Optimal scheduling of blockchain-based virtual power plants considering distribution network congestion

The development of renewable energy has brought new challenges to the power system. Virtual power plant based on blockchain is an effective means to solve the problem of new energy consumption. Solving the internal risk of line overload or voltage over-limit of a virtual power plant can make it play a better role. This paper proposes a scheduling method for a virtual power plant based on blockchain, which can solve the congestion problem. Firstly, the scheduling process of a blockchain-based virtual power plant is designed. Secondly, it is discretized as the basis of the congestion management scheme. Thirdly, the price penalty is introduced into the scheduling process to solve the congestion problem. Finally, an example is designed for simulation. The results show that the method can solve the congestion problem while realizing the virtual power plant scheduling.


Introduction
With the development of renewable energy sources (RES), how to solve the impact of their randomness and volatility on power systems has become an important issue.Virtual power plant (VPP) is an effective method to solve the impact of RES through resource integration and utilization [1].However, the traditional centralized VPP does not make targeted designs for RES, and the scheduling ability is limited [2].With the development of blockchain technology [3], blockchain-based decentralized VPP is expected to solve the problem of RES consumption better.[4].Thus, studying the VPP construction method and scheduling process based on blockchain is significant.
Several studies have explored the potential of blockchain technology in developing VPPs.Reference [5] solves the security problem in VPP and reduces the cost by constructing VPP based on the Ethereum platform.Reference [6] proposes a blockchain-based VPP smart contract, which verifies the method's superiority in trust cost and transaction efficiency.Reference [7] introduces blockchain technology into VPP to achieve more environmentally friendly resource scheduling.Reference [8] proposes a VPP trading mechanism based on blockchain technology for peak-shaving scenarios.The double auction mode is adopted to meet the diversified needs of peak-load buyers.
However, in the research of VPP based on blockchain, there is a lack of discussion on the possible line overload or voltage over-limit in the internal scheduling process of VPP.Therefore, this paper introduces the Dynamic Tariff Subsidy (DTS) [9] into the scheduling process of VPP and uses price 2 penalties to control the electricity consumption behavior to realize congestion management.In the meantime, the paper applies the distributed alternating direction method of multipliers (D-ADMM) to discretize the scheduling method, which provides the price signal for the DTS-based congestion management method.
The paper's structure is as follows: The second part of the paper designs a decentralized VPP based on blockchain.The third part further proposes a congestion management method.The fourth part prepares an example to realize the formulation of the scheduling strategy and the solution to the voltage congestion problem, thus verifying the effectiveness of the scheduling and congestion management methods.

design of a virtual power plant
VPP participants (VPs) include RES generators (such as photovoltaic and wind power), ESS, clusters, and demand response resources (such as air conditioning and electric vehicles).Clusters play an intermediary role in transferring price-related information.
Each VP sends the capacity that can participate in the scheduling to its cluster.The negotiation smart contract receives the price of the cluster broadcast and then negotiates based on the VPP scheduling task to feed back the new price information to the cluster.After several rounds of iterations, the price does not change to determine the demand/supply of each participant.After the electricity is delivered, the token is transferred from the user to the generator through the settlement contract.The scheduling process relies on four types of contracts: RES calculation contract (A), cluster's bidding contract (B), negotiation contract (C), and settlement contract (D).The location and transmission information of each smart contract is shown in Figure 1.

scheduling method based on blockchain
In each scheduling period, the power supply is represented by PGi, and PLi means the power consumed by power users.The scheduling model can be expressed as follows.The RES output as the target of VPP consumption is included in the PGi, which is sent to contract A in Figure 1. .
NL and NG are the number of power users and energy producers, respectively.CG is the cost function of power generation, and CL is the cost function that encourages users to reduce load or participate in consumption.The formula ( 2) is that VPs jointly achieve the task VPPorder of resource consumption or production.
The above equation is discretized by the ADMM method [10] and transformed into a two-layer process.The first layer is the negotiation between the VPs.The second layer is the data interaction between bidding contracts.Let xLi be the amount of electricity required by user i managed by cluster L; xGi is the output provided by the power producer i managed by cluster G. Let PRE=xG1+xG2+…+xGM, fLi(xLi)=CL(xLi), fGi(xGi)=CG(xGi), then the original function can be: NL and NG are the number of load clusters and generation clusters, respectively.PL is the power consumption matrix.
For different VPs, the cost calculation methods that make them participate in scheduling are also different.The processing methods corresponding to specific resource types are not all listed in this paper.Taking the air conditioning demand response resource as an example, the scheduling cost often has a quadratic function relationship with the capacity participating in the scheduling, as shown in (4).The cost of ESS participating in scheduling fE(PE) is not only related to electricity price λ and dispatching cycle T but also depends on the degradation cost fED(PE) of ESS.
d is the number of other bidding contracts negotiated with the bidding contract, and the value is NP−1.
k+1 P is the price adjustment signal, and λ k+1 P is the bidding price.k+1 P is the auxiliary variable.These three are dual variables.When all the bidding price λP values are the same and do not change, the total cost reaches the minimum.At this time, the scheduling behavior is performed, and the token payment is completed through contract D. The above price negotiation process is mainly carried out between contracts B and C in Figure 1.

The mechanism of congestion management
Blockchain-based VPP transmits transaction intention data to distribution system operators (DSO) for verification to perform congestion management.The value of DTS is realized by information iteration between DSO and VPP.DSO provides the information for calculating DTS and sends the initial DTS (all values are 0) to the VPP.The VPP performs optimal distributed energy and load scheduling according to the predicted electricity price and DTS.The VPP reports the load curve to the DSO, and the DSO issues a new DTS, repeating this process until there is no congestion.The above process is shown in Figure 2.

Congestion management on blockchain
The scheduling by VPP should not affect the long-term electricity purchase.Therefore, it is only necessary to examine the impact of the new contract on the predicted state of the power grid.The DTS penalty is superimposed on the node price if the power flow or voltage exceeds the limit.That is, in the scheduling process mentioned above, the price penalty signal is superimposed on the price signal to achieve the purpose of alleviating network congestion.
Without considering distributed resource scheduling, the simplified scheduling objective is as follows.
, is the cost function matrix of the generator.P is the matrix composed of the output Pi of the generator i. P max i and P min i are the upper and lower limits, respectively.M is the power limit of the line.D is the transmission silver submatrix of the power grid.Re(S) is the real part of the node injection power S._ Vmax and Vmin are the upper and lower voltage limits.V0 is the voltage of 0 node.Z is the inverse matrix of partial node admittance matrix YLL._S is the conjugate of S. The node admittance matrix YLL is a sub-matrix of the full admittance matrix Y (remove the relaxed bus-related admittance).
When considering distributed resource scheduling, the new injection power Sk of each node k is: P c and Q c are the active and reactive power of the total distribution network when there is no distributed resource scheduling.Ek is the mapping matrix with each scheduling and node.NB is a set of nodes.New power injection superimposes new costs at each node.The increment of unit price caused by power and voltage over-limit is as follows.This paper uses relative value to represent the amount of price penalty κk.LMPk is the marginal price of node k.
A distributed algorithm is used to solve the unit price increment.When the initial λ + , λ − , − , + =0, there is no over-limit situation.The result is a scheduling strategy that does not consider the grid constraints.Based on the result, the dual variable of the (j+1)th over-limit penalty is obtained by iterative method: Taking the line transmission power limit as an example, if the power exceeds the limit, λ + needs to increase to encourage users to buy less electricity (consume less power).If a node injection contributes more to the over-limit, the penalty is greater.Figure 3 shows the flow chart of the scheduling process considering congestion.The number in the figure is the number of the formula.

Example simulation
The following examples are used to verify the effectiveness of the scheduling method proposed in this paper.The example network is shown in Figure 4. MP1-MP5 are clusters forming a VPP composed of users and generators that access each node.An external power supply is set at node 5; the constant power is 2MW.This simulates the scheduling target that VPP needs to consume.The initial power flow calculation results are shown in Table 1 and Table 2.After the power flow calculation, the following data must be transmitted to the congestion cancellation smart contract: the active power value on each line, voltage amplitude, and active node electricity price.Take V0=1.Table 3 shows the scheduling results in violation of voltage constraints, and the congestion elimination is achieved through 4 iterations.The result of the first iteration is the result of unconstrained scheduling.After the first iteration, the voltage will exceed the limit.The positive number indicates that the node consumes the power supply of the power grid; the negative number means that this node injects electricity into the grid.
Due to the limited space, the price-related data are not all listed.The paper only shows the price penalty parameters of MP1 and MP5 in the iterative process in Table 4. Figure 5 shows the power of nodes participating in scheduling and penalty parameters under different iterations.It can be seen from the figure that as the number of iterations increases, the penalty parameters of nodes 8 and 10 rise, which indicates that the users on these two nodes face higher electricity costs at this time, prompting the user power consumption of these two nodes to decrease.This increases the power injected by MP4 to the node, which prompts the voltage to rise.The node 4 has a significant increase in power consumption, and the node 6 and 7 have a slight increase.This is due to the power balance constraint.Finding the power consumer for the supply of nodes 8 and 10 is necessary to reduce the terminal voltage.Therefore, the power consumption of nodes 4, 6, and 7 will increase.Figure 6 shows the final power consumption of the VPs connected to node 4 and node 10.Multiple color blocks stack each histogram, and each color block represents the power consumption of a VP.It can be seen that the price penalty parameters in Figure 6 have an impact on power consumption.MP5 represents one power producer and four users.It continues to receive positive penalties.The power generation of one internal power producer is rising, and the power consumption of four users is decreasing.MP1 has five users and has not received the penalty, but the power supply in the VPP increases, and the power consumption decreases, so its power consumption is also rising.
The results show that VPP effectively solves the voltage congestion problem through price penalty while completing the scheduling target.

Conclusion
This paper proposes a scheduling method considering congestion for the congestion problem of virtual power plants based on blockchain.By solving the congestion problem in the on-chain VPP, VPP can further play the role of resource aggregation and power consumption in the scenario of high penetration of new energy.Firstly, the VPP scheduling process running on the chain is proposed, and D-ADMM discretizes the scheduling method.Secondly, DTS is introduced into the scheduling process.Based on the discretized price information and price penalty parameters, the iterative operation is performed to obtain the optimal scheduling result without congestion.The example analysis verifies the effectiveness of the proposed method.

Figure 1 .
Figure 1.The interactive procedure of smart contracts.

Figure 2 .
Figure 2. The mechanism of congestion management through DTS.
fp(xp) is used to represent fL(PL) and fG(PG), and the original problem can be expressed as the dual problem.gP(λ) is the objective function of the dual problem.NP is the total number of clusters.AP is the row matrix; for load, all values are 1; for power generation, all values are −1.

Figure 3 .
Figure 3.The scheduling process considering congestion.

Figure 5 .
Figure 5.The power consumption and penalty parameters under different iterations.

Figure 6 .
Figure 6.The power consumption on node 4 and node 10.

Table 1 .
The flow data of the node before the scheduling.

Table 2 .
The Power flow data of the branch before the scheduling.

Table 3 .
Scheduling results of nodes in the iterative process.

Table 4 .
The price penalty parameters of MP1 and MP5.