CN108009693B

CN108009693B - Grid-connected micro-grid double-layer optimization method based on two-stage demand response

Info

Publication number: CN108009693B
Application number: CN201810004691.0A
Authority: CN
Inventors: 米阳; 李战强; 韩云昊; 何星瑭; 宋根新; 李振坤; 符杨; 苏向敬
Original assignee: Shanghai University of Electric Power
Current assignee: Shanghai University of Electric Power
Priority date: 2018-01-03
Filing date: 2018-01-03
Publication date: 2021-09-07
Anticipated expiration: 2038-01-03
Also published as: CN108009693A

Abstract

The invention relates to a grid-connected micro-grid double-layer optimization method based on two-stage demand response, wherein a demand response layer guides a user to actively track the power generation output of clean energy through primary optimization, promotes the efficient development and utilization of the clean energy, improves the proportion of the clean energy in terminal energy consumption, forms a more stable power load running state through the adjustment of demand load through secondary optimization, and improves the output stability of a controllable micro-source. The microgrid layer minimizes the comprehensive operation cost from the viewpoint of the economy of the microgrid on the basis of the microgrid layer. A reasonable grid-connected microgrid double-layer optimization model is constructed by fully considering the demand side response based on time-of-use electricity price, and the model is solved by adopting the proposed genetic algorithm based on simulated annealing, so that the convergence speed and precision of model solution are improved.

Description

Grid-connected micro-grid double-layer optimization method based on two-stage demand response

Technical Field

The invention relates to a microgrid economic optimization scheduling technology, in particular to a grid-connected microgrid double-layer optimization method based on two-stage demand response.

Background

With the increasingly prominent environmental and energy problems, distributed energy becomes a hotspot of research in the industry due to the characteristics of environmental friendliness, renewability and the like. However, when the direct Grid connection is adopted, the fluctuation and the intermittence of the direct Grid connection bring the power quality and the Operation stability and other problems to the power Grid (for example, document 1: in the establishment, late welfare construction, Xuke, and the like, the influence of the Distributed power supply access on the power Grid is analyzed [ J ]. the power system and the automatic chemical report thereof, 2012, 24(1): 138 + 141; document 2: Ahn S J, Nam S R, Choi J H, et al. As an effective organization form of a distributed Power supply connected to a Power distribution network, a micro-grid becomes one of key technologies for solving the problem (for example, document 3: Hatziargyrio N, Asano H, Iravani R, et al. Microgrids [ J ]. IEEE Power and Energy Magazine, 2007, 5 (4): 78-94; document 4: Katiraiei F, Iravani R, Hatziargyio N, et al. Microgridsmeage [ J ]. IEEE Power and Energy Magazine, 2008, 6 (3): 54-65).

The micro-grid is a small power system formed by a series of distributed power supplies, and the economic operation and the optimal scheduling of the micro-grid are important subjects in micro-grid research. The goal of optimal scheduling of the microgrid is generally divided into construction cost, operation and maintenance cost, environmental benefit cost, reliability, network loss and the like. In document 5 (zhulan, strictly, yang, etc.. wind-solar energy storage microgrid system storage battery capacity optimization configuration method research [ J ]. power grid technology, 2012, 36 (12): 26-31), a wind-solar energy storage complementary power generation system model is established, the storage battery is charged by utilizing sufficient wind energy at night and low-load time periods, and the solar energy, the wind energy and the energy storage equipment are utilized to generate power together to meet the requirement of power supply in the daytime and high-load time periods. Document 6 (wangchengshan, hong bovin, guo power, etc. general modeling method for optimization scheduling of combined cooling heating and power microgrid [ J ]. the report of electrical engineering in china, 2013, 33 (31): 26-33) proposes a bus-type structure based on combined cooling heating and power and designs a corresponding dynamic economic scheduling optimization model, and solves the model by adopting 0-1 mixed integer linear programming, thereby verifying the feasibility and effectiveness of the structure. Documents 7 to 9 (document 7: Chengjie, Yangxu, Zhulan, etc.. microgrid multi-target economic dispatch optimization [ J ]. China Motor engineering Proc, 2013 (19): 57 to 66; document 8: Penchunhua, Xipeng, Zhanwen, etc.. microgrid clubfenbane [ J ]. grid technology based on the improved bacterial foraging algorithm, 2014,38 (9): 2392 2398; document 9: Wuxiong, Wanglali, Wangjiangji, etc.. hybrid integer programming method of microgrid economic dispatch problem [ J ]. China Motor engineering Proc, 2013, 33 (28): 1 to 8) takes a microgrid comprising renewable energy, electricity storage, cogeneration system, etc. as a research object, and analyzes the optimal output of each unit under consideration of the operation of depreciation, maintenance, cogeneration fuel cost, etc. in the microgrid.

Because the optimization problem is a typical NP-hard problem and can simultaneously meet the constraint conditions of a plurality of indexes in the optimization process, the optimization method is suitable for solving the optimization problem by adopting an artificial intelligence algorithm (such as a document 10: Guo Li, Liu Jian, Qi-focused, and the like; a multi-target optimization planning design method of an independent micro-grid system [ J ]. Chinese Motor engineering report, 2014,34 (4): 524-536), and has excellent effect. Document 11 (liu xiao ping, dingming, zhang yue, etc. dynamic economic dispatch of microgrid system [ J ]. china motor engineering reports, 2011,31 (31): 77-84) proposes a microgrid system dynamic economic dispatch model based on opportunity constraint planning, and solves the model by using a genetic algorithm combined with monte carlo simulation. In document 12 (bear flame, Wujiekang, Wangqiang, etc.. Combined cooling, heating and power optimization coordination model and solving method [ J ]. Chinese Motor engineering report 2015, 35 (14): 3616 and 3625) combined with different rate structures, a combined cooling, heating and power optimization coordination model for wind-solar storage and natural gas complementary power generation is established, and the model is solved by adopting an NR-PSO algorithm, so that the function and the advantage of multi-source complementary are displayed. Document 13(Morais H, Kadar P, Faria P, et al. optimal scheduling of a Renewable micro-grid in an isolated load area using an integrated linear programming [ J ]. Renewable Energy, 2010, 35 (1): 151-156) proposes an Energy microgrid optimized mixed integer linear programming model, which is solved by using a CPLEX branch-and-bound algorithm, thereby obtaining a good precision while ensuring the computation time. In document 14 (poppy, aixin, treble and courage, etc.. micro-grid economic operation analysis containing multiple energy supply systems based on a particle swarm optimization algorithm [ J ]. a power grid technology, 2009, 33 (20): 38-42), a micro-grid economic model considering emission of greenhouse gases and pollutants is established according to steady-state characteristics of a distributed power supply by taking the lowest micro-grid operation cost as an objective function from 2 aspects of cost and benefit, and the model is solved by using the particle swarm optimization algorithm. Document 15 (zhao wave, kakan, xu zhi, etc.. optimization configuration of optical storage grid-connected microgrid considering demand-side response [ J ]. report of electrical engineering in china 2015, 35 (21): 5465-. 16-17 (16: Yuantiejiang, chaulmoogra, Tuersian Yibayno, and the like. modeling of dynamic cleaning economic optimization scheduling of large-scale wind power grid-connected power system [ J ]. Chinese Motor engineering newspaper 2010, 30 (31): 7-13; 17: expensive in terms of old money, rich in terms of old money, modeling and algorithm of environmental economic dynamic scheduling of wind power system [ J ]. Chinese Motor engineering newspaper 2013, 33 (10): 27-35) models dynamic optimization scheduling of wind power system, and solving the established scheduling model by respectively adopting genetic algorithm and improved quantum-particle swarm optimization algorithm.

Demand side response (DR) is a process in which an electric power company guides a user to change a consumption mode by an incentive mode such as discount of electricity price or high price compensation (for example, 18: zhulan, rigor, yang xiu, and the like, a microgrid comprehensive resource planning method considering demand side response [ J ], chinese motor engineering bulletin, 2014,34(16):2621 and 2628), so that supply and demand interaction is realized, a load rate is improved, and power balance is achieved. Demand-side responses generally include Direct Load Control (document 19: Kurucz C N, Brandt D, Sim S. A Linear programming Model for reducing system peak through customer Load Control programs [ J ]. IEEE Transactions on Power Systems, 1996, 11 (4): 1817 1824; document 20: Ruiz N, Cobelo I, Oyarzabal J.A Direct Load Control Model for Virtual Power Plant Management [ J ]. IEEE Transactions on Power Systems,2009,24(2):959-, Demand side bidding, interruptible load and emergency demand response, etc. Document 23 (Dian bridge column, King Lingyun. demand side response benefit estimation [ J ]. Chinese Motor engineering report, 2014,34(7): 1198-. The day-ahead scheduling plan model comprising an uncertain response mechanism of time-of-use electricity price and an interruptible load mechanism is constructed in a document 24 (Sun Yujun, Liyang, Wangbi, and the like, a day-ahead scheduling plan model [ J ] of power grid technology, 2014,38(10): 2708-.

Disclosure of Invention

The invention provides a grid-connected microgrid double-layer optimization method based on two-stage demand response, aiming at the problem that economic operation and optimal scheduling are important subjects in microgrid research, a reasonable grid-connected microgrid double-layer optimization model is built by fully considering demand side response based on time-of-use electricity price, the model is solved by adopting the proposed genetic algorithm based on simulated annealing, the operation cost of a microgrid is minimized, direct load control is mainly considered, and when the microgrid is scheduled, the corresponding load is directly removed.

The technical scheme of the invention is as follows: a grid-connected microgrid double-layer optimization method based on two-stage demand response specifically comprises the following steps:

1) inputting all parameters and constraint conditions of the model according to each micro-source model of the micro-grid system, and selecting an annealing temperature T and a maximum iteration number G_maxAnd a population size M;

2) initializing a first-level optimized population: selecting clean energy as primary optimization energy, and collecting the primary optimization energyThe source data is used as an optimization individual, and the data information contained in each individual is optimized in a first-level mode: electric power P output by wind turbine in same time period_wt,rOutput power P of photovoltaic module_pv,rLoad shift amount P in first-level optimization_Lin,1And the amount of discharge P_Lout,1；

3) Bringing each individual generated in step 2) into an objective function

Calculating the corresponding objective function value, wherein the individual with the minimum objective function value is the optimal, the individual adaptability value is the most optimal, and in order to obtain consistent solving results, the reciprocal of the objective function value is taken as the adaptability value of each individual;

4) sorting the fitness values of the individuals in the population in the step 2) from large to small by adopting a tournament selection pair, reserving the optimal individuals, then carrying out cross variation operation on other individuals to obtain new individuals corresponding to the optimal individuals, and forming a new population with the reserved optimal individuals;

5) then judging whether the number G of the previous iteration is larger than the maximum iteration number G_maxE.g. G is less than G_maxG +1, T is T tau, tau is a decreasing factor of annealing temperature, and the new population formed in the step 4) enters the step 3) to perform the next iteration cycle until G is greater than or equal to G_maxOutput the optimal individual P_wt,r、P_pv,rAnd P_L,1Substituting the determined value into secondary optimization and final microgrid layer optimization, and switching to secondary optimization;

6) initializing a second-level optimized population: resetting iteration times G and annealing temperature T, selecting energy with stable output and adjustable output as secondary optimized energy, collecting secondary optimized energy data as optimized individuals, obtaining an optimal output curve by adopting a similar fitting method according to historical operating data of each secondary optimized energy without considering a load state, setting a total output of a scheduling department after further correcting the optimal output curve according to the operating condition of the secondary optimized energy and a daily equipment maintenance plan as an expected supply curve, wherein each secondary optimized individual comprises a primary optimal individual P_wt,r、P_pv,rAnd P_L,1Simultaneously, performing secondary optimization on the output and expected supply curves of all energy sources;

7) bringing each individual generated in step 6) into an objective function

Calculating the corresponding objective function value, wherein

For the equivalent load of the microgrid system after the second-level optimization,

the method is expected to be supplied for the microgrid system, the individual with the minimum objective function value is the optimal individual, the maximum individual fitness value is the optimal individual, and in order to achieve the consistency of the solving results, the reciprocal of the objective function value is taken as the fitness value of each individual;

8) sorting the fitness values of the individuals in the population in the step 6) from large to small by adopting tournament selection, reserving the optimal individuals, then carrying out cross variation operation on other individuals to obtain new individuals corresponding to the optimal individuals, and forming a new population with the reserved optimal individuals;

9) then judging whether the number G of the previous iteration is larger than the maximum iteration number G_maxE.g. G is less than G_maxG +1, T is T tau, tau is a decreasing factor of the annealing temperature, and the new population formed in the step 8) enters the step 7) to carry out the next iteration cycle until G is larger than or equal to G_maxAnd outputting the total energy output and load transfer P in the optimal individual microgrid system_clAnd calculating the real-time electricity price rate P corresponding to the optimal solution^*；

10) And finally, determining the economic dispatching distribution of the output of each controllable secondary optimization micro-source according to the economic objective function of the micro-grid system and the corresponding constraint condition.

The invention has the beneficial effects that: according to the grid-connected micro-grid double-layer optimization method based on two-stage demand response, the demand response layer guides users to actively track the power generation output of clean energy through first-stage optimization, efficient development and utilization of the clean energy are promoted, the proportion of the clean energy in terminal energy consumption is improved, a more stable power load operation state is formed through adjustment of demand load through second-stage optimization, and the output stability of a controllable micro-source is improved. The microgrid layer minimizes the comprehensive operation cost from the viewpoint of the economy of the microgrid on the basis of the microgrid layer. In addition, the simulation annealing-based genetic algorithm is provided for solving the model, so that the convergence speed and precision of model solving are improved.

Drawings

FIG. 1 is a flow chart of the model solution of the present invention;

FIG. 2 is a graph of wind power, photovoltaic prediction data and optimization data according to the present invention;

FIG. 3 is a diagram of the results of various levels of load optimization according to the present invention;

FIG. 4 is a graph of the optimization results of the output of each distributed energy source;

FIG. 5 is a comparison graph of the algorithm solution results of the present invention;

FIG. 6 is a general clean energy optimization chart of the present invention;

FIG. 7 is a general optimization force diagram of the micro-combustion engine of the present invention;

fig. 8 is a load optimization graph of the present invention.

Detailed Description

The invention is further explained in the following aspects of a microgrid layer optimization model, a demand response layer optimization model, a model solving algorithm, validity verification and the like.

1. Optimization model of micro-net layer

In order to minimize the comprehensive cost, a comprehensive objective function comprising system investment depreciation cost, operation and maintenance cost, environmental benefit cost and electricity purchasing and selling cost is constructed for a grid-connected microgrid comprising a micro-gas turbine, wind-solar energy storage and demand side load.

1.1 objective function

The objective function for calibrating the comprehensive cost of the system is

Wherein F is the system day operationComprehensive cost;

depreciation cost for investment;

for operating and maintenance costs;

cost for environmental benefit;

and the electricity purchasing and selling cost of the system interacting with the power of the power grid.

1) Depreciation cost

The depreciation cost is usually obtained by converting the investment cost into the unit capacity cost by adopting an equal-annual-value method, belongs to a fixed component of the power generation cost of a power plant (such as a document 25: Wulian, profit distribution [ J ] based on kernel theory in Aiqing and large-scale multi-source combined delivery coordinated dispatching, a power grid technology, 2016,40(10):2975 and 2981.), and is closely related to the dispatching cost and the benefit distribution of the power plant, so that the depreciation cost is considered. the depreciation cost of the micro-source in the t period is (such as documents 26: Massaeli M, Javadian S A M, Khalesi N. environmental aspects of DGs and comparing the generation costs with thermal power generation requirements and human health [ J ].2011,15(2):1-6.)

In the formula, n_sThe number of the micro sources; r is_iFixed annual interest rate for the ith micro source; n is_iTaking the value of the average service life of the equipment for the investment repayment period; c. C_in,iIs the construction cost per unit volume; k is a radical of_iIs the annual utilization coefficient;

the output power of the micro source at the time t. In the formula r_i、n_iAnd k_iIs selected fromMethod reference [26]。

2) Operating and maintenance costs

The operation and maintenance cost comprises fuel cost and maintenance cost, and the model is

In the formula, c_o&m,iAnd the coefficient is the operation and maintenance cost coefficient of the ith micro-source unit output.

3) Cost of environmental benefit

Because the distributed micro-sources such as the micro-gas turbine and the like use non-renewable energy sources such as natural gas and the like as fuels to cause certain energy consumption and environmental pollution, the environmental benefit cost is considered, and the cost mainly comprises two aspects of cost of environmental value loss and fine cost of pollution discharge (such as reference 27: Ma xi Yuan, Wu Guanghe, Fang Hua, and the like, and wind/light/storage hybrid micro-grid power supply optimization configuration adopting an improved bacterial foraging algorithm [ J ] China Motor engineering Proc, 2011,31(25):17-25.), and the model is as follows:

in the formula, n_pIs a type of contaminant; q_ijThe discharge amount of j pollutants is g/(kW.h) when the ith distributed power supply outputs unit electric quantity;

the environmental value of the j-type pollutants; v_jThe penalty factor for class j contaminants.

The environmental benefits of the electricity purchasing part (thermal power) are taken into consideration, and the pollution emission data of each power generation technology is detailed in documents [28-29] (document 28: zhanxiahui, yangbao, chongqing, and the like. a power supply planning model considering environmental cost and demand side management items [ J ]. power grid technology 2015,39(10):2809 + 2814; document 29: Qian kejun, Yuan, Shixiandan, and the like. environmental benefit analysis of distributed power generation [ J ]. Chinese electro-mechanical engineering report, 2008,28(29):11-15.), and the environmental value standards and penalty orders of magnitude of each pollutant are shown in Table 1.

TABLE 1

4) Cost of electricity purchase and sale

The cost of electricity purchase and sale can be expressed as

In the formula (I), the compound is shown in the specification,

the power of the power purchased from the main network by the microgrid at the time t is shown, the power sold to the main network is shown when the value is negative, and the power purchased is otherwise;

representing the real-time electricity prices of the grid.

In addition, the transferable load in demand-side response is an autonomous response of the user to reduce own electricity consumption cost under the incentive of real-time electricity prices without compensation, so that DR compensation cost is not counted.

1.2 constraint conditions

1) Active power balance constraint

In the formula (I), the compound is shown in the specification,

respectively outputting the micro-gas turbine, the stored energy, the photovoltaic and the fan at each moment in real time;

the predicted load and the net load transfer at time t are respectively.

2) Junctor capacity constraint

In the formula (I), the compound is shown in the specification,

the upper and lower power limits of the tie line.

3) Micro-combustion engine output constraint and climbing constraint

In the formula (I), the compound is shown in the specification,

respectively representing the minimum and maximum output which can be achieved by the micro-combustion engine at the time t;

respectively the minimum rated power and the maximum rated power of the micro-combustion engine; delta_down、δ_upRespectively, the climbing rate and the descending rate.

4) Energy storage system restraint

During charging and discharging, the remaining capacity of an Energy Storage System (ESS) is usually characterized by a state of charge (SOC), and a model of the SOC of a battery and its constraints can be expressed as

SOC^min≤SOC^t≤SOC^max (13)

In the formula, epsilon is the self-discharge rate of the storage battery; Δ t is the time interval;

which indicates the power of charge and discharge,

the state is a discharging state, otherwise, the state is a charging state; alpha is the charge-discharge efficiency; v_ESSIs the total capacity of the storage battery.

Respectively representing the maximum and minimum charging and discharging power of the energy storage system; SOC^max、SOC^minRespectively, the upper and lower limits of the SOC.

5) Fan and photovoltaic output constraints

In the formula (I), the compound is shown in the specification,

wind curtailment and light curtailment power;

respectively for the predicted output of wind power and photovoltaic.

2-demand response layer optimization model

Although the traditional single-layer optimization model can effectively reduce the operation cost of the microgrid, the traditional single-layer optimization model is difficult to optimize and coordinate relevant indexes such as clean energy utilization rate, peak-valley difference on demand side, controllable micro-source output stationarity and the like, and even the indexes are used as the cost due to economy. The real-time matching degree of the load and the system output is improved by further optimizing the system by only adjusting the load of the demand side, the effect is improved, but the adjusting capability is limited and the demand influence on the demand side user is large. Therefore, a demand response layer optimization scheme based on source-load interaction is provided, and various indexes are coordinated and optimized while the operation cost of the microgrid is effectively reduced.

2.1 first order optimization

The investment cost of Distributed Renewable Energy (DRE) such as wind and light is high, the scheduling priority is difficult to obtain in the traditional economic scheduling model, the development and utilization of the DRE are not facilitated, and the strategic adjustment of the energy consumption structure in China is seriously hindered. In order to guide a user to actively track the generated output of the clean energy by using electricity, promote efficient development and utilization of the clean energy and improve the proportion of the clean energy in terminal energy consumption (such as a document 30: singing, Yang Yong Qi, Liu dun nan, and the like, an energy internet 'source-network-load-storage' coordination optimization operation mode and a key technology [ J ]. a power grid technology, 2016,40(1): 114-. The model is

In the formula (I), the compound is shown in the specification,

the equivalent load after the first-level optimization;

optimizing load net transfer capacity for a first level;

load amount transferred in and load amount transferred out in the first-level optimization respectively; n is a radical of_kIndicating the kind of transferable load,

is the number of cells shifted from time j to time t; p_kIs the unit transfer amount of the kth class load.

2.2 two-stage optimization

The conventional distributed controllable micro sources such as the micro gas turbine and the energy storage have the advantages of stable and adjustable output, but the large-range fluctuation of the load not only increases the spare capacity of the system, but also can lead the micro gas turbine to be in low-efficiency operation when the output is less or limit operation when the output is excessive, even to frequently start and stop the machine, and is difficult to ensure the economical efficiency of the operation of the unit and the safety of a power grid. Therefore, in order to form a more stable power load operation state and improve the output stability of the controllable micro-source, a DR secondary optimization scheme is provided.

And (3) obtaining an optimal output curve by adopting a similar fitting method according to historical operating data of each conventional unit without considering the load state, setting the total output of the unit after further correction as an expected supply curve by a dispatching department according to the unit operating condition and an equipment daily inspection and repair plan, and optimizing the load according to the curve by a secondary optimization scheme so as to expect that the load variation trend is adaptive to the expected supply curve, thereby improving the output stability of the conventional power supply.

Defining an expected supply curve as

The second-order optimization model is

In the formula (I), the compound is shown in the specification,

the equivalent load of the whole microgrid system after the second-stage optimization;

optimizing the net output of the energy load for the second level;

the load amount of the transfer-out and the transfer-in of the secondary optimized energy respectively can refer to primary optimized energy reference formulas (19) to (21).

The secondary optimization focuses on improving the output stability of the conventional power supply, although the load curve is only adjusted, the total output of the conventional controllable micro-sources of the micro-grid layer is indirectly adjusted, and the distribution of the load power among the micro-sources depends on the economic optimization scheduling of the micro-grid layer.

the maximum transfer capacity constraint in each stage optimization at the time t is

In the formula (I), the compound is shown in the specification,

the maximum transfer-in capacity and the maximum transfer-out capacity of the transferable load are respectively;

the total net load transfer is optimized for the demand side.

In the two stages, the user is stimulated to adjust the transferable load through real-time electricity price, and the user is guided to adjust the electricity utilization plan, so that the electricity safety and economic production are facilitated. The demand price elasticity reflects the sensitivity degree of the load demand to price change, and the expression is

In the formula, e is the demand price elasticity; l is_bAnd Δ L are the total load before optimization and the load variation after optimization, respectively; p is a radical of_bAnd Δ p are the adjustment range of the electricity prices before adjustment and the electricity prices, respectively.

The relationship between electricity price and demand side load can be expressed as (for example, document 31: P. Thimmamprami, J. Kim, A. Botterud, and Y. Nam, "Modeling and relationship of price elasticity of demand use agent-based model," in Proc. IEEE Innovative Smart Grid Technologies (ISGT),2010, pp.1-8.)

L＝ap^e (31)

Wherein a is a constant; l is the load capacity; and p is the optimized real-time electricity price. The load and the real-time electricity price are in a per unit value form on the basis of literature research such as [ 7-9, 31 ]:

L^*＝L_a/L_b (32)

p^*＝p/p_ref (34)

in the formula, L^*、p^*Electricity rate which is the desired load rate and the real-time electricity rate, respectively; l is_a、L_bRepresenting the expected load after optimization and the load amount before optimization at a certain time; p is a radical of_refIs the conventional electricity price (is a fixed value).

Although equations (25) to (29) indirectly constrain the load transfer rate and the real-time electricity rate by constraining the load transfer amount, constraint p satisfies:

p^min≤p≤p^max (35)

in the formula, p^max、p^minThe real-time electricity price is constrained by the upper limit and the lower limit.

Since DR is the subjective intention of users, uncertainty of DR has certain influence on power grid dispatching (such as documents 32: Rochun Jiang, Liyao Yao, Schuohanping, and the like)]Power system automation, 2017(5):22-29.), thus introducing an uncertainty factor R_DThe desired load factor is corrected. Corrected load factor

Is composed of

And the dispatching department formulates reasonable electricity price according to the expected load rate and by combining the relation between the real-time electricity price and the response of the demand side, stimulates the user to change the consumption mode, and achieves the expected effect of adjusting the transferable load time sequence.

3 model solving algorithm

GA (genetic algorithm) is a parallel random search intelligent optimization algorithm formed by simulating a natural genetic mechanism and a biological evolution theory, does not need knowledge of a search space or other auxiliary information, and iterates from a string set to seek a global optimal solution. However, the phenomenon that the local search ability of GA is weak and early maturing is easy to occur is a problem which is difficult to deal with for a long time.

SA (simulated annealing algorithm) simulates the thermodynamic process of high-temperature metal cooling, and Metropolis criterion is adopted as a search strategy, so that a new high-quality solution is unconditionally accepted, and meanwhile, a poor new individual is selected with a certain probability, so that the situation that the local optimum is trapped is avoided, and the method has strong local search capability. Metropolis criterion is

In the formula, p_rA selection probability for j individuals; x is the number of_jFor the current decision variable x_iThe new decision variables generated; f (x)_j)、f(x_i) Respectively corresponding fitness values; k is a Boltzmann constant; t is the annealing temperature.

Will be referred to herein as f (x)_i) Set to a constant of 0, and since the three objective functions of the scheme herein are all non-negative, the combination (37) can be derived:

from the formula (38): 1) at the same temperature, the larger the fitness value, the smaller the probability that the individual is selected; 2) iteration is initial, T is very high, the probability difference of each feasible solution is very small, and SA can be regarded as performing wide area search to avoid trapping in local optimum; 3) in the later period of iteration, the temperature is very low, and the fitness value of each feasible solution is even very smallThe difference will also rapidly expand the p of the good individuals and the poor individuals_rThe SA can be regarded as performing a local search to refine the optimal solution. And T controls the solving process to be carried out towards the optimal value.

Therefore, the GA parallel search provides the largest possible search space for the SA, the SA effectively improves the local search capability of the GA, and the high-efficiency and strong-robustness Simulated Annealing Genetic Algorithm (SAGA) can be obtained by combining the two.

When the model is solved, all the constraints are added into the corresponding objective function in a penalty function mode and converted into an unconstrained optimization model. The flow of the algorithm to solve the model is shown in FIG. 1, wherein G, G_maxRespectively representing the current iteration times and the maximum iteration times; m represents the population size; τ represents the annealing temperature decreasing factor. The crossover and mutation probabilities depend on the SA algorithm (38), but because the mutation probability is small, the Boltzmann constant k corresponding to the crossover and mutation probabilities is smaller than that in the crossover probability.

The concrete solving steps are as follows:

a, according to each micro-source model of the micro-grid system, inputting all parameters and constraint conditions of the model, selecting an annealing temperature T and a maximum iteration number G_maxAnd a population size M;

initializing a first-stage optimization population (one feasible solution of each objective function is an individual of the objective function, all the individuals form the population, and the optimization finds an optimal solution in a plurality of feasible solutions), and performing first-stage optimization on data information contained in each individual: electric power P output by wind turbine in same time period_wt,rOutput power P of photovoltaic module_pv,rLoad shift amount P in first-level optimization_Lin,1And the amount of discharge P_Lout,1；

C, bringing each individual generated in the step B into a formula (16) to calculate a corresponding objective function value, wherein the individual with the minimum objective function value is the optimal individual, and the individual adaptability value is the most optimal individual, so that in order to obtain consistent solving results, the reciprocal of the objective function value is taken as the adaptability value of each individual;

d: sorting individual fitness values in the population from large to small by adopting tournament selection, reserving optimal individuals, then carrying out cross variation operation on other individuals to obtain new individuals corresponding to the optimal individuals, and forming a new population with the reserved optimal individuals;

e, judging whether the number of previous iterations G is greater than the maximum number of iterations G_maxE.g. G is less than G_maxAnd G is G +1, T is T tau, tau is a decreasing factor of the annealing temperature, and the new population formed in the step D is put into the step C to carry out the next iteration cycle until G is more than or equal to G_maxOutput the optimal individual (where P_wt,r、P_pv,rAnd P_L,1Namely, the optimal individual determination value is substituted into the secondary optimization and the final microgrid layer optimization), and the secondary optimization is switched to. F: similarly, the solving steps of the second-level optimization and the microgrid layer optimization are similar to the steps B, C, D and E of the first-level optimization, except that the information contained in each individual in the initialized population is different, and the information contained in each individual of the second-level optimization and the microgrid layer optimization is respectively P_Lin,2、P_Lout,2And P_ess、P_grid、P_μgThe output results are respectively P_cl、P^*And P_ess、P_grid、P_μg。

As can be seen in FIG. 1, the load is primarily adjusted by the demand response mechanism indicated by the first-level optimization to actively track the clean energy to generate power, and the optimized P is used_wt,r、P_pv,rAnd P_L,1The parameters are stored in the structure body and used as parameters for secondary optimization calling; similarly, the secondary optimization improves the load and the total expected output of the controllable micro-source by further adjusting the load on the demand side

Determining the final total load transfer amount P_clAnd calculating the electricity rate P therefrom^*Same level of optimization P_wt,r、P_pv,rThe results are stored in the structure body together for economic dispatching optimization of the micro-grid layer; finally, the economic dispatching distribution of the output of each controllable micro source such as a micro combustion engine and energy storage is determined by the micro network layer according to an economic objective function and corresponding constraint conditions (the most important secondary optimization is)The total load transfer quantity of demand response is obtained and combined with P of first-level optimization_wt,r、P_pv,rAnd obtaining the sum of the output of all the traditional energy power supplies, wherein the microgrid layer definitely distributes the total power required to be provided by the traditional energy to each traditional power supply according to economic dispatching). The red solid lines with arrows in fig. 1 indicate the optimized information transfer situation of each stage.

4 example analysis

1) Model parameters and data

In the calculation example, a micro-grid-connected system formed by wind, light, gas and storage is adopted, and the micro-grid-connected system comprises 1 wind power plant, 1 photovoltaic power plant and 1 energy storage power plant, 2 gas turbine units and parameters of various devices shown in a table 2. And a real-time electricity price made according to the load rate is adopted in the electricity purchasing and selling cost model, and the time scale is 15 min.

TABLE 2

4.2 analysis of results

Based on the parameters, firstly, the improved SAGA algorithm is adopted to carry out solving analysis on the model, and the feasibility and the effectiveness of each level/layer of optimization target are verified. And secondly, comparing the four different schemes respectively to verify the coordination optimization effect of the scheme on each index.

4.2.1 model solution

And the demand side response is optimized in two stages, the result of the first-stage optimization is used as the input of the second-stage optimization, and the result of the second-stage optimization is used as the input of the microgrid layer optimization. The primary optimization carries out time sequence adjustment on the transferable load according to the predicted total output of clean energy such as wind, light and the like, and determines the actual reference values of the output of the fan and the photovoltaic. The results of the SAGA algorithm for fan and photovoltaic output optimization are shown in FIG. 2. The load optimization of each level of the demand response layer is shown in figure 3.

As can be seen from the wind-solar power output characteristics shown in fig. 2 and 3 and the first-stage optimization result at the demand side, the first-stage optimization can improve the utilization rate of clean energy in the low load period and reduce the load transfer amount of the clean energy in the high peak period by changing the source-load interaction mode of the demand curve and the clean energy supply curve.

In addition, as can be seen from fig. 3, the second-level optimization is smoother than the first-level optimization load curve, which indicates that the second-level optimization further optimizes DR to improve the output stability of the controllable micro-source.

The real-time electricity prices (integration time) obtained from the expected load rates determined from the initial load and the secondary optimized equivalent load are shown in table 3.

TABLE 3

And the dispatching department can formulate a reasonable electricity price according to the optimized load curve to transfer the user to participate in demand side management. Obviously, the demand price elasticity e and the constant a are optimal parameter values determined in long-term investigation of the response law of the user to the real-time electricity prices. The real-time electricity prices made according to the load rates in the table 3 are all in a reasonable interval and are easy to implement.

The optimization result of the real-time output of the controllable micro-source is shown in fig. 4. As can be seen from fig. 4, the energy storage unit purchases electricity from the power grid in the low electricity price section at night and discharges electricity in the high electricity price section in the daytime, the output is particularly obvious during the peak electricity price period, but the output is obviously reduced in the time period of high load but relatively low electricity price in the 45-60 interval, which indicates that in the economic optimization scheduling: 1) the effect of energy storage regulation is that the load expected transfer rate of real-time electricity price is determined instead of the actual load capacity, and the larger the difference between the peak and the valley of the electricity price is, the more obvious the effect of energy storage peak regulation is; 2) the microgrid can reasonably utilize the characteristics of the energy storage units to realize indirect load transfer. In addition, because the unit power cost of the micro-combustion engine 1 is higher than that of the micro-combustion engine 2, the output of the micro-combustion engine 2 is preferentially increased to supply power to the load in the system when the load is increased, and the output of the micro-combustion engine is obviously increased when the unit power cost of a power grid is higher than that of the micro-combustion engine.

Therefore, the demand side two-stage optimization scheme and the real-time electricity price making scheme can fully guide energy storage to play a self peak regulation role, can effectively realize load peak clipping and valley filling, and reduce cost.

In addition, the results of the improved SAGA and the traditional GA solution to the model are compared here. To follow the single variable principle and reliability, the other operations of each algorithm were performed after initializing the population and the average of the 40 solution results of each algorithm were compared. The algorithmic comparison and optimization costs are shown in fig. 5 and table 4, respectively.

TABLE 4

As can be seen from fig. 5 and table 4, SAGA not only maintains the simple and easy features of the GA algorithm and accelerates the convergence rate, but also reduces the cost by about 8.9% compared with the solution result of the GA algorithm, enhances the local search capability of the GA, and improves the robustness and convergence accuracy of the algorithm.

4.2.2 protocol comparison

To verify the validity and feasibility of the protocol, four comparison protocols were set up herein, of which:

scheme 1: single-layer optimization with only piconet layer scheduling.

Scheme 2: two-layer optimization including only DR first-level optimization. Wherein DR primary optimization result determination

And

and then the three are used as the optimized input of the microgrid layer to optimize the output of other conventional power supplies. The scheme focuses on improving the utilization rate of new energy, so the scheme is brought into comparison for verification.

Scheme 3: two-layer optimization including only DR second-level optimization. Since there is no first-order optimization, the load will be predicted

Wind and light predicted contribution

Direct substitution of the corresponding amount in formula (23) with

The two are taken together as the input quantity of the secondary optimization of the scheme, and the optimization result is obtained

All micro-source output is optimized as input for optimizing the micro-grid layer. The proposal is included in comparison to verify the influence of pure addition of secondary optimization on each index, particularly the influence on improving the stability of the controllable micro-source.

Scheme 4: double layer optimization with two-stage response to DR. The coordination optimization of the scheme on each index is verified through comparison.

The simulation analysis of the index comparison of the four schemes is as follows:

1) power rate of clean energy

Under the existing clean energy conversion efficiency, the clean energy generating capacity is exerted to the maximum extent, and the reduction of the electric quantity discarded by the clean energy is one of effective ways for improving the utilization rate of the clean energy. Defining the clean energy power abandonment rate as follows:

in the formula, r_CSThe power rate of the clean energy is abandoned;

and respectively the predicted output and the optimized output of the ith clean energy at the moment t.

Each scheme is optimized as shown in figure 6, and the clean energy power curtailment rate is shown.

TABLE 5

Quantitative analysis shows that the DRE power abandonment rate can be reduced by two-stage optimization, and the comprehensive power abandonment rate of the scheme four is reduced by about 10 percent compared with that of the scheme one. The effect of reducing the power abandonment rate of the first-stage optimization is more obvious than that of the second-stage optimization.

2) Controllable micro-source output stability index

The stationarity index expression is as follows:

in the formula, r_stableIs a stability index;

optimizing output for the ith controllable micro source;

is the average force over the sampling period.

The optimization results of the schemes are shown in FIG. 7, and the stationarity indexes are shown in Table 6.

TABLE 6

According to the formula (39), the smaller the stationarity index is, the smaller the output fluctuation of the micro source is, and the more the maintenance of the controllable micro source is facilitated. The comparison results of the four schemes show that the secondary optimization can effectively reduce the stability index and improve the stability of the controllable micro-source output; comparison of the stationarity indexes of the scheme 3 and the scheme 4 shows that the first-level optimization has a certain negative effect on the second-level optimization effect, but the first-level optimization is still obviously improved compared with the scheme 1 and the scheme 2, and the effectiveness of the scheme is proved.

3) Average load factor

The power system load fluctuation condition can be represented by an average load rate

In the formula, r_loadIs the average load rate;

the optimized maximum load;

to optimize the average value of the afterload.

The optimization results of each scheme are shown in fig. 8. The load fluctuation situation is shown in table 7.

TABLE 7

The results of fig. 8 and table 7 show that any stage of optimization of the demand side optimization scheme can effectively realize timing adjustment of transferable loads, and transfer the demand response load from a high electricity price interval to a low electricity price interval, so as to achieve the purpose of peak clipping and valley filling. FIG. 8 shows that the fourth scheme not only reduces the peak-to-valley difference, but also improves the average load rate, so that the load curve is smoother.

4) Cost of each scheme

Each solution optimizes cost pairs such as table 8.

TABLE 8

The optimization cost of each scheme shows that although the clean energy power generation cost is high, the requirement-side two-stage optimization-based micro-grid double-layer optimization scheme provided by the invention can coordinate and optimize each index (particularly, remarkably reduce the clean energy power discard amount) and simultaneously effectively reduce the comprehensive operation cost of the micro-grid.

5 conclusion

Establishing a comprehensive operation cost optimization scheduling model of the micro-grid; and simultaneously introducing a two-stage optimization model of DR, and solving the model by adopting an improved SAGA algorithm, wherein the result shows that:

1) the genetic algorithm based on simulated annealing better improves the convergence rate and accuracy of GA.

2) The method for formulating the real-time electricity price based on the load rate is simple, convenient and reasonable, is easy to implement, and can guide the energy storage system to fully exert the peak regulation effect of the energy storage system.

3) Comparison of indexes of the four schemes shows that the DR first-stage optimization can improve the utilization rate of clean energy, the second-stage optimization can reduce the stability index of the controllable micro-source, the two-stage optimization can fully exert the potential of peak clipping and valley filling of demand response, and the comprehensive economic benefit of micro-grid operation is further improved on the premise of coordinating and optimizing the indexes.

Uncertainty of access of distributed devices such as DR, wind, light and electric vehicles brings opportunities and challenges to dynamic economic dispatching research of micro-grids and energy Internet, and a reasonable dynamic response mechanism and optimized modeling of the distributed devices still need to be deeply explored and researched.

Claims

1. A grid-connected microgrid double-layer optimization method based on two-stage demand response is characterized by specifically comprising the following steps:

2) initializing a first-level optimized population: selecting clean energy as primary optimized energy, collecting primary optimized energy data as optimized individuals, and optimizing data information contained in each individual at the primary level: electric power P output by wind turbine in same time period_wt,rOutput power P of photovoltaic module_pv,rLoad shift amount P in first-level optimization_Lin,1And the amount of discharge P_Lout,1；

3) Bringing each individual generated in step 2) into an objective function

Calculating the corresponding objective function value, wherein the individual with the minimum objective function value is the optimal, the individual adaptability value is the most optimal, and in order to obtain consistent solving results, the reciprocal of the objective function value is taken as the adaptability value of each individual, wherein

The equivalent load after the first-level optimization;

4) sorting the fitness value of each individual in the population in the step 2) from large to small by adopting tournament selection, reserving the optimal individual, then carrying out cross variation operation on other individuals to obtain a new individual corresponding to the optimal individual, and forming a new population with the reserved optimal individual;

7) bringing each individual generated in step 6) into an objective function

Calculating the corresponding objective function value, wherein

8) sorting the fitness value of each individual in the population in the step 6) from large to small by adopting tournament selection, reserving the optimal individual, then carrying out cross variation operation on other individuals to obtain a new individual corresponding to the optimal individual, and forming a new population with the reserved optimal individual;