Study on optimal scheduling of energy storage participation in power market considering source-load uncertainty

With the increasing penetration of new energy sources such as wind and light in the power grid, the selection of new energy-storage joint scheduling and operation methods can reduce the phenomenon of wind and light abandonment, but the new energy-storage joint scheduling and operation methods need to take into account the market factors, characteristics, and other aspects of a variety of problems. Given these difficulties, this paper takes into account the uncertainty of wind power, adopts the Latin hypercube sampling method to generate and consider the probabilistic distance fast reduction method, selects the typical scenarios, establishes the energy storage market operation model by minimizing the comprehensive operation cost, and responds to the source-load demand through the integrated response of the energy storage and the power grid. Finally, the effectiveness of the proposed method is demonstrated through simulation examples.


Introduction
In recent years, with the continuous development of new power systems to wind and other new energy installed capacity continues to rise [1] , the stochastic nature of new energy scenarios for the development of the power system has brought great challenges [2] .With the addition of an energy storage system to realize the decoupling of the source load, grid supply, and demand imbalance, and new energy and energy storage participation in the power market to increase the proportion of energy storage, there are certain problems in joint new energy as a market subject [3][4] .How to effectively solve the problem of wind and light uncertainty and to achieve the new energy-storage joint scheduling and operation based on the results of comprehensive consideration of the energy storage participation in the power market are of great significance.
For new energy-energy storage joint scheduling, Wang et al. comprehensively considered the stability, economy, and other factors, building a source-storage-load multi-objective optimization model, but its energy storage system does not realize the depth of coupling with the power market, resulting in economic maximization can not be achieved [5] .Xu et al. considered wind energy consumption, considering the cost of power generation, the cost of aging batteries, and the rate of wind energy consumption as the objective function of optimal scheduling of the microgrid research, but the system only considered simple microgrid model, did not consider the stochastic nature of wind and other new energy output [6] .Xu et al. proposed an optimization model to minimize the microgrid system operation IOP Publishing doi:10.1088/1742-6596/2771/1/012015 2 and maintenance costs as well as emission costs as the goal, for the system to save costs, but did not consider the grid time-sharing tariffs, and can not be new energy into a schedulable energy to achieve the full use of the economy of energy storage [7] .Wang et al. considered new energy participation in green power trading and established a two-layer optimization model for green power trading power decomposition considering market force inhibition, which provides references for the design of domestic green power trading [8] .To summarize, the current research on new energy-energy storage joint scheduling is mainly aimed at wind power consumption and operating costs and does not consider the uncertainty of new energy and energy storage participation in the power market for the benefits of the system.
Aiming at the stochasticity of wind and light characteristics and the advantages of energy storage in power scheduling and other factors, this paper comprehensively considers the uncertainty of wind and light as well as the storage of joint new energy as the main market.First of all, we use the improved Latin hypercube sampling method to achieve the generation of wind and light scenes and use the probability-based distance method to achieve the scene cuts and solve the problem of wind and light uncertainty.Secondly, we comprehensively consider the coupling relationship between the participation of storage in the power market, the establishment of the source of the load, storage, and the purchase and sale of electricity energy storage market operation model, to set the optimal cost of the system as the goal, to achieve the participation of storage in the power market optimization of the scheduling, and utilize the results to demonstrate this paper's models and the validity of the algorithms.

Modeling of new energy scenario outputs
The wind turbine scenario characteristics satisfy the Weibull distribution and the wind power versus wind speed is shown below [9][10] : where PWT is the wind power output at moment t; Pc is the rated power; Vin, Vout, and Vc are the wind power cut-in, cut-out wind speeds, and rated wind speeds; v is the wind speed, and a and b are the shape and scale parameters of the wind turbine, respectively.The light intensity of the photovoltaic scene satisfies the Beta distribution, and the relationship between its output power and light intensity is [10] : where Ppv is the PV output at moment t; rt is the light intensity at moment t; n is the number of PV modules in the PV system; Șn and Sn are the photovoltaic efficiency and area of the nth PV module, respectively; Į and ȕ are the shape parameters of the PV Beta distribution, respectively; rmax is the maximum light intensity.

Improved latin hypercube scene generation with probability reduction optimization
Latin hypercube sampling (LHS) is a method of approximate random sampling from a multivariate parameter distribution: firstly, we assumed that in the n-dimensional space A n = [0, 1], we take m samples which are divided into several small intervals [(k-1)/n, k/n], where k is the boundary range to take an integer; in several small intervals of random values, we avoid the large-scale data sampling overlap effect; each interval extracted values are then randomly combined to form a random matrix, the ith dimension of the random matrix is denoted by (x1i, x2i, …xmi)', the dimensional matrices are independent of each other, taking yki=(xki -0.5+pki)/m,where pki obeys a uniform distribution in the interval [-0.5, 0.5] and is independent of x, generating the Latin hypercube sample.We let the original set of scenarios be B, and the set of scenarios J whose number of iterations needs to be reduced to be the empty set, and consider the probabilistic distances under the scenarios, calculate the scene distance of any 2 scenes to form the scene distance matrix D[w(Si), w(Sj)] =|| w(Si)-w(Sj)||2, where w(Si) is the uncertain scenario and Si is the scenario number.Then we let the probability of scenario Si occurring be Pi, find the minimum distance probability PDi, where PDi = pi min{D[w(Si), w(Sk)]}, between scene Si and the nearest scene Sk, search for the smallest PDm = min PDi in the current scene, update the set B of scenarios and the set J of planned cut scenarios, and update the scenario probabilities pi = pi+PDm.

Objective function
The objective function to minimize the total cost of operation is given in the following equation: where Ce is the cost of wind power generation; CGrid is the cost of grid interaction; Cbuy and Csell are the cost of grid transaction for buying and selling electricity, respectively; Cbat is the cost of storage response interaction; CGrid, t is the time-of-day tariff for selling electricity to realize the trading of energy storage with the electricity market; char ess,t P and dischar ess,t P are the outgoing and absorbed power of the energy storage at the moment of t, respectively.

Constraints
where PminM, t and PmaxM, t are the upper and lower limits of grid interaction power handling.
3) Energy storage output power constraints  (9)   where SOC(t) is the charge state of the energy storage at moment t, Ș is the charge/discharge efficiency of the energy storage, which is taken as 90%, and SOCmin and SOCmax are the upper and lower limits of the charge state.
5) Energy storage capacity state constraints min 0 max 1 1 where E0 is the initial capacity of the energy storage system; Emax and Emin are the maximum and minimum capacities.

Model solution
In this paper, we study that in the new energy joint energy storage participation in the power market optimization scheduling model, the energy storage output state as well as the energy storage output state of the grid under the power market have the 0-1 variable and continuous variable multiplication term as well as the absolute value term, which leads to the model with complex mixed-integer nonlinear planning characteristics, and in this paper, we realize the nonlinear constraints of the nonlinear model by transforming all the nonlinear models into the linear model through the Big-M method. ^`^0

Example data and scenario generation
To solve the system source-side stochasticity problem, a local new energy-energy storage joint system is selected as the research object, which contains a photo-voltaic system, wind turbine system, energy storage system, and supply loads, and the parameters of the microgrid equipment are shown in Table 1.
The actual measurement of the local new energy system is one day out of the power situation, with a time interval of 1 h, and 1, 000 groups of scenarios of wind and light out of the power situation are generated through the improvement of the Latin superlikeness method.Five typical scenarios are finally obtained by probabilistic cuts and the power situations under the various scenarios, as shown in Figure 1.
As can be seen from Figure 1, five typical scenarios are obtained after Latin hyper-cube generation and probabilistic cuts, among which scenario 3 has the largest probability in the wind power scenario, and scenario 1 has the largest probability in the photovoltaic scenario.Therefore, this paper selects these two sets of data as the wind scenario out of the wind and validates the scheduling impacts brought by uncertainty scenarios.

Optimized scheduling results
To verify the effectiveness of the optimal scheduling model of energy storage participation in the power market in this paper, we select the highest probability scenario in 3.1, with the difference of source-load certainty as well as source-load uncertainty, respectively.Energy storage is used for system regulation and not use energy storage as the system optimization objective.
Option 1: the optimal scheduling only considers the new energy uncertainty; Option 2: the optimal scheduling considers only the participation of the storage in Scenario 1: the optimal scheduling only considers the new energy uncertainty; Option 3: the optimal scheduling considers both the new energy uncertainty and energy storage participation.The operating costs under the three scenarios are shown in Table 1, The results of the three scenario runs are shown in Figure 2.
Table 1.Costs of individual programs.
Operating cost/yuan Option 1 1849 Option 2 1250 Option 3 1240 Combined comparison of Table 1 and Figure 2 reveals that: option 1 only considers wind and light uncertainty and does not consider the optimization effect of the system after energy storage is involved in scheduling, which affects the system's wind and light consumption rate as well as the energy quality issue, and the system can only meet the supply by buying power from the grid and also needs to buy power from the grid when the feed-in tariffs are higher, which increases the system's operating costs.Option 2 only considers the optimization effect on the system after energy storage is involved in scheduling, and does not consider the wind and light uncertainty, which leads to the inaccuracy of energy storage scheduling precision, the increase of wind and light consumption rate, the reduction of the cost of purchasing electricity from the grid, and the uncertainty of wind and light, which leads to the storage power not being able to be better coupled with grid power, resulting in the existence of the problem of abandoning the wind and light.Option 3 proposes the comprehensive consideration of the participation of energy storage in scheduling as well as wind and light uncertainty, increasing the consumption rate and reducing the cost of buying electricity from the grid will lead to the existence of abandoned wind IOP Publishing doi:10.1088/1742-6596/2771/1/0120156 and light.With consumption rate increases, the cost of purchasing power from the grid decreases.It shows that the cost of energy can be reduced through the energy storage output when the time-sharing tariffs are high, which not only ensures the supply of loads at the same time but also ensures that the storage energy participates in the power market, ensuring that the system economy is optimal.

Conclusion
Aiming at the stochasticity of new energy and the uncertainty of market factors in the joint scheduling operation of new energy and energy storage, this paper constructs a new energy scenario generation model, considers the combination of energy storage and power market to realize the system equipment scheduling, verifies the validity of the method through the examples, and obtains the following conclusions: 1) Based on the locally measured data, the new energy uncertainty scenario is generated based on the Latin superlattice method, which takes into account the stochastic nature of the wind power as well as the temporal sequence, can better characterize the uncertainty of the new energy power wind, and is close to the actual application scenarios.
2) This paper considers the relationship between power operation costs and time-sharing tariffs, which can improve the utilization rate of energy; at the same time, it takes into account the economics of purchasing and selling electricity.Compared with conventional energy storage scheduling and operation methods, it can improve the rate of new energy consumption such as wind and light, promote the local consumption of new energy in the microgrid, and reduce the cost of new energy system operation.
bch and bdis are storage charge/discharge power mutual exclusion coefficients.After transforming the nonlinear model into a linear model by solving in Matlab 2021, we use the Gurobi toolbox of commercial solvers.

Figure 2
Figure 2 Diagram of the Operating Status of Each System under the Three Options