1. A power grid multi-time scale multi-target energy optimization scheduling method is characterized by comprising the following steps: the method comprises the following steps:

s1, collecting historical operation data of the power grid;

s2, establishing an uncertainty description power grid weak robust multi-objective optimization model, a day-ahead optimization scheduling objective function and a day-ahead optimization scheduling constraint condition based on the historical operation data, and establishing a day-ahead long-time weak robust multi-objective optimization scheduling model;

s3, solving the long-time weak robust multi-objective optimization scheduling model in the day ahead to obtain a day ahead scheduling plan;

s4, establishing an intra-day control time domain and performance indexes based on the actual scheduling plan of the power grid and the day-ahead scheduling plan, and constructing an intra-day MPC rolling optimization scheduling model;

and S5, modifying the day-ahead scheduling plan in a rolling manner in real time based on the intraday MPC rolling optimization scheduling model.

2. The power grid multi-time scale multi-target energy optimization scheduling method according to claim 1, wherein: the objective function of the optimized scheduling comprises a cost operation function and a total pollutant gas emission quantity function, wherein the cost operation function is used for describing the operation cost of the power grid, the operation cost comprises the power grid operation cost and a risk cost, and the cost operation function is as follows:

F₍₁₎＝min(C₀+C_cut)，

wherein, F₍₁₎For cost operationNumber, C₀Representing the operating costs of the grid, C_cutRepresents a risk cost;

the total pollutant gas emission function is as follows:

in the formula, F₍₂₎The total amount P converted from the discharge of the pollution gas generated in the power grid to the light abandoning of the wind abandoning_DGTo the output power of the grid, gamma_qwtTo discard the air quantity, gamma_qpvIn order to discard the light quantity, a, b and c are emission coefficients of the polluted gas when the gas turbine operates, and d is a pollution coefficient calculated by discarding wind and light.

3. The power grid multi-time scale multi-target energy optimization scheduling method according to claim 2, wherein: the power grid operation cost comprises power generation cost, energy storage charging and discharging cost and demand response load scheduling cost:

C₀＝C_DG+C_BA+C_DDR，

wherein, C₀Representing the operating costs of the grid, C_DGRepresenting the cost of electricity generation of the grid, C_BARepresenting the energy storage charge-discharge cost of the grid, C_DDRCosts are scheduled for demand response loads.

4. The power grid multi-time scale multi-target energy optimization scheduling method according to claim 2, wherein: the risk cost comprises wind abandoning cost, light abandoning cost and load shedding cost, and the expression of the risk cost is as follows:

C_cut＝C_loss+C_qwt+C_qpv，

wherein, C_qwtCost for wind abandonment, C_qpvCost for light rejection, C_lossCost for load shedding;

λ_qwtpenalty factor for wind curtailment, λ_qpvFor light rejection penalty factor, λ_lossA penalty factor for load shedding; gamma ray_lossThe load was cut.

5. The power grid multi-time scale multi-target energy optimization scheduling method according to claim 4, wherein: the day-ahead optimized scheduling constraint conditions comprise day-ahead power balance constraint, demand response constraint and wind-abandoning and light-abandoning load shedding constraint,

the expression of the day-ahead power balance constraint is:

wherein the content of the first and second substances,representing the actual value of the load in the grid a day before,the actual value of the wind power output before the day is shown,representing the actual photovoltaic output value, P, before the day_DGRepresenting the gas turbine output, P, day ahead_BARepresenting the energy storage output before the day. P_DDRRepresenting the actual dispatching power of the demand response load in the final dispatching plan before the day, and t representing a certain period in 24 hours;

the expression of the demand response constraint is:

wherein, P_DDR(t) actual dispatch power for the demand response load at the final dispatch plan; k_DDRGiving compensation cost and scheduling cost of demand response load to the power grid;responding to expected power consumption provided by a user for a demand in the day ahead, wherein delta t is a scheduling step length;

the expression of the wind abandoning light abandoning load shedding constraint is as follows:

wherein the content of the first and second substances,the upper limit of the load shedding power is shown,the upper limit of the wind curtailment power is shown,respectively, the upper limit of the dump optical power.

6. The power grid multi-time scale multi-target energy optimization scheduling method according to claim 1, wherein: the S4 includes:

s4.1, constructing a power grid state space prediction model based on the historical operation data of the power grid and the day-ahead scheduling plan;

s4.2, setting a control time domain and performance indexes in a day, establishing a rolling optimization objective function, converting the rolling optimization objective function into a quadratic programming standard model, solving the quadratic programming standard model to obtain M control quantities in a prediction time domain, and executing a first control quantity in the prediction time domain;

and S4.3, feedback correction is carried out, the power grid state space prediction model is updated, and S4.1 is returned.

7. The power grid multi-time scale multi-target energy optimization scheduling method according to claim 6, wherein: the power grid state space prediction model comprises a state variable, a control variable and a disturbance variable, and comprises the following steps:

wherein G (t) is the system state quantity of the controlled object, and Δ G (t) represents the control quantity or related input quantity of the object at the time t; h (t) represents the system output quantity at the time t; A. b, C, D are respectively a system matrix, an input matrix, an output matrix, and a disturbance matrix.

8. The power grid multi-time scale multi-target energy optimization scheduling method according to claim 7, wherein: the rolling optimization objective function is:

ΔG(t)＝[ΔG(t)^T，ΔG(t+1)^T，...，ΔG(t+M-1)^T，ΔG(t+M)^T]^T

wherein, F_objIs a tracking objective function; x (t) is a reference quantity issued by an upper layer at the time t; Δ G (t) is a control quantity at the time t, and M is a prediction time domain; δ and θ are weight coefficient matrices.

9. The power grid multi-time scale multi-target energy optimization scheduling method of claim 8, wherein: the feedback correction updates the power grid state space prediction model as follows:

ρr(t+1)＝H(t+1)-H(t+1|t)，

wherein rho is an error coefficient matrix; h (t +1| t) is the predicted output for t +1 at time t.

Background

With the economic development, energy becomes an indispensable important resource. However, since the world enters an industrialized age, traditional energy sources such as coal mines and petroleum are developed in large quantities, so that the global energy resources are exhausted. Meanwhile, burning conventional fossil energy emits a large amount of pollution gas, which causes global environmental deterioration. Therefore, renewable energy becomes the key point of research for solving the problem of energy shortage and environmental protection, and meanwhile, the distributed power generation technology is widely applied. Under the ambitious goals of 2030 carbon peak and 2060 carbon neutralization in China, the low-carbonization transformation of energy becomes an important development strategy of energy systems in China, and the high-proportion renewable energy grid connection becomes an important characteristic of future power systems in China. Aiming at the recent target of carbon peak reaching in 2030, the total installed wind power and solar energy is more than 12 hundred million kilowatts in 2030, the non-fossil energy consumption percentage reaches 25%, and the proportion of renewable energy in energy consumption is further increased to more than 60% in 2050 in China. Intermittent renewable energy sources such as wind power, photovoltaic and the like are gradually changed from supplementary energy sources in a power system into main energy sources, and the consumption of the renewable energy sources becomes one of main tasks of a power grid in China.

Intermittent renewable energy sources such as wind power, photovoltaic and the like have obvious difference from the traditional thermal generator set in power generation characteristics, and in order to absorb high-proportion renewable energy sources, a power system undergoes profound changes. The proportion of installed capacity of new energy power generation in a power grid is increased, so that the number of unit combinations and operation modes is increased sharply, and the new energy power generation has randomness, so that great challenges are brought to the system for making a power generation plan, an operation mode and scheduling operation, the spare capacity of the system is increasingly deficient, the power grid regulation capacity is reduced, and the impact generated by the power fluctuation of the new energy power generation is difficult to balance. Meanwhile, the intermittent and fluctuating nature of the generated power of the new energy also changes the tidal current distribution and flow direction of the accessed power grid, and the voltage and tie line power near the Point of Common Coupling (PCC) of the new energy station in the power grid will exceed the safety range. With the large-scale access of new energy power generation to a power grid, due to the characteristics of randomness, intermittence, fluctuation and the like of factors such as climate, environment and the like, the output of the new energy power generation changes frequently, the output of the new energy power generation still lacks sufficient stability, and the adverse factors cause the output fluctuation of a new energy station during grid connection so as to bring adverse effects to the stable operation of a power system, thereby seriously restricting the full utilization of the new energy power generation. Taking 2019 as an example, the electricity quantity abandoned by China reaches 169 hundred million kilowatts, and the direct economic loss is about 101.4 million yuan.

Therefore, a power grid multi-time scale and multi-target energy optimization scheduling method is needed to solve the problem that the power grid cannot achieve expected operation economy due to source load uncertainty and prediction error influences in the traditional scheduling method.

Disclosure of Invention

The invention aims to provide a power grid multi-time scale and multi-target energy optimal scheduling method, which is used for solving the problems in the prior art and the problem that the power grid cannot achieve expected operation economy due to source load uncertainty and prediction error in the traditional scheduling method, can be well applied to power grid optimal scheduling, and obtains good economy and environmental protection on the basis of ensuring certain robustness.

In order to achieve the purpose, the invention provides the following scheme: the invention provides a power grid multi-time scale and multi-target energy optimal scheduling method, which comprises the following steps of:

s1, collecting historical operation data of the power grid;

s2, establishing an uncertainty description power grid weak robust multi-objective optimization model, a day-ahead optimization scheduling objective function and a day-ahead optimization scheduling constraint condition based on the historical operation data, and establishing a day-ahead long-time weak robust multi-objective optimization scheduling model;

s3, solving the long-time weak robust multi-objective optimization scheduling model in the day ahead to obtain a day ahead scheduling plan;

s4, establishing an intra-day control time domain and performance indexes based on the actual scheduling plan of the power grid and the day-ahead scheduling plan, and constructing an intra-day MPC rolling optimization scheduling model;

and S5, modifying the day-ahead scheduling plan in a rolling manner in real time based on the intraday MPC rolling optimization scheduling model.

Preferably, the objective function of the optimized scheduling includes a cost operation function and a total amount of pollutant gas emission function, wherein the cost operation function is used for describing operation costs of the power grid, the operation costs include power grid operation costs and risk costs, and the cost operation function is:

F₍₁₎＝min(C₀+C_cut)，

wherein, F₍₁₎As a function of cost operation, C₀Representing the operating costs of the grid, C_cutRepresents a risk cost;

the total pollutant gas emission function is as follows:

in the formula, F₍₂₎The total amount P converted from the discharge of the pollution gas generated in the power grid to the light abandoning of the wind abandoning_DGTo the output power of the grid, gamma_qwtTo discard the air quantity, gamma_qpvIn order to discard the light quantity, a, b and c are emission coefficients of the polluted gas when the gas turbine operates, and d is a pollution coefficient calculated by discarding wind and light.

Preferably, the grid operation cost includes power generation cost, energy storage charging and discharging cost and demand response load scheduling cost:

C₀＝C_DG+C_BA+C_DDR，

wherein, C₀Representing the operating costs of the grid, C_DGRepresenting the cost of electricity generation of the grid, C_BARepresenting the energy storage charge-discharge cost of the grid, C_DDRCosts are scheduled for demand response loads.

Preferably, the risk cost includes a wind curtailment cost, a light curtailment cost, and a load shedding cost, and the expression of the risk cost is:

C_cut＝C_loss+C_qwt+C_qpv，

wherein, C_qwtCost for wind abandonment, C_qpvCost for light rejection, C_lossCost for load shedding;

λ_qwtpenalty factor for wind curtailment, λ_qpvFor light rejection penalty factor, λ_lossA penalty factor for load shedding; gamma ray_lossThe load was cut.

Preferably, the day-ahead optimized scheduling constraint conditions comprise a day-ahead power balance constraint, a demand response constraint and a wind curtailment and light curtailment load shedding constraint,

the expression of the day-ahead power balance constraint is:

wherein the content of the first and second substances,representing the actual value of the load in the grid a day before,the actual value of the wind power output before the day is shown,representing the actual photovoltaic output value, P, before the day_DGRepresenting the gas turbine output, P, day ahead_BARepresenting the energy storage output before the day. P_DDRRepresenting the actual dispatching power of the demand response load in the final dispatching plan before the day, and t representing a certain period in 24 hours;

the expression of the demand response constraint is:

wherein, P_DDR(t) actual dispatch power for the demand response load at the final dispatch plan; k_DDRGiving compensation cost and scheduling cost of demand response load to the power grid;responding to expected power consumption provided by a user for a demand in the day ahead, wherein delta t is a scheduling step length;

the expression of the wind abandoning light abandoning load shedding constraint is as follows:

wherein the content of the first and second substances,the upper limit of the load shedding power is shown,the upper limit of the wind curtailment power is shown,respectively, the upper limit of the dump optical power.

Preferably, the S4 includes:

s4.1, constructing a power grid state space prediction model based on the historical operation data of the power grid and the day-ahead scheduling plan;

s4.2, setting a control time domain and performance indexes in a day, establishing a rolling optimization objective function, converting the rolling optimization objective function into a quadratic programming standard model, solving the quadratic programming standard model to obtain M control quantities in a prediction time domain, and executing a first control quantity in the prediction time domain;

and S4.3, feedback correction is carried out, the power grid state space prediction model is updated, and S4.1 is returned.

Preferably, the power grid state space prediction model includes a state variable, a control variable and a disturbance variable, and the power grid state space prediction model is:

wherein G (t) is the system state quantity of the controlled object, and Δ G (t) represents the control quantity or the related input quantity of the object at the time t; h (t) represents the system output quantity at the time t; A. b, C, D are respectively a system matrix, an input matrix, an output matrix, and a disturbance matrix.

Preferably, the rolling optimization objective function is:

ΔG(t)＝[ΔG(t)^T，ΔG(t+1)^T，...，ΔG(t+M-1)^T，ΔG(t+M)^T]^T

wherein, F_objIs a tracking objective function; x (t) is a reference quantity issued by an upper layer at the time t; Δ G (t) is a control quantity at the time t, and M is a prediction time domain; δ and θ are weight coefficient matrices.

Preferably, the feedback correction updates the grid state space prediction model as:

ρr(t+1)＝H(t+1)-H(t+1|t)，

wherein rho is an error coefficient matrix; h (t +1/t) is the predicted output of t +1 at time t.

The invention discloses the following technical effects:

according to the power grid multi-time scale and multi-target energy optimization scheduling method provided by the invention, a weak robustness multi-target optimization method is adopted to obtain a compromise scheduling plan meeting the economical efficiency and the environmental protection performance in a day-ahead long-time scale scheduling stage, and an MPC (multi-control processor) is adopted to track the day-ahead scheduling plan in a day-ahead short-time scale stage, so that the problem that the power grid cannot reach the expected operation economical efficiency due to the influence of source load uncertainty and prediction error in the traditional scheduling method can be solved. The method can be well applied to power grid optimized dispatching in an isolated grid mode, and good economy and environmental protection are obtained on the basis of ensuring certain robustness.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.

FIG. 1 is a schematic diagram of a day-ahead-day multi-objective optimization scheduling process of a power grid in the embodiment of the invention;

FIG. 2 is a schematic diagram of grid wind and light power in three operation scenarios in this embodiment;

fig. 3 is a schematic diagram of the power grid load power in three operation scenarios in the present embodiment;

FIG. 4 is a schematic diagram of the output power of the gas turbine of the power grid in an operating scenario of the present embodiment;

fig. 5 is a schematic diagram of energy storage charging and discharging power of a power grid in a certain operation scenario in this embodiment;

fig. 6 is a schematic diagram of a demand response actual/expected power consumption plan of a power grid in an operation scenario in this embodiment;

FIG. 7 is a pareto frontier diagram of a weak robust multi-objective optimization model;

FIG. 8 is a schematic diagram of a scheduling plan in this embodiment;

FIG. 9 is a comparison of the power at risk of operating at 3% error in an embodiment of the present invention;

FIG. 10 is a comparison of the running power at 5% error for the risk in the embodiment of the present invention;

FIG. 11 is a comparison of the power of the operation at 10% error in the embodiment of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.

The invention provides a power grid multi-time scale and multi-target energy optimal scheduling method which is used for solving the problem of power grid optimal scheduling, and the power grid multi-time scale and multi-target energy optimal scheduling method in the embodiment is shown in figure 1.

Historical operation data of the power grid are collected, and an uncertainty description power grid weak robust multi-objective optimization model is established according to the historical operation data. In order to meet the load requirement in a power grid and improve the wind and light consumption, a translatable load is introduced in the day-ahead scheduling to perform excitation type demand response, the power load in the peak period is transferred, and peak clipping and valley filling are realized. Aiming at the problem of uncertainty in a power grid, the power grid optimization scheduling method adopts a weak robust optimization method to carry out power grid optimization scheduling, and a relaxation variable gamma is added on the basis of the traditional robust optimization, so that the method has the physical significance of risk operation variables of wind abandonment, light abandonment and load shedding, and the system is allowed to have certain risk operation. The embodiment selects the minimum operation cost as a first objective function, and the objective is the operation cost C of the power grid network₀And a risk cost C_cutThe composition is shown in formulas (1) to (2):

F₍₁₎＝min(C₀+C_cut) (1)

C₀＝C_DG+C_BA+C_DDR (2)

in the formula, C_DGRepresenting the cost of power generation of the gas turbine, C_BACost of charging and discharging for energy storage, C_DDRCosts are scheduled for demand response loads.

In this embodiment, the pollutant emissions of the gas turbine are mainly CO₂、SO₂And NO_xThe gas turbine has different output, the discharged pollution gas is not used, and the wind and light abandoning parts are used as gasTurbine emissions, so the emission model used in this embodiment is shown in equation (3):

in the formula, F₍₂₎The total amount P converted from the discharge of the pollution gas generated in the power grid to the light abandoning of the wind abandoning_DGIs the output power of the gas turbine, gamma_qwtTo discard the air quantity, gamma_qpvIn order to discard the light quantity, a, b and c are emission coefficients of the polluted gas when the gas turbine operates, and d is a pollution coefficient calculated by discarding wind and light.

In this embodiment, the various costs are calculated as follows:

(1) gas turbine

In this embodiment, because the optimized scheduling in the isolated grid mode is considered, and a controllable unit which responds faster needs to participate, the operating efficiency of the gas turbine is selected as shown in formula (4):

wherein P is_DG(t) is the output power of the gas turbine, η_DG(t) is its operating efficiency.

The gas turbine generates electricity by consuming fuel, and fuel cost and operation and maintenance cost are generated in the operation process, and the expressions are shown as formulas (5) and (6):

C_DG(t)＝K_DGP_DG(t)+C_G.F(t) (5)

in the formula: c_DG(t) represents the cost of power generation by the gas turbine; c_G.F(t) is its fuel cost; k_DGThe operation and maintenance cost is the same; t represents a t period; c_ONRepresenting the gas cost and LHV representing the gas lower heating value.

(2) Demand side response

When the power grid adjusts the demand response load power utilization plan, the normal power utilization plan of the demand response load can be influenced, so the power grid can give a certain compensation to demand response users according to a new scheduling plan, and the scheduling cost C is achieved_DDR(t) may be represented as shown in formula (7):

in the formula: p_DDR(t) actual dispatch power for the demand response load at the final dispatch plan; k_DDRGiving compensation cost and scheduling cost of demand response load to the power grid;Δ t is the scheduling step size for the day ahead demand to respond to the desired power usage provided by the user.

(3) Cost of risk

The risk cost in this embodiment refers to wind abandon, light abandon cost and load shedding cost generated when the renewable energy output or load scene does not satisfy the power balance of the power grid, and the cost is as shown in formulas (8) and (9):

C_cut＝C_loss+C_qwt+C_qpv (8)

in the formula, C_qwtCost for wind abandonment, C_qpvCost for light rejection, C_lossCost for load shedding; lambda [ alpha ]_qwtPenalty factor for wind curtailment, λ_qpvFor light rejection penalty factor, λ_lossA penalty factor for load shedding; gamma ray_lossThe load was cut.

The day-ahead optimized scheduling constraint conditions comprise a day-ahead power balance constraint, a gas turbine power constraint, a demand response constraint and a wind curtailment and light curtailment load shedding constraint.

After introducing the relaxation variable, the power balance constraint before the day that the operation of the power grid needs to meet is as shown in equation (11):

in the formula (I), the compound is shown in the specification,representing the actual value of the load in the grid a day before,the actual value of the wind power output before the day is shown,representing the actual photovoltaic output value, P, before the day_DGRepresenting the gas turbine output, P, day ahead_BARepresenting the energy storage output before the day. P_DDRRepresenting the actual dispatch power of the demand-response load at the final dispatch plan by day, and t represents a certain period of 24 hours.

Wherein the total load after the power grid dispatching is as shown in formula (12):

in the formula, P_LOADIs the total load after scheduling.

The daily power balance is as shown in equation (13):

an actual value representing the load in the power grid during the day;representing an actual value of wind power output in a day;representing the actual value of the photovoltaic output in the day;representing the actual output power of the gas turbine in the day;and the actual output of the stored energy in the day is shown.

Since the gas turbine responds quickly, only the output power constraint, neglecting its ramp-up constraint, is considered, as shown in equation (14):

in the formula (I), the compound is shown in the specification,representing the maximum/minimum output power of the gas turbine.

And (3) constraint of demand response:

the demand response load mainly considers the load translation problem, and the following conditions should be satisfied, as shown in formulas (15) to (16):

in the formula: s_DDRFor the total power demand within the scheduling period:andrespectively representing the maximum and minimum loads required to meet the demand response in each time period.

Wind abandoning, light abandoning and load shedding restraint:

the wind abandoning light abandoning load shedding power is required to meet the constraint condition in scheduling as shown in the formula (17):

in the formulaAndrespectively representing the upper limit of the tangential load, the upper limit of the curtailed wind and the upper limit of the curtailed optical power.

According to the peer-to-peer conversion theory, the power balance constraint of equation (10) translates to the equation (18):

P_LOAD(t) represents a predicted value of the load in the grid before the day; p_WT(t) representing a predicted value of the wind power output before the day; p_PV(t) represents a predicted photovoltaic output value before the day;representing maximum fluctuation of wind power output before the day;representing maximum photovoltaic output fluctuation before the day;representing the maximum fluctuation of the regular load after removing the demand response load in the grid a day before.

In conclusion, in the day-ahead stage, by using a weak robustness multi-objective optimization scheduling method, the minimum running cost of the power grid and the total emission amount of the pollution gas are taken as optimization objectives, and the NSGA-II is used for solving and obtaining a day-ahead scheduling plan for balancing economy and environmental protection in each hour in the day-ahead stage.

In the in-day stage, in order to meet the problem of deviation between the actual scheduling plan of the power grid and the current weak robust economic optimization scheduling plan, a MPC rolling optimization scheduling model is established by taking 15min as a control time domain and 1h as a prediction time domain and adopting traceability as a performance index. The basic principle of MPC is to predict the system output in the future finite time domain according to the current state of the system, obtain the optimal control quantity in the future finite time domain by an optimization mode, use the first component for system control, and repeat the above process at each moment. The model predictive control can well solve the optimization problem containing uncertainty factors, and has stronger anti-interference capability and robustness. In addition, when MPC is used, the acceptability of the prediction model is high and no specific model is required. Multiple constraints can be added, so that the method is more suitable for the power system optimization scheduling considering the uncertainty. The power grid contains a large number of uncontrollable distributed power sources, and the randomness and the uncertainty of the uncontrollable distributed power sources are suitable for tracking correction by using the MPC. MPC is a closed-loop control method with good tracking and anti-interference capability, and comprises three links of model prediction, rolling optimization and feedback correction. By introducing feedback correction, the deviation of the optimized scheduling result caused by random factors can be timely and effectively corrected, so that the MPC is adopted to correct the day-ahead scheduling plan in a real-time rolling manner in the embodiment so as to improve the accuracy of the scheduling plan.

(1) Prediction model

The model capable of predicting the future output response of the system according to the historical information and the future input information of the controlled object is called a prediction model, and the prediction model is composed of a state variable, a control variable and a disturbance variable. In this embodiment, a power grid state space prediction model is established according to the actual values of the gas turbine, the stored energy output and the risk operating power of the power grid at the time t, as shown in formula (19):

in the formula: g (t) is the system state quantity of the controlled object, and Δ G (t) represents the control quantity or related input quantity of the object; h (t) represents the system output; A. b, C, D are respectively a system matrix, an input matrix, an output matrix, and a disturbance matrix.

(2) Roll optimization

And solving according to the optimal performance index in the rolling time domain, wherein the rolling optimization is optimized by taking the selected economic cost or the minimum deviation from the plan proposed in the day as a target, and because the optimization in the time domain is limited, the rolling optimization can be repeated for many times so as to make up the influence caused by the uncertainty problem in the model. In this embodiment, a rolling finite time domain optimization method is adopted for prediction control of the power grid state space prediction model. The optimization performance index function is a quadratic optimization function, and is expressed by the following formula (20):

in the formula: f_objIs a tracking objective function; h (t) is output quantity at the moment t, namely output power and risk operation power of the gas turbine and the stored energy; x (t) is a reference quantity issued by an upper layer at the time t; Δ G (t) is a control quantity at the time t, namely a quantity change value of the gas turbine and the energy storage output; m is a prediction time domain; δ and θ are weight coefficient matrices.

After the equation (20) is converted into the quadratic programming standard form, the Hessian matrix of the quadratic programming is a positive definite matrix, so that the global minimum value is unique and cannot fall into the local optimum. And optimizing the performance indexes on M time domains at the time t to obtain a control sequence in the prediction time domain from t to t + M as the formula (21), only executing the control quantity at the first time, and repeating the steps at the time t + 1.

△G(t)＝[△G(t)^T,△G(t+1)^T，...,△G(t+M-1)^T,△G(t+M)^T]^T (21)

(3) Feedback correction

In the rolling optimization process in the day, errors are generated between the actual output quantity and the predicted value due to insufficient prediction errors, uncertainty and modeling precision, and the corrected scheduling control method still possibly has deviation from the actual operation of the system, so that the actual value and the predicted value of the output variable are detected at each sampling moment, the errors are updated through an equation (4), and the prediction model is subjected to feedback correction and then subjected to next rolling optimization.

ρr(t+1)＝H(t+1)-H(t+1|t) (22)

In the formula: rho is an error coefficient matrix; h (t +1| t) is the predicted output for t +1 at time t.

Simulation analysis: simulating three operation scenes, and respectively offsetting 3%, 5% and 10% of the predicted values of the source load from the day before to the inferior (rigid load is increased, wind and light output is reduced) and then superposing 3%, 5% and 10% of random errors as actual values of the source load, wherein the specific output is shown in fig. 2 and 3; the load prediction fluctuation proportion is 10%, and the wind and light output prediction error fluctuation proportion is 15%. Each punishment cost is lambda_t＝[λ_loss,λ_qwt,λ_qpv]^T＝[1.2，1，1]^TDischarge coefficients [ a, b, c ] for any time instant]^TIs [0.00079,0.025,21.9 ]]^TFrom calculating a robust uncertainty Γ_t2.4867 allows a maximum out-of-limit percentage of maximum risk operation of 10%.

The scheduling result of the scheduling method provided by the embodiment under the scene 1 (the source load predicted value shifts to be inferior (rigid load is increased, wind and light output is reduced) by 3%) is shown in fig. 4-6, fig. 4 shows the output of the gas turbine, and the gas turbine selects the minimum mode to operate because the power load is small and the wind and light output is large in 1-7 h and 24 h; fig. 5 shows energy storage charging and discharging power, and energy storage is charged for 2 hours, 4-5 hours, 7 hours and 24 hours and discharged for 9 hours and 18-21 hours due to low load demand at night, so that load power utilization in peak period is met. Demand response load scheduling is illustrated in fig. 6, which presents a desired power usage plan primarily focused on peak load hours. On the premise that the power grid meets the total power requirement and power constraint at each time interval, the power requirements of 10-12 h and 16-22 h are respectively adjusted to 1-7 h and 24h, and the insufficient load at the peak time interval and the wind curtailment power at the valley time interval are reduced. FIG. 7 is a pareto frontier of a weak robust multi-objective optimization model; the running cost is 673.11 yuan, and the environmental pollution amount is 846.77 kg.

During scheduling in the day, prediction errors and uncertainty factors are coped by a gas turbine and stored energy together, during actual operation, due to random factor interference, a scheduling plan in the day needs to be adjusted to meet system power balance, and feedback correction improves accuracy of a control strategy, and a scheduling plan after MPC rolling optimization in the day is adopted under three scenes as shown in FIG. 8, wherein a risk operation positive value represents a load shedding amount, and a negative value represents a wind curtailment amount and a light curtailment amount. As can be seen from the figure, because the actual conditions in the day have the influence of uncertain factors, the MPC can correct the day-ahead scheduling plan according to the actual conditions in the day, adjust the output plan and meet the requirement of system power balance.

In order to further verify the feasibility of the optimized scheduling method provided by the method in the isolated network mode, 4 different scheduling methods are respectively set. The 4 scheduling methods are respectively. 1) The invention provides a power grid multi-time scale optimization scheduling method; 2) adopting weak robust economic optimization scheduling in the day ahead, and selecting an open-loop control method for wind abandoning, light abandoning and load shedding processing for errors in the day; 3) optimizing and scheduling by adopting a random optimization algorithm in the day ahead, and adopting an MPC rolling optimization method in the day; 4) the traditional robust optimization scheduling is adopted in the day ahead, and the MPC rolling optimization method is adopted in the day.

MPC performance comparison verification:

TABLE 1

As can be seen from the comparison in table 1, the operation cost of the scheduling method 1) provided in this embodiment is lower than 2) in all three scenarios. Because isolated network operation and day-ahead plan making consider the fluctuation of the source load in the day, when the fluctuation in the day is too small, redundant wind and light can be abandoned, and because certain punishment is carried out on the behavior of wind abandoning and light abandoning, the wind abandoning and light abandoning in the day causes the environmental pollution amount to be slightly higher than that in the day-ahead under the condition of small fluctuation. Under the actual operation condition in the day, the output of the unit can be reduced through MPC control, so that the situations of wind abandonment and light abandonment are reduced, and the comprehensive cost of the power grid in the whole day is reduced compared with a method of not adopting MPC in the day. The intra-day risk operating power versus the pre-day risk operating plan deviation ratios are shown in table 2:

TABLE 2

As can be seen from the comparison in table 2, in the method 1) provided in this embodiment, the absolute mean of deviation is lower than 2) in the three scenarios, the method 1) and the risk operating rate are shown in fig. 9 to 11, and when the current scheduling plans are the same, the economy and the trackability of the MPC rolling optimization adopted in the day are better than those of the non-MPC rolling optimization. Simulation results show that in an actual operation state in the day, by adopting MPC rolling optimization, a day-ahead scheduling plan can be well tracked and corrected to meet the actual operation state, so that the power grid still has good economy when the actual operation state fluctuates in a large range, and the total amount of risk operation power correction in the whole day is smaller, thereby reducing the change of the power grid risk operation power scheduling plan and ensuring that the power grid can meet the economic operation of the power grid.

Comparison of robust performance:

the comparison of the combined cost and the environmental pollution amount of the three methods 1), 3) and 4) under the same three errors is shown in Table 3:

TABLE 3

As can be seen from the comparison in table 3, although the cost of the method 1) is higher than that of the method without using the weak robustness in the past day, under the actual operation scene where the source load prediction error and the uncertainty are unavoidable, the cost in the day is lower than that of the method without using the weak robustness, and the larger the fluctuation degree is, the lower the cost of the method 1) and the environmental pollution amount are. Compared with the method 4), the method 1) can be seen, although the traditional robust optimization can meet the feasibility of the scheduling plan under any condition when the day-ahead scheduling plan is made, and the system safety is ensured. But the risk operation condition can better adapt to various scenes under the allowable condition, and the pursuit of the optimal comprehensive benefit is facilitated. The deviations of the intra-day risk operating power and the pre-day risk operating plan of the different methods are shown in table 4:

TABLE 4

As can be seen from a comparison of Table 4, the risk total power of operation of method 1) is less than that of the other methods. The source load prediction error and uncertainty influence are inevitably caused in an actual operation scene through comparative analysis, a random optimization method is adopted in the day ahead, the obtained scheduling plan is an economic optimal scheduling plan in a specific scene, when the scheduling plan is actually operated, the deviation of the scheduling plan becomes large along with the increase of the random error, and the continuous adoption of the MPC to track the non-weak robust scheduling plan causes the increase of the day operation cost and the risk operation power deviation of the power grid. And by adopting traditional robust optimization, the planned day-ahead economic optimal scheduling plan is too conservative and has poor applicability, so that the economic optimal scheduling plan cannot be ensured during tracking period scheduling in the day. Therefore, the weak robust economic optimization scheduling plan is adopted in the day-ahead scheduling stage, so that the tracking scheduling in the day is more practical.

Under the condition of considering operating cost and environmental factors, a weak robustness multi-objective optimization method is adopted to calculate a compromise scheduling plan meeting economical efficiency and environmental protection performance in a day-ahead long-time scale scheduling stage, and an MPC (multi-control processor) is adopted to track the day-ahead scheduling plan in a day-in short-time scale stage, so that the problem that the power grid cannot achieve expected operating economical efficiency due to source load uncertainty and prediction error influence in a traditional scheduling method can be solved. Compared with random optimization, the method has the advantages that better tracking effect can be obtained by adopting MPC rolling optimization under the scene of considering source load prediction error and uncertainty through comparison of simulation results; compared with the traditional robust optimization method, the wind and light absorption capacity and the load demand of the power grid under the condition of large probability can be ensured, the economy and the environmental protection performance of the power grid can be improved, the conservation of the traditional robust optimization is effectively improved, and the economy and the environmental protection performance are considered while the better robustness is ensured. Therefore, the method provided by the invention can be well applied to power grid optimization scheduling, and good economy and environmental protection are obtained on the basis of ensuring certain robustness.

The above-described embodiments are merely illustrative of the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various modifications and improvements of the technical solutions of the present invention can be made by those skilled in the art without departing from the spirit of the present invention, and the technical solutions of the present invention are within the scope of the present invention defined by the claims.