High energy load park affine optimization scheduling method considering wind and light correlation

Abstract

The invention relates to a high-energy-load park affine optimization scheduling method considering wind and light correlation, and belongs to the field of energy optimization scheduling. The method comprises the following steps: acquiring prediction intervals of output and load demands of wind and light units in a park, constructing an initial parallelogram correlation model, and converting the initial parallelogram correlation model into a multi-constraint affine form model in combination with an improved adaptive polygon algorithm; a particle swarm optimization algorithm is used for dynamically adjusting constraint conditions, redundant intervals are eliminated, and irregular polygon affine constraints fitting the reality are generated; the method comprises the following steps: converting a deterministic model of electrolytic aluminum, energy storage and thermal power generating units into an affine form, and constructing an optimal scheduling model containing power balance, equipment operation and cross-time constraint; and solving by adopting a mixed integer linear programming solver in order to minimize the operation cost. According to the method, by dynamically correcting the wind-light correlation area, the conservative property of the uncertainty interval is reduced, the new energy consumption rate is improved, the starting and stopping cost of the thermal power generating unit is reduced, and collaborative optimization of economical efficiency and reliability is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial park optimal scheduling, in particular to an affine optimal scheduling method for high-load energy park considering wind-solar correlation. BACKGROUND

[0002] The industrial park with high-load energy as the mainstay has higher requirements for power grid stability and energy supply reliability due to its huge power demand. In the face of the continuous rise of new energy penetration rate, the industrial park urgently needs to rely on optimal scheduling strategy to resolve the contradiction between supply and demand. In view of this, for the high-load electrolytic aluminum industrial park, through the adjustable characteristics of electrolytic aluminum and the cooperation of energy storage device adjusting load peak and valley and regular load in a short time, the demand power is flexibly adjusted, the renewable energy consumption capacity is enhanced, and the carbon emission is reduced, so as to realize the green and sustainable development goal.

[0003] In the research field of traditional power system optimal scheduling, there are various uncertain factors, such as intermittency of wind power and photovoltaic power generation and random fluctuation of load demand, which bring challenges to traditional deterministic optimization method. By considering the uncertainty and correlation of new energy unit output through affine optimal scheduling, the interval width is significantly reduced compared with interval optimization, and the conservativeness is reduced. However, there are still some problems in considering the correlation problem. The correlation region is limited by the correlation coefficient, and the actual operation rule is not reflected. Further combining the actual production data with the affine scheduling considering the wind-solar correlation, it is of great significance to cope with the challenge of new energy fluctuation, obtain more flexible and economic optimal scheduling scheme, and improve the overall efficiency of the industrial park while ensuring the continuity of production. SUMMARY

[0004] The purpose of the present application is to provide an affine optimal scheduling method for high-load energy park considering wind-solar correlation, which uses wind-solar unit historical operation data and future prediction data to ensure the comprehensiveness and continuity of the data, so as to more accurately reflect the uncertain output of new energy unit, obtain more flexible and economic optimal scheduling scheme, and improve the overall efficiency of the industrial park while ensuring the continuity of production.

[0005] To achieve the above purpose, the present application provides an affine optimal scheduling method for high-load energy park considering wind-solar correlation, comprising the following steps:

[0006] (1) Collecting the wind turbine, photovoltaic unit output prediction interval and load demand prediction interval of the industrial park, constructing an initial parallelogram correlation model based on the correlation of wind-solar historical operation data, and defining the wind and light output fluctuation interval and its correlation coefficient through constraint conditions;

[0007] (2) Based on the cutting line equation and the adaptive polygon algorithm for irregular convex polygon region, the initial parallelogram correlation model is converted into an affine form model containing wind and light output fluctuation interval, correlation coefficient and reality constraint;

[0008] (3) The particle swarm optimization algorithm is adopted to minimize the area of the modified polygon region as the optimization objective, dynamically adjust the slope coefficient and constant coefficient of the cutting line, and eliminate redundant intervals to obtain an irregular polygon affine constraint that fits the actual operation;

[0009] (4) The deterministic model of electrolytic aluminum equipment, energy storage equipment and thermal power unit in the park is converted into an affine form model, combined with the affine constraint of step (3), an affine interval optimization scheduling model of high energy-consuming industrial park containing power balance constraint, equipment operation constraint and cross-time period constraint is constructed;

[0010] (5) Taking the minimum operation cost of the park as the objective function, the Gurobi mixed integer linear programming solver is used to solve the optimization scheduling model, and the optimization scheduling result is output.

[0011] Further, the step (1) of constructing the initial parallelogram correlation model comprises the following steps:

[0012] Step 1.1 According to the correlation parallelogram

[0013] |ρ -1 T -1 R -1 D|≤e (1)

[0014] ρ xy =(b-a) / (a+b) (2)

[0015] Where: a and b are the length of two half-diagonal of parallelogram, a is the half-length of short diagonal, b is the half-length of long diagonal Where: a and b are the length of two half-diagonal of parallelogram, a is the half-length of short diagonal, b is the half-length of long diagonal

[0016] The correlation coefficient matrix of two uncertain variables ρ It can be constructed as:

[0017]

[0018] Where: T is the correlation matrix; R is the interval matrix, controlling the interval size; D is the translation matrix; e is the unit matrix.

[0019] Step 1.2: Define the auxiliary matrix A = ρ -1 T -1 R -1 ;

[0020] That is

[0021]

[0022]

[0023] Step 1.3: The original parallelogram satisfies the constraints as follows

[0024] -1≤ineq1≤1

[0025] -1≤ineq2≥1 (6)

[0026] Further, step (2) converts the wind-solar parallelogram correlation model into constraints based on improved adaptive polygon, and constructs an affine form model considering wind-solar uncertainty, correlation and integrity; the irregular convex polygon is analytically transformed and adaptively corrected, and the parallelogram and convex polygon are unified.

[0027] Step 2.1: Define the cutting line equation

[0028] k1ineq1+k2ineq2≤R3

[0029] k3ineq1+k4ineq2≥R4 (7)

[0030] If k2 and k3 are both 1, the equality holds, and the expression is represented as two straight lines.

[0031] Step 2.2: Correlation-based multi-deformation region constraint condition

[0032] Overall constraint:

[0033]

[0034] Step 2.3 According to the formula (6)

[0035] -2≤ineq1+ineq2≤2 (10)

[0036] Step 2.4: The value ranges of k1, k2, R3, k3, k4 and R4 are respectively:

[0037]

[0038] When k2 and k3 are both 1, and ineq1 and ineq2 are two variables, then formula (8) represents the area surrounded by 6 straight lines.

[0039] Step 2.5 On the basis of considering the correlation between wind and light, the wind-solar output affine form is rewritten as

[0040]

[0041] Step 2.6: Convert formula (12) into inequality form, and get: ​

[0042]

[0043] Where inf(·) represents the lower bound of interval number; sup(·) represents the upper bound of interval number.

[0044] Step 2.7: Construct k1A i +B i , and the calculation of which obtains an interval number That is:

[0045]

[0046] When two intervals are subtracted in an affine form, the operation is very close to the true value when the correlation coefficient is negative. The true conversion region is the mirror image of. At this time, it can be considered that it is in Figure 3 In the affine form, the straight lines AF and CE can be represented, and the straight lines CF and AE can be selected by a suitable k 22 Obtained. Two cutting GH, IJ are also selected by a suitable k value.

[0047] Step 2.8: Adaptive polygon interval affine form complete constraint considering correlation:

[0048]

[0049] In the formula,

[0050]

[0051] Further, step (3) uses a particle swarm algorithm to convert the correlation into a straight line slope coefficient and constant coefficient in an inequality constraint for optimization:

[0052] Step 3.1: Area area correction needs to determine the values of k1, R3, k4, R4, and the particle swarm optimization algorithm is used to determine the slope coefficient and constant coefficient of the two straight lines. The optimization goal is to minimize the area of the cutting region.

[0053] Particle velocity update formula:

[0054]

[0055] In the formula is the velocity of particle i in d-dimensional space in the t+1 generation; w is the inertia weight; c1 is the individual learning factor; r1, r2 are random numbers generated in the range of [0, 1]; respectively represent the velocity and global extreme value of the tth generation.

[0056] Particle position update formula:

[0057]

[0058] In the formula, is the position of the particle i in the d-dimensional space, the t+1 generation.

[0059] Step 3.2: The real range of wind and light output is a parallelogram area under the consideration of the effect of correlation without considering the effect of reality. And the real range of wind and light output is an irregular polygon area under the consideration of the effect of reality and correlation, as can be seen. The method can take into account the correlation between wind and light output and the reality of wind and light operation, so that the affine value range of wind and light output is obtained by the first correction from the initial parallelogram area .

[0060] Step 3.3: Obtain the value of k 33 , k 44 , R 33 , R 44 when the area of the region is the smallest by the particle swarm optimization algorithm. The excess area is removed to show the modified irregular polygon area, which is more consistent with the real operation range and helps to obtain an affine optimization result with lower conservatism. Set the number of particles to 50, the maximum number of iterations to 100, the cognitive factor C1=1.4, the social factor C2=1.4, and the inertia weight w=0.7.

[0061] Further, step (4) the remaining equipment in the park is converted from a deterministic model to an affine form model, and an affine model of the high-load industrial park is constructed for optimization and solution.

[0062] Step 4.1: Establish the model of the aluminum electrolysis equipment:

[0063] Power-current constraint: The output power of the aluminum electrolysis load is controlled by controlling the current of the electrolytic cell, and its power-current constraint is as follows:

[0064]

[0065] Power-temperature constraint: Because the electrolytic cell has thermal inertia, when the power changes, it will produce a large fluctuation to the electrolytic cell in a short time. The power-temperature constraint for adjacent time is as follows:

[0066]

[0067]

[0068] In the formula: T t al is the production temperature of the aluminum electrolysis load at time t; are the upper and lower limits of the temperature of the aluminum electrolysis load, generally taken as 970 and 950℃ respectively. al , m al are the specific heat capacity coefficient and mass of the electrolyte respectively.

[0069] Step 4.2: Establishing the energy storage device model:

[0070] Power constraint:

[0071]

[0072] wherein are the upper and lower limits of the energy storage capacity, respectively, is the energy storage capacity value at time t.

[0073] State uniqueness constraint:

[0074]

[0075] wherein are the energy storage charging and discharging states at time t, respectively. is the energy storage charging and discharging power value at time t. Upper and lower limits of charging power value. Upper and lower limits of charging power value.

[0076] Power-capacity constraint:

[0077]

[0078] wherein μ bess is the energy storage self-discharge rate, η bch , η bdis is the charging and discharging efficiency.

[0079] Step 4.3: Establishing the thermal power unit model:

[0080] Upper and lower limits of output power constraint:

[0081]

[0082] wherein P gmin , P gmax represent the minimum and maximum output limits of the thermal power unit, represents the output of the thermal power unit at time t.

[0083] Ramp constraint:

[0084]

[0085] wherein represent the upper and lower limits of the ramp of the thermal power unit, represents the output of the thermal power unit at time t-1.

[0086] Step 4.4: Building system power balance constraints

[0087]

[0088] where, is the regular load except the electrolytic aluminum load, are the actual output of photovoltaic and wind turbine respectively.

[0089] Further, step (5) defines the interval affine form of the demand load, and the electrolytic aluminum, energy storage device, and thermal power unit constraints are converted into interval affine form, to construct an affine interval optimization scheduling model for high-load industrial parks considering wind and light uncertainty, correlation, and completeness.

[0090] Step 5.1: Define the affine form of the load

[0091]

[0092] where is the affine variable corresponding to the load demand power at time t; are the center values of the corresponding affine variables respectively; is obtained from the fluctuation range; The value range is [-1, 1]

[0093] Step 5.2: All device constraints in the park are converted into affine form constraints, mainly including three types of constraints: equality constraints, inequality constraints, and cross-time constraints. For equality constraints of other decision variables not including state variables, the equality condition is converted according to the corresponding between the center values on both ends of the equality and the corresponding between the noise elements; for example:

[0094]

[0095] where is the variable of the electrolytic aluminum affine form, is the variable center value of the electrolytic aluminum affine form; is the noise element coefficient of the electrolytic aluminum affine form.

[0096] Step 5.3: Inequality constraints are mainly the upper and lower limit constraints of the output of the park model. The minimum and maximum values that can be taken are constrained, as shown in equation (32).

[0097]

[0098] Step 5.4: For, the park device model with cross-time constraints will be converted as follows:

[0099]

[0100] where: and are the affine forms of the energy device output in the industrial park at time t and t-1 respectively; R down and Rup are the set lower and upper limits of the slope, respectively; and are the center value and noise coefficient of the i-th uncertain factor, respectively. and are the center value and noise coefficient of the i-th uncertain factor, respectively.

[0101] Step 5.4: The objective function of the model is

[0102] min F total = F G + F BESS (34)

[0103]

[0104]

[0105] According to the basic theory of affine algorithm, the objective function in affine form can be further converted equivalently, which is divided into two parts of operation cost and center value and affine radius. Therefore, formula (34) can be further defined as:

[0106]

[0107] In the formula, F total,0 , F total,i are the center value and noise coefficient of the i-th uncertain factor of the industrial park operation cost, respectively; F G,0 , F BESS,0 , F G,i , F BESS,i are the center value and noise coefficient of the i-th uncertain factor of the t-period industrial park energy storage and thermal power unit operation cost, respectively; ω is the optimization weight, and its value range is [0, 1].

[0108]

[0109]

[0110] Step 5.5: The optimization scheduling model is

[0111]

[0112] The industrial park day-ahead affine optimization model shown in formula (40) is a mixed integer linear programming problem. Wherein formula (19)-(29) are all converted into affine constraint form.

[0113] The optimization scheduling model is solved by using the Gurobi solver, and an interval affine optimization scheduling result of the high-load industrial park is output.

[0114] Compared with the prior art, the following beneficial effects are achieved:

[0115] 1. Reduce the conservatism and narrow the uncertainty interval: by constructing an affine model considering the correlation between wind and light output, and combining the particle swarm optimization algorithm to dynamically correct the constraint region, the extreme values with extremely low probability are eliminated, the uncertainty interval width is significantly narrowed, and the conservatism of the scheduling scheme is reduced.

[0116] 2. Improve the new energy consumption rate: through the flexible adjustment of electrolytic aluminum load and the coordinated scheduling of energy storage equipment, the wind and light output fluctuation is dynamically matched, the phenomenon of abandoned wind and light is reduced, and the new energy consumption rate is significantly improved.

[0117] 3. Optimize the economy: taking the minimum operation cost as the objective function, combining affine optimization and mixed integer programming, while ensuring the continuity of production, the central value of operation cost is reduced, and the economy and conservatism are considered.

[0118] 4. Dynamically adapt to actual operation demand: based on historical data and predicted data, the adaptive polygon algorithm is used to dynamically correct the model region, so that the scheduling scheme is more suitable for the real distribution of wind and light output, and the scheduling accuracy is improved.

[0119] 5. Reduce the start-stop cost of thermal power units: through the coordinated optimization of energy storage and electrolytic aluminum load, the load peak-valley difference is smoothed, the frequent peak regulation demand of thermal power units is reduced, the start-stop frequency is reduced, the equipment life is prolonged, and the operation cost is reduced by about 30%.

[0120] 6. Flexibly cope with complex correlation: break through the traditional rectangular or parallelogram region limit, construct irregular polygon affine constraint, more truly reflect the nonlinear correlation characteristics between wind and light output, and improve the robustness of the scheduling scheme. BRIEF DESCRIPTION OF DRAWINGS

[0121] Figure 1 It is a flowchart of the method of the present application;

[0122] Figure 2 It is the predicted interval of wind power output of the present application;

[0123] Figure 3 It is the predicted interval of load demand of the present application;

[0124] Figure 4 It is the predicted interval of photovoltaic output of the present application;

[0125] Figure 5 It is an effect diagram of the correction process 1 of the present application;

[0126] Figure 6 Effect diagram of the revision process 2 of the present application;

[0127] Figure 7 Different weight operation cost intervals of the present application;

[0128] Figure 8 Electric power affine scheduling result of the present application. DETAILED DESCRIPTION

[0129] Embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0130] The present embodiment can effectively eliminate the possible values with extremely low probability in the uncertain interval and reduce scheduling conservatism by using the wind-light correlation model in two-dimensional plane affine form and according to the correlation and dependence between noise elements.

[0131] The adaptive polygon affine modeling method considering correlation proposed in the present application can obtain a more actual scheduling scheme by combining wind-light historical output and prediction data and dynamically revising the polygon region, and provide a high-precision and low-conservative solution for industrial park multi-energy collaborative optimization.

[0132] By introducing high-load adjustable electrolytic aluminum load in the industrial park to improve the overall efficiency of the industrial park, on the basis of ensuring production continuity, through the collaborative interaction of electrolytic aluminum high-load load and energy storage equipment, the new energy consumption rate is improved, the start-stop cost of thermal power units is reduced, the new energy consumption rate is improved, and the economy and new energy utilization efficiency are considered.

[0133] As shown in Figure 1 comprise the following steps:

[0134] S1: Collect the source-load prediction interval of the industrial park, and construct an initial parallelogram correlation model.

[0135] S1.1 According to the correlation parallelogram

[0136] |ρ -1 T -1 R -1 D|≤e (1)

[0137] ρ xy =(b-a) / (a+b) (2)

[0138] Where: a and b are the length of two half-diagonal of parallelogram, a is the half-length of short diagonal, b is the half-length of long diagonal Where: a and b are the length of two half-diagonal of parallelogram, a is the half-length of short diagonal, b is the half-length of long diagonal

[0139] The correlation coefficient matrix of two uncertain variables

[0140]

[0141]

[0142] It can be constructed as:

[0143] S1.2: Define auxiliary matrix A = p -1 T -1 R -1

[0144] i.e.

[0145]

[0146]

[0147] S1.3: The original parallelogram satisfies the constraints as follows

[0148] -1≤ineq1≤1

[0149] -1≤ineq2≥1 (6)

[0150] S2: Based on the improved adaptive polygon, the wind-light parallelogram correlation model is converted into constraints, and an affine form model considering wind-light uncertainty, correlation and integrity is constructed; the irregular convex polygon is analytically transformed and adaptively corrected, and the parallelogram and convex polygon are unified.

[0151] S2.1: Define the cutting line equation

[0152] k1ineq1+k2ineq2≤R3

[0153] k3ineq1+k4ineq2≥R4 (7)

[0154] In the formula, k2 and k3 will take 1, then when the equation is established, the expression is represented as two straight lines.

[0155] S2.2: Correlation consideration of multi-deformation area constraint condition

[0156] Overall constraint:

[0157]

[0158] S2.3 According to the formula (6) know

[0159] -2≤ineq1+ineq2≤2 (10)

[0160] S2.4: i.e. the value range of k1, k2, R3, k3, k4, R4 is respectively

[0161]

[0162] When k2, k3 are both 1, and ineq1, ineq2 are two variables, then formula (8) represents a region surrounded by 6 straight lines.

[0163] S2.5 On the basis of considering the correlation between wind and light ， Rewrite the wind and light output in affine form as ：

[0164]

[0165] S2.6: Transform formula (12) into inequality form, and obtain:

[0166]

[0167] Where inf(·) represents the lower bound of the interval number; sup(·) represents the upper bound of the interval number.

[0168] S2.7: Construct k1A i +B i , and calculate an interval number That is:

[0169]

[0170] When two intervals are subtracted in affine form, the operation is very close to the true value when the correlation coefficient is negative. The true transformation region is the mirror image of. At this time, it can be considered that the operation is in Where: T is the correlation matrix; R is the interval matrix, controlling the interval size; D is the translation matrix; e is the unit matrix. The straight lines AF and CE can be represented in, and similarly, the straight lines CF and AE are selected with appropriate k 22 . Two cuts GH, IJ are also selected with appropriate k values.

[0171] S2.8: Considering the correlation, the complete constraint of the adaptive polygon interval affine form is as follows:

[0172]

[0173] In the formula,

[0174]

[0175] S3. Use the particle swarm algorithm to transform the correlation into inequality constraints of the slope coefficient and constant coefficient of the straight line to optimize;

[0176] S3.1: Area correction needs to determine the values of k1, R3, k4, and R4. The particle swarm optimization algorithm is used to determine the slope coefficient and constant coefficient of the two straight lines, and the optimization goal is to minimize the area after cutting.

[0177] Particle velocity update formula:

[0178]

[0179] where is the velocity of particle i in d-dimensional space at the t+1 generation; w is the inertia weight; c1 is the individual learning factor; r1, r2 are randomly generated numbers in the range of [0, 1]; and represent the velocity of the t generation and the global extreme value, respectively.

[0180] Particle position update formula:

[0181]

[0182] where, is the position of particle i in d-dimensional space at the t+1 generation.

[0183] S3.2: In the case where the effect of correlation is not considered, the real range of wind and light output is a parallelogram region. In the case where the effects of correlation and reality are considered, the real range of wind and light output is an irregular polygon region, as can be seen. The proposed method can take into account the correlation between wind and light output and the reality of wind and light operation, so that the affine value range of wind and light output is changed from the initial parallelogram region to after the first correction.

[0184] S3.3: Obtain the values of k 33 , k 44 , R 33 , R 44 by the particle swarm optimization algorithm. The excess area is removed to form a modified irregular polygon region, which is more consistent with the real operation range and helps to obtain an affine optimization result with lower conservatism. Set the number of particles to 50, the maximum number of iterations to 100, the cognitive factor C1 = 1.4, the social factor C2 = 1.4, and the inertia weight w = 0.7.

[0185] S4: The remaining equipment in the park is converted from the deterministic model to the affine form model, and the affine model of the high-load industrial park is constructed for optimization.

[0186] S4.1: Establish the model of the aluminum electrolysis equipment:

[0187] Power-current constraint: The power output of the aluminum electrolysis load is controlled by controlling the current of the electrolytic cell, and its power-current constraint is as follows:

[0188]

[0189]

[0190]

[0191] Power-temperature constraint: Because the electrolytic tank has thermal inertia, when the power changes, it will produce a large fluctuation to the electrolytic tank in a short time. The power-temperature constraint for adjacent time is as follows:

[0192]

[0193]

[0194] wherein T t al is the production temperature of the electrolytic aluminum load at time t; are the upper and lower limits of the temperature of the electrolytic aluminum load, respectively, and the temperature is set to 970, 950 ℃, respectively.c al , m al are the specific heat capacity coefficient and mass of the electrolyte, respectively.

[0195] S4.2: Establishing a model of energy storage equipment:

[0196] Power constraint:

[0197]

[0198] wherein are the upper and lower limits of the energy storage capacity, respectively, is the energy storage capacity value at time t.

[0199] State unique constraint:

[0200]

[0201]

[0202]

[0203] wherein are the charging and discharging working states of the energy storage at time t. is the charging and discharging power value of the energy storage at time t. is the upper and lower limit value of the charging power. is the upper and lower limit value of the charging power.

[0204] Power-capacity constraint:

[0205]

[0206] wherein μ bess is the self-discharge rate of the energy storage, η bch , η bdis is the charging and discharging efficiency.

[0207] S4.3: Establishing a model of thermal power generating unit:

[0208] Upper and lower limit constraint of output power:

[0209]

[0210] Pmin gmin Pmax gmax Pmin P(t)

[0211] Ramp constraints:

[0212]

[0213] Pmin Pmax P(t-1)

[0214] S4.4: Construct system power balance constraints:

[0215]

[0216] P(t) P(t) P(t)

[0217] S5: Define the interval affine form of demand load, convert the constraints of electrolytic aluminum, energy storage devices, thermal power units into interval affine form, and construct an affine interval optimization scheduling model for high-load industrial parks considering wind and light uncertainty, correlation and completeness.

[0218] S5.1: Define the affine form of load

[0219]

[0220] P(t) P(t) P(t) P(t) [-1, 1]

[0221] S5.2: Convert all device constraints in the park into affine form constraints, mainly three types of constraints conversion of equality constraints, inequality constraints and cross-time constraints. For equality constraints of other decision variables not including state variables, according to the corresponding between the center values of both ends of the equality and the corresponding between the noise elements, the equality condition is converted, for example:

[0222]

[0223] P(t) P(t) is the center value of the affine form of the variable of electrolytic aluminum; is the noise element coefficient of the affine form of the noise element of electrolytic aluminum.

[0224] S5.3: The inequality constraint is mainly the upper and lower limit constraint of the output of the park model, which constrains the minimum and maximum values that can be taken, as shown in equation (32).

[0225]

[0226] S5.4: For, the park equipment model with cross-time constraints will adopt the following transformation:

[0227]

[0228] In the formula: and are the affine forms of the energy equipment in the industrial park at time period t and time period t-1, respectively; R down and R up are the set ramp-down lower limit and upper limit, respectively; and are the center value, noise element coefficient of and .

[0229] S5.4: Taking the park operating cost as the target, the storage and high-load energy adjustment power of electrolytic aluminum are optimized under the conditions of satisfying various constraints such as thermal power units, wind power, photovoltaic units, energy storage and load. The objective function of the model is

[0230] min F total = F G + F BESS (34)

[0231]

[0232]

[0233] According to the basic theory of affine algorithm, the affine form of the objective function can be further equivalently converted into two parts of operating cost and center value and affine radius. Therefore, equation (34) can be further defined as:

[0234]

[0235] In the formula, F total,0 , F total,i are the center value and noise element coefficient i of the uncertain factor of the industrial park operating cost; F G,0 , F BESS,0 , F G,i , F BESS,i, are the central value and noise element coefficient of the uncertain factor i of the operation cost of the energy storage and thermal power unit in the industrial park for the time period t, respectively; ω is the optimization weight, and the value range is [0, 1].

[0236]

[0237]

[0238] S5.5: The optimization scheduling model is:

[0239]

[0240] The industrial park day-ahead affine optimization model shown in formula (40) is a mixed integer linear programming problem. Wherein formula (19)-(29) are all converted into affine constraint form.

[0241] S6: The Gurobi solver is used to solve the optimization scheduling model, and the interval affine optimization scheduling result of the high-load industrial park is output.

[0242] The source and load prediction values of the industrial park within 24 hours are shown in formula (41)-(43). Figure 3 , Figure 2 and Figure 3 , wherein the fluctuation degree of the fan and photovoltaic output is set to 10% of the predicted value, and the fluctuation degree of the electric load is set to 5% of the predicted value; the rated power of the aluminum electrolysis plant is 600 MW, and the current aluminum electrolysis has a regulation range of 5%-10% above the rated point and 25%-30% below the rated point, and the regulation range of the aluminum electrolysis is set to 5% of its rated value. The 10% of the rated power of the thermal power unit 500 MW is set as its regulation range for regular peak shaving. The charging and discharging power of the energy storage unit 100 MW is set to 20%-80% of its capacity, and the remaining 20% rated capacity is set as the limit regulation fluctuation range.

[0243] Figure 4 , Figure 5 are the correlation region schematic diagram after the first correction and the correlation region schematic diagram after the second correction of the method, respectively.

[0244] In order to reflect the advantages of the affine optimization method and considering the correlation of wind and light, the following three methods are used for comparison, and the operation cost interval of the three methods is shown in Table 1.

[0245] Method 1: Interval affine optimization algorithm without considering the correlation of wind and light;

[0246] Method 2: Interval affine optimization algorithm considering the correlation of wind and light (ρ xy =-0.5);

[0247] Method 3: Interval affine optimization scheduling method considering correlation and true self-adaptive polygon;

[0248] Table 1. Operating cost range for different methods

[0249]

[0250] Because the affine optimization method does not consider correlation, the interval values ​​are taken as the entire rectangle, resulting in a strong conservatism that easily leads to extreme cases. Therefore, as shown in Table 1 above, Method 1's operating cost interval completely covers that of Methods 2 and 3, exhibiting the greatest conservatism. Method 2, by using the PM model to narrow the fluctuation range of new energy units, achieves a narrower operating cost interval than the interval optimization method. Furthermore, Method 3, considering the realism of wind and solar power output, eliminates redundant value regions by considering correlation in the PM model, thereby further narrowing the operating cost interval.

[0251] The different correlation coefficient values ​​for Method 2 are as follows, from Figure 6 It can be seen that the closer the correlation coefficient is to -1, the smaller the parallelogram region, and the less conservative it is. Since the central value of operating costs represents the deterministic optimization result, the interval width represents the degree of fluctuation in operating costs considering uncertainty. Method 3 takes into account the realism of wind and solar operation, and its interval width can be compared with ρ in Method 2. xy The results for the operating cost range of -0.8 are similar, and the median operating cost of Method 3 is similar to that of Method 2. xy = -0.9 is closer. Its result is closer to the deterministic operation result, while also having an appropriate interval width to adapt to the uncertain output on the source load side.

[0252] Table 2. Operating cost ranges for different correlation coefficients in Method 2.

[0253]

[0254] Setting different weight values ​​in the objective function will shift the optimization scheduling scheme towards either an economical or conservative approach. For example... Figure 6 As shown, when the weight When the weights are large, the results tend to be conservative, resulting in a wider interval. At that time, the interval width increased significantly, and the central value of operating costs and The differences are not significant, combining good economic efficiency with conservatism.

[0255] like Figure 7 By setting different weight values ​​in the objective function shown, the optimized scheduling scheme will shift towards either an economical or conservative approach. For example... Figure 7 As shown, when the weight When the weights are large, the results tend to be conservative, resulting in a wider interval. At that time, the interval width increased significantly, and the central value of operating costs and The differences are not significant, combining good economic efficiency with conservatism.

[0256] The electrolytic aluminum load is coordinated with the energy storage device through current regulation, which significantly improves the system flexibility. As shown in Figure 7 (a), when the wind turbine output is high and the photovoltaic output is zero (0:00-1:00), the electrolytic aluminum load actively reduces the power. During the period (3.00-4.00), the wind turbine output increases, the electrolytic aluminum increases the power, the energy storage device is charged, and the minimum power operation of the thermal power unit is ensured to reduce the unit start-stop cost. During the period (9.00-17.00), the wind turbine output fluctuates sharply, and the photovoltaic unit gradually reaches the peak output and then gradually decreases. The remaining power of the load is reasonably distributed by the energy storage system, which weakens the peak-valley difference of the net load at this time, as shown in Figure 8 (c). During the period (18.00-23.00), the wind turbine output gradually increases, and the photovoltaic unit output is almost zero. At this time, the electrolytic aluminum reduces the power. During the whole process, the electrolytic aluminum adjusts the power according to the yield constraint and the energy storage coordinated scheduling enhances the consistency of supply and demand in the park. This strategy increases the new energy consumption rate and reduces the start-stop cost of the thermal power unit. Figure 8 Figure 8 (b) represents the scheduling interval of the remaining devices in the park under the fluctuation of source and load. The optimal scheduling interval can be obtained by adjusting the affine radius according to the uncertain output of new energy units. It can be seen that the electrolytic aluminum provides sufficient scheduling margin for the dispatchers.

Claims

1. A high-load energy park affine optimization scheduling method considering wind-light correlation, characterized in that, The method comprises the following steps: (1) collecting the wind turbine, photovoltaic unit output prediction interval and load demand prediction interval of the industrial park, constructing an initial parallelogram correlation model based on the correlation of wind and light historical operation data, and defining the wind and light output fluctuation interval and its correlation coefficient through constraint conditions; (2) based on the cutting line equation and the adaptive polygon algorithm for irregular convex polygon region, the initial parallelogram correlation model is converted into an affine form model containing the wind and light output fluctuation interval, the correlation coefficient and the authenticity constraint; (3) using the particle swarm optimization algorithm, taking the minimum area of the modified polygon region as the optimization objective, dynamically adjusting the slope coefficient and constant coefficient of the cutting line, and removing the redundant interval to obtain the irregular polygon affine constraint fitting the actual operation; (4) converting the deterministic model of the electrolytic aluminum equipment, energy storage equipment and thermal power unit in the park into an affine form model, combining the affine constraint of step (3), and constructing an affine interval optimization scheduling model of the high-load industrial park containing power balance constraints, equipment operation constraints and cross-period constraints; (5) taking the minimum operation cost of the park as the objective function, using the Gurobi mixed integer linear programming solver to solve the optimization scheduling model, and outputting the optimization scheduling result.

2. The affine optimization scheduling method of claim 1, wherein, The step (1) of constructing the initial parallelogram correlation model comprises the following steps: Step 1.1: according to the correlation coefficient of wind and light output historical data, a parallelogram model satisfying the following constraints is constructed; |ρ -1 T -1 R -1 D|≤e (1) p xy = (b - a) / (a + b) (2) In the formula: a and b are the lengths of two half diagonals of the parallelogram, a is the half length of the short diagonal, and b is the half length of the long diagonal The correlation coefficient matrix ρ of the two uncertain variables can be constructed as: Wherein, T is the correlation matrix; R is the interval matrix, which controls the interval size, D is the translation matrix, and e is the unit matrix; Step 1.3: verify whether the constraint conditions of the parallelogram model satisfy the correlation distribution of wind and light output historical data: Step 1.2: Define the auxiliary matrix A = p -1 T -1 R -1 for calculating the joint volatility interval of wind, light output -1≤ineq1≤1 -1≤ineq2≥1 (6). The step (2) of generating the affine form model comprises the following steps:

3. The method of claim 1, wherein the method further comprises: Step 2.1: define the cutting line equation: k1ineq1+k2ineq2≤R3 k3ineq1+k4ineq2≥R4 (7) In the formula, k2 and k3 are both 1, and the equation is established, which is expressed as two straight lines; Step 2.2: consider the multi-transformation region constraint condition of correlation: Overall constraint: -2≤ineq1+ineq2≤2 (10) Step 2.3 according to formula (6) known: Step 2.4: the value range of k1, k2, R3, k3, k4 and R4 is: When k2 and k3 are both 1, and ineq1 and ineq2 are two variables, then formula (8) represents the region surrounded by 6 straight lines; Wherein, inf(·) represents the lower bound value of the interval number; sup(·) represents the upper bound value of the interval number; Step 2.5 on the basis of the wind and light correlation ， Rewrite the wind and light output in affine form to a linear expression containing the correlation coefficient: Step 2.6: Transforming equation (12) into inequality form gives expression (13): (12) Transforming equation (12) into inequality form gives expression (13): Step 2.8: consider the adaptive polygon interval affine form complete constraint of correlation as follows: Step 2.7: Construction of k1A i + B i for which an interval number In the formula, The step (3) of the irregular polygon affine constraint comprises the following steps:

4. The method of claim 1, wherein the method further comprises: ​ Step 3.1: The area of the region needs to be determined k1, R3, k4, R4 value, using particle swarm optimization algorithm, the area of the region minimum as the optimization goal, determine the slope coefficient and constant coefficient of two straight lines; in the optimization process, according to the particle velocity update formula and particle position update formula, the particle position is constantly iterated; Particle velocity update formula: wherein is the velocity of particle i in d-dimensional space at the t+1th generation; w is the inertia weight; c1 is the individual learning factor; r1, r2 are randomly generated numbers in the range [0, 1]; respectively denote the velocity of the tth generation and the global extreme value. Particle position update formula: wherein is the position of particle i in d-dimensional space for the t+1 generation; Step 3.2: The real range of wind and light output is a parallelogram area without considering the effect of correlation and reality, and is an irregular polygon area with considering the effect of correlation and reality, as can be seen; the proposed method can take into account the correlation between wind and light output and the reality of wind and light operation, so that the affine value range of wind and light output is changed from the initial parallelogram area to k 11 、 22 、 ​ Step 3.3: k is obtained by particle swarm optimization algorithm when the area of the region is minimum 33 , k 44 , R 33 , R 44 ; the excess area is removed to show the modified irregular polygon region, which is more consistent with the real running range, and helps to obtain an affine optimization result with lower conservativeness; the number of particles is set to 50, the maximum number of iterations is 100, the cognitive factor C1 = 1.4, the social factor C2 = 1.4, and the inertia weight w = 0.

7.

5. The method of claim 1, wherein the method further comprises: The construction of affine interval optimization scheduling model of high load industrial park in step (4) includes the following steps: Step 4.1: Establishing electrolytic aluminum equipment model: Power-current constraint: electrolytic aluminum load controls the output power size by controlling the current of electrolytic cell, and its power-current constraint is as follows: Power-temperature constraint: for the power-temperature constraint of adjacent time, the following is obtained: In the formula: is the production temperature of the electrolytic aluminum load at time t; are the upper and lower limits of the temperature of the electrolytic aluminum load, respectively, and the temperature is set to 970, 950°C;c al , m al are the specific heat capacity coefficient and the mass of the electrolyte, respectively; Step 4.2: Establishing energy storage equipment model: Power constraint: In the formula respectively, upper and lower limits of the energy storage capacity, is the energy storage capacity value at time t. State unique constraint: In the formula respectively, the energy storage charge and discharge working state at time t; is the energy storage charge and discharge power value at time t; upper and lower limit values of the charge power; upper and lower limit values of the charge power; Power-capacity constraint: wherein μ bess is the self-discharge rate of the energy storage, η bch , η bdis is the charge-discharge efficiency; Step 4.3: Establishing thermal power unit model: Output power upper and lower limit constraint: wherein P gmin , P gmax represents the minimum output of the thermal power unit, the maximum output limit, represents the output of the thermal power unit at time t; Climbing constraint: In the formula represents the upper and lower limits of the climbing of the thermal power unit, represents the output of the thermal power unit at time t-1; Step 4.4: Constructing system power balance constraint: wherein is the conventional load other than the load of electrolysis of aluminum, are the actual outputs of the photovoltaic and the wind turbine, respectively.

6. The method of claim 1, wherein the method further comprises: The construction of affine form model in step (5) includes the following steps: Step 5.1: Defining load affine form: In the formula is an affine variable corresponding to the load demand power of the time period t; is a center value of the affine variable, respectively; is obtained from the fluctuation range; The value range is [-1, 1]. Step 5.2: All device constraints of the park are converted into affine form constraints, including three types of constraints: equality constraints, inequality constraints and cross-time constraints; for the equality constraints of other decision variables not including state variables, the equality constraints are converted according to the corresponding between the center values on both sides of the equality and the corresponding between the noise elements: wherein is an affine form of a variable for electrolytic aluminum, is an affine form of a variable center value for electrolytic aluminum; is an affine form of a noise element coefficient for electrolytic aluminum; Step 5.3: Inequality constraints are the upper and lower limit constraints of the park model output, and the minimum and maximum values obtained are constrained: Step 5.4: For the cross-time constraints of the park device model, the following conversion is adopted: In the formula: and are the affine forms of the energy equipment output in the industrial park at time period t and time period t-1, respectively; R down and R up are the set lower limit and upper limit of the ramp, respectively; and are the central value, noise element coefficient of and , respectively. Step 5.4: Taking the park operation cost as the target, the storage and electrolytic aluminum high load adjustment power are optimized under the constraint conditions of thermal power unit, wind power, photovoltaic unit, storage and load; the objective function of the model is: minF total = F G + F BESS (34) According to the basic theory of affine algorithm, the affine form target function can be further converted into two parts: operation cost, center value and affine radius; therefore, formula (34) can be further defined as: In the formula, F total,0 , F total,i are the central values of the operation cost of the industrial park and the noise element coefficient of the uncertain factor i, respectively; G,0 , F BESS,0 , F G,i , F BESS,i are the central values of the operation cost of the energy storage and the thermal power unit of the industrial park at the t period and the noise element coefficient of the uncertain factor i, respectively; ω is the optimization weight, and the value range thereof is [0, 1]. Step 5.5: The optimization scheduling model is:

7. The method of claim 1, wherein the method further comprises: The scheduling result output in step (5) includes wind, light output interval, storage charging and discharging plan, electrolytic aluminum load adjustment range and thermal power unit output interval. 8.The method of claim 1-7, wherein, The method realizes dynamic matching with wind and light output through the power-current constraint of electrolytic aluminum load and the charging and discharging power constraint of energy storage device, improves new energy consumption rate and reduces the start-stop cost of thermal power unit.

Images