy Internet. Through optimal scheduling of regional EI, large-scale utilization and sharing of distributed renewable energy can be realized, while taking into consideration the uncertainty of both demand side and supply side. Combing the multiagent system with noncooperative game theory, a novel electricity price mechanism is adopted to maximize the profit of the regional EI. We prove that Nash equilibrium of theoretical noncooperative game can realize consensus in the multiagent system. The numerical results of real-world traces show that the regional EI can better absorb the renewable energy under the optimized control strategy, which proves the feasibility and economy of the proposed decentralized energy flow control framework.

The third industrial revolution is emerging, represented by new energy technologies and Internet technologies. Construction of Energy Internet (EI) can promote industrial technology upgrading and structural adjustment in modern energy industry [

As a subnetwork of Energy Internet, energy local network (ELN) has a variety of energy absorption methods. The ELN is a multienergy operation system, where different energy networks have strong coupling. The complementarity of energy can greatly improve the energy efficiency of the system to achieve cascade utilization of diverse energy sources [

Due to the diversification of load patterns and stochastic nature of renewable energy sources in EI, the traditional centralized optimization scheduling method is difficult to apply in practice in actual operation. How to cope with the problem has become urgent to be solved in the management and optimization operation of EI [

Motivated by the above facts, we analyze the operation pattern of energy flow in a typical architecture of the ELN framework in order to obtain the optimal operation of multienergy. A strategy of optimal operation in EI is proposed based on the multiagent system (MAS) combined with noncooperative game theory to realize the decentralized control of the ELN system. The real-time electricity price is obtained by iterative optimization, which maximizes the overall profit of the EI system. For each ELN, a quadratic programming problem is developed with the aim to increase the individual economic benefit. Through energy trading and conversion between ELNs, the balance of supply side and demand side of the overall EI system is enhanced. As a result, the resilience and economy of the EI system are significantly enhanced.

The remainder of the paper is organized as the following. In Section

ELN is a collection of a complete future-oriented energy system constructed from the aspects of energy production, transmission, distribution, transformation, and consumption. It is a power-centric interactive and shared platform for all kinds of energy which enables smart mutual supply of different types of loads. Figure

The typical architecture of ELN framework.

The EI consists of ELNs, which are highly coupled products of multiple type energy and information networks. With huge number of diverse energy equipment in EI, centralized control cannot deal with the rapidly increasing data complexity, and it is difficult to make full use of the energy in EI. As a result, the stability and economy of the power system will be impacted. Based on the aforementioned, we propose a decentralized energy flow control framework of multiagent EI to effectively improve computing and execution of system operating.

A discrete-time model is considered in this paper. Assume that the optimal range (e.g., 24 h) is divided into

As the basic scheduling unit in EI, the energy storage system (ESS) can compensate for the power difference caused by the volatility of the RESs and the load. The model of the ESS can be described as follows:

We assumed that all energy storage systems have the same lithium-ion battery pack and the charging/discharging power over a single period of time is considered constant [

Since the SOC at period

Gas turbine is a vital device for EI with high efficiency, which can fully utilize natural gas energy and contribute to reducing environmental pollution. The output of the gas turbine is expressed as follows:

The active power output of the gas turbine is bounded by the ramping constraints, denoted as

The gas turbine waste heat is mainly recycled by heat exchangers and adsorption refrigerators for refrigeration. Specific physical modeling is shown as follows:

Heat exchanger:

Adsorption refrigerator:

Gas boiler:

When the thermal power of the system is insufficient, it is supplemented by the heat generated by the gas-fired boiler. The output thermal power of the gas-fired boiler is expressed as follows:

The scenario in power grid is a kind of operation state of the power system [

The random distribution error is obtained by the forecasting error and its probability distribution which is determined based on historical data. The random variable of RES is converted to output power based on the output characteristic curve. In this paper, the predicted value of RES output at period

RES system operates in the Maximum Power Point Tracking (MPPT) mode which can adapt to environmental changes in real time to achieve maximum output [

The total power of RES output in this paper is

According to the components of the power supply side and the demand side in the ELN, the power balance model can be obtained:

Multiagent is a network structure composed of agents with the characteristics of autonomy, decentralized control, and bidirectional communication with other agents [

Intelligent measurement agent (IMA): it monitors and reports the operation status, power output status, and load demand status of internal equipment in the ELN system, which is responsible for the monitoring for the balance of the supply side and the demand side.

Scheduling management agent (SMA): according to the information uploaded by the IMA and electricity price agent (EPA), the internal equipment output optimization is executed. When there is a shortage or excess load, the information is reported to the EPA for further addressing.

Electricity price agent (EPA): it receives information of each ELN by the IMA and SMA. According to the real-time supply-demand balance of the system, the global optimal equilibrium solution is calculated. As the most essential agent, the strategy of maximizing EI benefits is the consensus reached by all EPAs in decentralized decision-making [

Different from the previous method of static electricity price that determines its own electricity price by the power grid [

Since multiple energy sources can be converted into electricity, we use electricity as the core of trading in the energy flow control mechanism. In the electricity market of EI, the agent of electricity price takes part in the bidding to maximize the benefit of ELN.

The key problem is to obtain the optimal electricity price which maximizes the profit of ELN:

The mixed integer model is established by the problem of optimization in optimization periods; the following are the detailed components of the game model.

The players are all the agents of electricity price in the set

Action: for any

Information: it includes RESs and various demand load and strategies adopted by other players.

Strategies: each participant’s revenue is maximized by an optimized strategy, which can be expressed as a feasible strategy set

It is used to measure the benefit of the players in the game; the payoff of each player is maximized, expressed as

Based on the above set of strategy,

The strategy

Prove that

Since

In the game model, each agent can know others strategies in each round of decision-making [

According to the aforementioned models, the proposed decentralized energy flow control strategy will be used to determine the regional ELN scheduling plan, which is shown in Algorithm

Initialization:

Transmit initial strategy according to players’ requirements to its neighbors.

A new iteration starts when the strategy information is updated.

Input predicted value of

Update

Send the new strategy to the MAS.

Denied, retain the previous control strategy.

To verify the validity of the proposed energy flow control strategy, four typical ELNs with different structures are chosen for case analysis [

The basic parameters of ELN.

Types | PV (kW) | WT (kW) | BESS (kWh) | Gas turbine (kW) |
---|---|---|---|---|

ELN 1 | 3000 | 2500 | 13000 | 2000 |

ELN 2 | 3300 | 3550 | 13000 | 2000 |

ELN 3 | 4000 | 3900 | 13000 | 1000 |

ELN 4 | 3500 | 3750 | 13000 | 1000 |

The parameters of energy conversion equipment.

Types | Heat exchanger (kW) | Absorption chiller (kW) | Gas boiler (kW) | Electric refrigerator (kW) |
---|---|---|---|---|

ELN 1 | 2000 | 2000 | 1500 | 1000 |

ELN 2 | 2000 | 0 | 1500 | 0 |

ELN 3 | 0 | 2000 | 0 | 1000 |

ELN 4 | 0 | 0 | 0 | 0 |

The output power of RESs is given in Figure

Active power of PV and WT output in a typical day.

The cooling and heating loads of ELNs.

In order to precisely quantify the optimization effect of the proposed energy flow control strategy, in this section, we simulate the following three models simultaneously in the EI. The three cases are demonstrated as follows:

Analyze and compare the energy net load characteristics and economy of the three operating modes. The load curve of the three cases is depicted in Figure

The net load of the Energy Internet under three cases.

As shown in Figure

It can be observed from Figure

Optimal scheduling results of four ELNs.

The detailed comparison of time-of-use pricing and real-time pricing obtained by the proposed game theoretic model as mentioned above is illustrated in Figure

Comparison of two types of electricity price.

Table

Economic statistics of three cases.

Types | Case I | Case II | Case III |
---|---|---|---|

Loss cost of RES abandoning (yuan) | 2240.5 | 0 | 17.5 |

Transaction cost of power grid (yuan) | 25991.7 | 17172.2 | 4804. 4 |

Charging and discharging loss (yuan) | 0 | 1129.6 | 1173.2 |

Photoelectric subsidy (yuan) | 77532.6 | 81296.6 | 81267.2 |

Operational and maintenance cost | 6268.2 | 6587.6 | 6662.9 |

Gas cost (yuan) | 66357.5 | 73663.3 | 69358.9 |

Multienergy supply income (yuan) | 31453.4 | 31453.4 | 31453.4 |

Power loss (yuan) | 0 | 2233.2 | 2319.5 |

Generation cost of gas turbine (yuan/kWh) | 0.3243 | 0.3889 | 0.3733 |

Total revenue (yuan) | 8128 | 14197.3 | 31703.8 |

Comparing Case II and Case III with Case I, the RES utilization rate has been significantly improved, when the abandoned RES power loss is close to 0. The photoelectric subsidy has increased by 4.85% and 4.82%, respectively. In addition, Case 3 has a slight increase in charging/discharging loss, power loss, and operation and maintenance costs compared with Case II, indicating that the optimization process has little influence on the economy with less transaction cost.

Compared with Case I and Case II, Case III has increased the total revenue by 290.06% and 123.31%, respectively, indicating that the game-theoretical decentralized optimization process can significantly improve the economics of EI. It is obvious that the overall benefit is significantly increased by fully utilizing the output of RES with real-time electricity price mechanism.

Compared with Case I, natural gas costs of Case II and Case III have increased by 11.01% and 4.52%, respectively. The average power generation cost of gas turbines has increased by 19.92% and 15.11%, respectively, indicating that Cases II and Case III can transform heating and cooling power in more effective ways. Comparing Case III with Case II, it can be observed that the increase in natural gas costs is less which indicates that multienergy can actively participate in system optimization to reduce gas turbine output. The utilization of RES has also greatly improved the economic and environmental performance of the EI system.

In this paper, a decentralized energy flow control framework of optimal operation considering the uncertainty of the supply side and demand side has been proposed for the Energy Internet. A typical architecture of ELN is established with system models which can better reflect the characteristics and requirements of EI. In addition, a novel electricity price mechanism for energy interaction is proposed to respond to the supply-demand difference. The theoretical noncooperative game is proposed with the objective to minimize the daily operational cost of the EI system. Through iterative calculation, the game reaches the Nash equilibrium, which is the consensus reached by the MAS. The case study based on real-world data proves the feasibility and effectiveness of the proposed framework. The proposed decentralized framework combining with optimized operational strategy can contribute to reducing the system load volatility and decreasing the operating economic cost as well as improving the reliability and resilience of the EI system.

The data used to support the findings of this study are included within the article.

The authors declare no conflicts of interest.

This work was supported by the National Natural Science Foundation of China under grant 51777193.