Inner layer process dynamic optimization method based on RBF coherent method

Abstract

The invention discloses an inner-layer process dynamic optimization method based on an RBF coherent method. The inner-layer process dynamic optimization method comprises the following steps: giving a control variable initial value u0, a state variable initial value x0, optimization time t0 and te and boundary constraints of all variables; performing variable discretization; simulation of a DAE model equation is carried out once in each optimization iteration, and model calculation is carried out on a discrete nonlinear algebraic equation set in a simulation layer by adopting a Newton iteration method to obtain a value of a state variable; during model calculation, obtaining a sensitivity information matrix of a state variable to a control variable and a time variable; and in each optimization iteration process, the previous steps are repeated, and an NLP solver is utilized to solve the nonlinear programming problem processed by the simulation layer until the convergence precision of optimization iteration is achieved, and an optimal solution is found. According to the method, optimal scheduling of the fast climbing ability of the CHP unit is considered, wind power consumption can be promoted on the premise of ensuring the feasibility of a CEHS scheduling scheme, and the operation cost is saved.

Description

Technical Field

[0001] The invention relates to the technical field of optimization scheduling of an electrothermal coupling system, and in particular to a dynamic optimization scheduling method for an electrothermal coupling system taking into account the fast ramping capability constraint of a CHP unit. Background Art

[0002] Due to production process requirements, large industrial enterprises are often equipped with combined heat and power (CHP) power plants to supply all thermal needs and obey the grid dispatch to participate in regional power supply. However, the rigid power output of the enterprise's own CHP units seriously squeezes the space for new energy consumption, hindering the enterprise's response to the carbon quota policy. In recent years, the coupled electric heat system (CEHS) has improved the rapid ramping capability of power output through the flexibility transformation of CHP units, but the characteristics of the transformed units and their CEHS optimization scheduling have become more complex. Therefore, the study of the CEHS dynamic optimization scheduling method considering the rapid ramping capability of CHP units is of great significance to support the operation of the transformed CHP units and improve the flexibility of system scheduling.

[0003] According to the inherent characteristics of CHP units, the optimization control method is used to improve the short-term rapid adjustment capability of the units. It is a low-tech cost and comprehensive advantage solution suitable for the flexibility transformation of self-provided CHP units of enterprises. This solution can significantly change the ramping capability of the unit's power output in a short period of time without affecting the thermal load supply of the enterprise. The precise dynamic modeling of the optimization control is adopted on the differential-algebraic equation (DAE) mechanism model of the CHP unit. The improvement of the unit's ramping rate can compensate for wind power disturbances and adapt to the rapid frequency regulation of the power grid. The rapid ramping capability of the CHP unit can complement the rapid fluctuation of wind power. Making full use of the flexibility of unit scheduling will be conducive to improving the CEHS wind power consumption capacity. However, existing studies rarely apply the precise model of the rapid ramping capability of CHP units to the CEHS optimization scheduling, and deeply analyze its role in wind power consumption.

[0004] At present, the flexibility transformation of CHP units is mainly achieved through direct decoupling solutions such as flexible removal of low-pressure cylinders, installation of heat storage, configuration of electric boilers, and bypass heating of high- and medium-pressure cylinders. Although it can significantly increase the flexibility of the unit, the transformation cost is relatively high. The characteristics of the transformed unit are often modeled using algebraic equation constraints to participate in CEHS scheduling optimization calculations. This modeling based on non-precise algebraic equations cannot describe the dynamic nonlinear characteristics of the CHP unit's ramping process, such as electric-thermal coupling, pressure safety, and heating stability.

[0005] The CEHS optimal scheduling considering the precise dynamic modeling of the rapid ramping capability of CHP units is a dynamic optimization problem with DAE constraints. Its highly nonlinear, non-convex and non-smooth characteristics make it difficult to apply the traditional linear programming direct solution method. Most DAE optimization solutions are to discretize it into a nonlinear programming (NLP) problem, and then use various NLP solvers or intelligent optimization algorithms for optimization and solution. However, the large-scale problems generated after discretization often require special decomposition techniques and complex mathematical processing. When taking into account both solution efficiency and accuracy, the existing CEHS dynamic optimization scheduling problem solutions are faced with challenges in solution strategies and the applicability of discrete methods. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides a dynamic optimization scheduling method for an electric-thermal coupling system taking into account the fast climbing capability constraint of the CHP unit. Under the premise of ensuring the feasibility, safety and reliability of the scheduling scheme, it effectively promotes the maximum consumption of wind power generation, thereby improving the effectiveness of the CHP units with flexibility transformation participating in CEHS scheduling, and has the advantage of low cost.

[0007] The technical solution adopted by the present invention is:

[0008] The dynamic optimization scheduling method of the electric-thermal coupling system considering the fast ramping capability constraint of the CHP unit includes the following steps:

[0009] Step 1: Establish a CHP unit rapid ramping capability constraint model;

[0010] Step 2: Considering the fast ramping capability constraint of CHP units, the CEHS optimization scheduling model of the electric-thermal coupling system is constructed;

[0011] Step 3: Design a two-layer dynamic optimization scheduling strategy for CEHS, and use the quasi-sequential method based on radial basis function to solve the dynamic optimization problem of the inner CHP unit control process; use the dynamic adaptive particle swarm algorithm to generate the steady-state optimization plan for the outer system scheduling.

[0012] In step 1, the CHP unit rapid ramping capability constraint model includes:

[0013] Mathematical model of the dynamic relationship between electric and thermal coupling of CHP units:

[0014]

[0015] Where: r m is the actual amount of coal entering the coal mill, u B is the fuel quantity of the unit, t is the time variable in the control time domain, and ε is the delay time of the pulverizing process; T s is the milling inertia time, r B is the boiler combustion rate; C bis the boiler heat storage coefficient, p d is the drum pressure, K3 is the static parameter of the turbine gain, p t is the pressure before the turbine, u T is the opening of the steam inlet regulating valve of the high-pressure cylinder of the steam turbine, K1 is the static parameter of the fuel gain; K2 is the resistance coefficient of the superheater; T t is the turbine inertia time, P H is the generating power of the unit, K4 is the proportion of high and medium pressure cylinders in the turbine work, K5 is the static parameter of low pressure cylinder gain, p z Extraction steam pressure for heating turbine, u H Adjust the butterfly valve opening for heating extraction steam; C h is the heat storage coefficient of the unit heater, K6 is the effective specific heat capacity of the circulating water in the heat network, τ H is the circulating water return temperature of the unit; m H is the heating extraction steam flow rate, K7 is the effective heat of heating extraction steam converted into steam flow coefficient, m x is the circulating water flow rate of the unit; p1 is the first stage pressure of the turbine.

[0016] Rapid climbing ability optimization control algorithm model:

[0017] Design steam turbine u T (t), boiler u B (t) and heating u H (t) Three variable PID controllers to achieve three key functions: accurate tracking of electric power, rapid recovery of thermal power, and safe and stable system operation;

[0018]

[0019] Where: K PT To control u T (t) Proportional coefficient of the variable, K IT To control u T (t) Integration coefficient of the variable, K DT To control u T (t) differential coefficient of the variable; K PB To control u B (t) Proportional coefficient of the variable, K IB To control u B (t) The integral coefficient of the variable, K DB To control u B (t) differential coefficient of the variable; K PH To control u H (t) Proportional coefficient of the variable, K IH To control u H (t) The integral coefficient of the variable, K DH To control u H(t) differential coefficient of the variable;

[0020] During the control optimization cycle, is the main steam setpoint, is the set value of the unit's power generation. is the set value of the heating extraction steam flow rate; p t (t) is the current value of the main steam pressure control process, P H (t) is the current value of the unit power control process, m H (t) is the current value of the heating extraction steam flow control process.

[0021] Considering the safety and stability of the system, the adjustment process needs to consider the impact of output variable fluctuations to satisfy the path constraint of formula (4); at the same time, the steady-state value of each output variable cannot exceed the allowable error range to satisfy the final value constraint of formula (5).

[0022]

[0023] Where: M pt is the fluctuation range of main steam pressure, M PH is the fluctuation range of electric power, M mH is the fluctuation range of extraction steam flow;

[0024]

[0025] Where: pt is the error range of the main steam pressure, δ PH is the error range of electric power, δ mH is the error range of the extraction steam flow rate; p t (t e ) is the final value of the main steam pressure control process, P H (t e ) is the final value of the unit power control process, m H (t e ) is the final value of the heating extraction steam flow control process.

[0026] Thermal Fluctuation Model of Thermodynamic System:

[0027] The effect of heat changes caused by extraction steam flow fluctuations on heat load demand is calculated as follows:

[0028] Q Load =m L ·c w ·(τ out -τ in )(6);

[0029] Where: Q Load is the heat load demand, m Lis the circulating water flow rate of the heating network, c w is the specific heat capacity of water, τ out Heating network water supply temperature, τ in Heat network return water temperature;

[0030] Q H,equ =m H,equ △h(7);

[0031] Where: Q H,equ is the thermal power of the CHP unit, Δh is the enthalpy drop of the heating extraction steam, m H,equ is the average equivalent flow rate of the regulation process;

[0032]

[0033] Where: t0 and t e are the start and end time points of unit optimization control, m H (t) is the process value of heating extraction steam flow control;

[0034]

[0035] Where: C sys is the heat capacity of the heating area, τ indoor is the temperature of the heating area, dτ indoor / dt is the temperature change rate. In step 2, the CEHS optimization scheduling model of the electrothermal coupling system includes:

[0036] (1) Objective function:

[0037] The goal of optimizing scheduling is to minimize the operating cost of CEHS:

[0038]

[0039] In formula (10), The output of the i-th thermal power unit in the T-th period; is the power of the jth CHP unit in the T period, is the thermal output of the jth CHP in the T period; It is the output of the lth wind turbine in the T period.

[0040] In formula (10), f i G , f l W The function form is as follows:

[0041]

[0042] Among them, f i Gis the operating cost of the i-th thermal power unit, a 2,i 、a 1,i 、a 0,i are the quadratic cost coefficient, linear cost coefficient and constant coefficient of the i-th thermal power unit respectively; is the operating cost of the jth CHP unit, b 2,j 、b 1,j 、b 0,j are the secondary cost coefficient, primary cost coefficient and constant coefficient of the jth CHP unit output, respectively, 2,j 、c 1,j are the secondary cost coefficient and the primary cost coefficient of the thermal output of the jth CHP unit, c 0,j is the cost coefficient of the electric-thermal coupling item of the j-th CHP unit; is the actual power of the first wind turbine in the T period, f l W is the penalty cost of the lth wind turbine unit, θ is the penalty cost coefficient of wind power abandonment;

[0043] (2) Constraints:

[0044] 1) The operation constraints of thermal power units include: unit output constraints and unit climbing constraints;

[0045] Wind turbine operation constraints;

[0046] CEHS electrical power balance and thermal power balance constraints;

[0047] Power system cybersecurity constraints;

[0048]

[0049] in: The output of the i-th thermal power unit in the T-th period; is the minimum output of the i-th thermal power unit in the T period, is the maximum output of the i-th thermal power unit in the T period;

[0050]

[0051] Where: R G,i is the electrical ramp rate of the i-th thermal power unit; The i-th thermal power unit is put on standby in the T-th period, and △T is the system scheduling interval;

[0052]

[0053] in: The i-th thermal power unit in the T period is put into standby mode;

[0054]

[0055] in: The output of the i-th thermal power unit in the T-1 period;

[0056]

[0057] in: The output of the first wind turbine in the T period, is the actual power of the first wind turbine in the T period;

[0058]

[0059] Where: i is the variable of the number of thermal power units, j is the variable of the number of CHP units, l is the variable of the number of wind turbine units, is the electrical load in the T period; Ω G is the set of thermal power units, Ω H is the CHP unit set, Ω W Assemble for wind turbines;

[0060]

[0061] in: is the heat load in period T; is the thermal output of the jth CHP in the T period;

[0062]

[0063] in: is the transmission power of branch s in period T; are the lower and upper bounds of the maximum transmission power of branch s, respectively, and Ω B for all transmission lines in the power system;

[0064] 2): Dynamic constraints on the rapid ramping capability of the CHP unit, written in the standard form of the DAE model:

[0065]

[0066] In formula (20), u is the vector of CHP unit control variables, y is the vector of CHP unit algebraic state variables, x is the vector of CHP unit differential state variables, p is the parameter to be optimized that is independent of t, t0 and t e are the starting and ending time points of the unit optimization control respectively; x L 、x U are vectors consisting of the upper and lower limits of the differential state variables, y L ,y U are vectors consisting of the upper and lower limits of the algebraic state variables, uL 、u U are vectors composed of the upper and lower limits of the control variables, respectively; f is the differential algebraic equation dynamic model of the CHP unit, corresponding to equation (2); g(t) is the path constraint of the CHP unit system, corresponding to equations (3)-(4); h(t e ) is the CHP unit system at time t e The final value constraint corresponds to equation (5).

[0067] In the DAE model:

[0068]

[0069] In formula (21), x1, x2, x3, x4 are the four differential state variables of the CHP unit model, corresponding to p d is the drum pressure, P H is the generating power of the unit, p z Extraction steam pressure for steam turbine heating, r B is the boiler combustion rate; u1, u2, u3 are the three control variables of the CHP unit model, corresponding to u T is the opening of the steam inlet regulating valve of the high-pressure cylinder of the steam turbine, u B is the fuel quantity of the unit, u H is the opening of the butterfly valve for heating extraction steam regulation; y1, y2, y3 are the four algebraic state variables of the CHP unit model, corresponding to p t is the pressure before the turbine, P H is the generating power of the unit, m H is the extraction steam flow for heating;

[0070] 3) Calculate the heat fluctuation and temperature change of the CHP unit, and convert the fluctuation constraint into heat change constraint:

[0071]

[0072] Among them, Δq is the allowable heat change value within the scheduling period (5 minutes); ΔQ is the allowable heat change value within the scheduling cycle (2 hours); and M is the number of scheduling periods of the CEHS system.

[0073] In step 3, the CEHS two-layer dynamic optimization scheduling strategy is designed as follows:

[0074] ①. Outer optimization model:

[0075]

[0076] In formula (24), J1 is the objective function of the scheduling model, X = [X1, X2, X3, X4] is the decision vector composed of the optimized scheduling variables, which are the power output of the thermal power unit, the power output of the wind turbine unit, and the power output and heat output of the CHP unit respectively; d(X) is the constraint of the scheduling optimization decision variable.

[0077] ②Inner layer optimization model:

[0078]

[0079] In formula (25), u is the vector of CHP unit control variables, y is the vector of CHP unit algebraic state variables, x is the vector of CHP unit differential state variables, p is the parameter to be optimized that is independent of t, t0 and t e are the starting and ending time points of the unit optimization control respectively; x0 is the initial value of the differential state variable; x L 、x U are vectors consisting of the upper and lower limits of the differential state variables, y L ,y U are vectors consisting of the upper and lower limits of the algebraic state variables, u L 、u U are vectors consisting of the upper and lower limits of the control variables, respectively; f is the differential algebraic equation of the CHP unit; g(t) is the path constraint of the CHP unit system; h(t e ) is the CHP unit system at time t e The final value constraint. x1, x2, x3, x4 are the four differential state variables of the CHP unit model, corresponding to p d is the drum pressure, P H is the generating power of the unit, p z Extraction steam pressure for steam turbine heating, r B is the boiler combustion rate; Q, R and S are positive definite weight matrices, and the initial values ​​can be determined by normalizing the variables.

[0080] In the formula,

[0081] X3(t)=x2(t) (26);

[0082] Among them, X3 is the power output of the CHP unit determined by the dispatch model, and x2 is the power output of the CHP unit controlled by the CHP unit model optimization;

[0083] X4(t)=△h·(K7K6m x (96x3(t)-τ H +103))(27);

[0084] Among them, X4 is the thermal output of the CHP unit determined by the dispatch model, x3 is the steam extraction pressure of the steam turbine for heating under the optimal control of the CHP unit model, Δh is the enthalpy drop of the heating extraction steam, and m x is the circulating water flow rate of the unit, τ H is the circulating water return temperature of the unit, K6 is the effective specific heat capacity of the circulating water in the heat network, and K7 is the effective heat of the heating extraction steam converted into the steam flow coefficient;

[0085] In step 3, the outer scheduling steady-state optimization algorithm is as follows:

[0086] The dynamic adaptive particle swarm algorithm DAPSO is applied to the CEHS dynamic optimization scheduling problem with DAE constraints. Its mathematical description is:

[0087]

[0088] In formula (28): is the velocity of the i-th particle at the k-th iteration; is the position of the i-th particle at the kth iteration; is the velocity of the ith particle in the k+1th iteration, is the current position of the ith particle in the k+1th iteration, n is the population size; c1 and c2 are acceleration coefficients, generally c1=c2=2; r1 and r2 are two random numbers that vary in the range [0,1]; P i k is the best position searched by the i-th particle in k iterations; is the best position searched by the entire group in k iterations; is the inertia weight of the ith particle at the kth iteration; α and β are selected in [0,1], and α = β = 0.5 is generally taken; a and b are the thresholds for controlling the aggregation factor, and a = 0.9 and b = 0.5 are generally taken; is the evolution speed factor; s1 is the fitness value aggregation factor; s2 is the spatial aggregation factor.

[0089]

[0090] In formula (29): P i k is the best position searched by the i-th particle in k iterations, P i k-1 is the best position searched by the i-th particle in k-1 iterations, F(P i k ) is P i k The fitness value, F(P i k-1 ) is P ik-1 The fitness value, F k is the best fitness value in the kth iteration, is the average fitness value in the kth iteration, N is the population size, L is the longest radius of the search space, D is the dimension of the solution space, g id is the d-th dimension coordinate of the ith particle, is the average value of the d-th dimension coordinates of all particles.

[0091] The steps of the outer scheduling steady-state optimization algorithm based on DAPSO are as follows:

[0092] 1) Set DAPSO parameters according to the CEHS dynamic optimization problem and generate decision variables for the scheduling solution;

[0093] 2) Load the data and decision variables of the CEHS scheduling model Substitute the constraints of the algebraic equations of non-CHP units for verification;

[0094] 3) Decision variables Input the DAE model of the CHP unit to perform dynamic optimization calculation of the process and verify whether the decision variables meet the control characteristic constraints of the unit;

[0095] 4) If the decision variables of the scheduling scheme do not meet any of the constraints in step 2) and step 3), the decision variables are multiplied by a larger penalty coefficient and then the output is used to calculate the objective function value of the scheduling model;

[0096] 5) If the decision variables of the scheduling scheme meet all constraints, the calculated objective function value is output and the minimum value of the comparison function is retained;

[0097] 6) The number of iterations k = k + 1, and the next population scheduling scheme is updated using equations (28) and (29) until the set number of iterations termination condition is met.

[0098] In step 3, the inner process dynamic optimization method based on the RBF quasi-sequential method includes the following steps:

[0099] Step 1), given initial value:

[0100] Given the initial value of the control variable u0, the initial value of the state variable x0, the optimization time t0 and t e and bounds on all variables.

[0101] Step 2), variable discretization:

[0102] The DAE model of discrete formula (20) discretizes each continuous variable along the time axis, so only the univariate numerical approximation method needs to be considered for discretization. Using RBF to implement discretization, the discretized NLP model has a simpler formula description and is easier to implement programming. First, the derivative function shown in formula (20) Make an approximation, and then get the approximation of the original function x(t) by integration.

[0103] For a given CHP unit DAE initial value problem:

[0104] x(t0)=x0(30);

[0105] Among them, x is the differential state variable of the CHP unit, f is the differential algebraic equation of the CHP unit, t0 is the starting time point of the unit optimization control, and x0 is the initial value of the differential state variable of the CHP unit;

[0106] It is known that {(x i ,f i )|i=1,2,...,n}, for any given x, equation (31) can be used for interpolation, and the condition of equation (32) is satisfied:

[0107]

[0108] Among them, i is the real coefficient of the i-th sampling point, φ(·) is a one-dimensional basis function, i is the sampling point, n is the total number of sampling points in the control period, x i is the differential state variable value of the i-th sampling point;

[0109]

[0110] Among them, f k is the differential algebraic equation function value at the kth sampling point, k is the sampling point, x k is the differential state variable value of the kth sampling point;

[0111] The CHP unit optimal control time is discretized into a time series {t0, t1, ..., t n}, the following formula can be used to approximate formula (31).

[0112]

[0113] Among them, i is the real coefficient of the i-th sampling point, φ(·) is a one-dimensional basis function, t i is the time point of the i-th sampling point for optimal control;

[0114] right Performing indefinite integral, we can get:

[0115]

[0116] Among them, i is the real coefficient of the i-th sampling point, φ(·) is a one-dimensional basis function, and C0 is a constant;

[0117] make Then we can get:

[0118]

[0119] Among them, t i is the time point of the i-th sampling point for optimal control, is the function after integration of φ(·);

[0120] The discretization format of the initial value problem is easy to obtain:

[0121]

[0122] Among them, i is the real coefficient of the i-th sampling point, φ(·) is a one-dimensional basis function, is the function of φ(·), C0 is a constant, t i is the time point of the i-th sampling point of the optimization control, j is the sampling point, t j is the time point of the jth sampling point of the optimal control, x0 is the initial value of the differential state variable;

[0123] Solving equation (36) together, we get Substituting the value of C0 into equation (35) yields the variable value at the discrete point: In the calculation, C0 can be eliminated by taking the difference. i )) is denoted as x i , then the simplified formula (36) is:

[0124]

[0125] Among them, k is the real coefficient of the kth sampling point, t i is the time point of the i-th sampling point for optimal control; t j is the time point of the jth sampling point of the optimal control; t j-1 is the time point of the j-1th sampling point of the optimized control; t k is the time point of the kth sampling point of the optimization control, φ(·) is a one-dimensional basis function, is the function after integration of φ(·);

[0126] The basis function φ(·) is often in the form of RBF, among which the inverse Multi-Quadric function is most suitable for solving DAE. Let c = γt e / n, for most DAE optimization models, when γ = 2, the solution process can be easily converged. That is:

[0127]

[0128] Among them, c is a given positive number, r is an intermediate variable, and φ(·) is a one-variable basis function;

[0129] Taking the indefinite integral of φ(·) we get:

[0130]

[0131] Where c is a given positive number, t i is the time point of the i-th sampling point for optimal control; is the function after integration of φ(·);

[0132] The dynamic optimization problem shown in (25) is that all variables are time t∈[t0,t e ]. Without loss of generality, the present invention assumes that the optimized DAE has continuous second-order partial derivatives for each variable. Only the discretization of equation (25) is considered, and the optimization control period [t0,t e ] is divided equally, and the time series can be obtained And there is t n =t e , t i -t i-1 =t e / n. i The corresponding variables are denoted as x i ,y i and u i , discretize equation (25) according to equation (37), and we can get the following NLP model:

[0133]

[0134] Where: s is the variable dimension of the vector, u i is the control variable of the i-th sampling point, y i is the algebraic state variable of the i-th sampling point, x s,i is the s-dimensional differential state variable of the i-th sampling point, p is the parameter to be optimized that is independent of t, x L 、x U are the upper and lower limits of the differential state variable, y L ,y U are the upper and lower limits of the algebraic state variables, u L 、u U are the upper and lower limits of the control variables, respectively; f is the dynamic model of differential algebraic equations; g(.) is the path constraint; h(.) is the final value constraint. i is the time point of the i-th sampling point for optimal control; t j is the time point of the jth sampling point of the optimal control; t j-1 is the time point of the j-1th sampling point of the optimized control; tk is the time point of the kth sampling point for optimal control; x s,i is the s-dimensional differential state variable of the ith sampling point, ζ s,k is the real coefficient of the s-dimensional variable at the k-th sampling point, is the function of the integral of the one-dimensional basis function φ(·); x s,0 is the s-th dimension differential state variable value of the initial sampling point; x s,n u is the s-th dimension differential state variable value of the termination sampling point; n is the control variable value of the termination sampling point, y n is the algebraic state variable value of the termination sampling point; x s,j is the s-dimensional differential state variable at the j-th sampling point, x s,j-1 is the s-dimensional differential state variable of the j-1th sampling point;

[0135] Step 3), model calculation:

[0136] In each optimization iteration, a simulation of the DAE model equation is performed. In the simulation layer, the Newton iteration method is used to calculate the model of the discretized nonlinear algebraic equation group to obtain the value of the state variable.

[0137] Step 4), sensitivity calculation:

[0138] While the model is being calculated, the sensitivity information matrix of the state variables to the control variables and time variables is obtained.

[0139] Step 5), optimization layer calculation:

[0140] In each optimization iteration, steps 4) and 5) are repeated, and the existing NLP solver is used to solve the nonlinear programming problem after the simulation layer processing until the convergence accuracy of the optimization iteration is reached and the optimal solution is found.

[0141] The present invention provides a dynamic optimization scheduling method for an electric-thermal coupling system considering the fast ramping capability constraint of a CHP unit, and the technical effects are as follows:

[0142] 1) Numerical simulation of the fast ramping dynamic model of the CHP unit with optimized control is carried out. Compared with the static model, the dynamic model has a faster ramping rate and a larger ramping capacity;

[0143] 2) From the perspective of improving the wind power absorption capacity of the comprehensive energy system, the role of the rapid ramping capability of the CHP unit in the wind power absorption in the CEHS optimization dispatch is fully considered, and the current situation that the rigid power output of the CHP unit limited by the electric-thermal coupling characteristics seriously squeezes the space for new energy absorption is improved;

[0144] 3) Under the premise of ensuring the feasibility, safety and reliability of the dispatching plan, effectively promote the maximum consumption of wind power generation, thereby improving the effectiveness of CHP units with flexibility transformation in participating in CEHS dispatching.

[0145] 4) The method of the present invention takes into account the optimal scheduling of the rapid ramping capability of the CHP unit, which can promote wind power consumption and save operating costs while ensuring the feasibility of the CEHS scheduling scheme. BRIEF DESCRIPTION OF THE DRAWINGS

[0146] Figure 1 This is the structural diagram of CEHS including self-contained CHP unit.

[0147] Figure 2 This is the structure diagram of the dynamic model of the CHP unit's rapid ramping capability.

[0148] Figure 3 (1) is a comparison of the CHP unit power output of the static model and the dynamic model;

[0149] FIG3(2) is a comparison diagram of the main steam pressure of the CHP unit of the static model and the dynamic model;

[0150] Figure 3 (3) is a comparison diagram of the extraction steam flow of the CHP unit between the static model and the dynamic model.

[0151] Figure 4 Solve the flow chart for the CEHS dynamic optimization scheduling model.

[0152] Figure 5 It is the structure diagram of the quasi-sequential algorithm;

[0153] Figure 6 This is the CEHS structure diagram.

[0154] Figure 7 Graphs showing optimization scheduling results under different examples.

[0155] Figure 8 The distribution diagram of CHP unit operation points under different examples;

[0156] Fig. 9 The electrical output diagram of CHP unit under different calculation examples;

[0157] Fig.10 This is the main steam pressure curve of the CHP unit.

[0158] Fig.11 The extraction steam flow diagram of CHP unit under different calculation examples. DETAILED DESCRIPTION

[0159] A CEHS dynamic optimization scheduling method considering the rapid ramping capability of CHP units. First, a CHP unit mechanism model is constructed, an optimization control strategy is proposed, and the impact of variable load disturbances on the unit operating characteristics is analyzed. Then, dynamic constraints on the rapid ramping capability of the CHP unit are formed, and on this basis, a CEHS dynamic optimization scheduling model with differential algebraic equation constraints is established. Finally, a two-layer optimization structure combining the outer layer DAPSO-based scheduling steady-state optimization and the inner layer radial basis quasi-sequential process dynamic optimization is used to solve the dynamic problem. The specific implementation process is as follows:

[0160] Step 1: Establish the CHP unit rapid ramping capability constraint model:

[0161] The CEHS studied consists of three parts: thermal system, power system, and CHP unit energy supply and control system. The structure is as follows: Figure 1 As shown. B Control the boiler combustion to produce high-temperature steam, by adjusting the butterfly valve u T Control the steam entering the high-pressure cylinder; install a steam extraction regulating butterfly valve in the connecting pipe between the medium-pressure cylinder and the low-pressure cylinder H , control the extraction steam to supply the heat source to output thermal power, and control the steam to enter the low-pressure cylinder to do work and output electrical power. The heat source exchanges heat with the first heating station, and supplies the internal heat load of the industrial enterprise through the heat exchanger. Due to the small area of ​​the enterprise, the spatial distribution of the heat pipe network is relatively limited, and the impact of the pipeline heat transfer process can be ignored. At the same time, the CHP unit bears part of the electricity load of the production and self-use electricity of the self-contained power plant enterprise and the production and life of the power supply radiation area. The insufficient power supply is met by the large network dispatching wind turbines and conventional thermal power units.

[0162] In the traditional power grid CEHS optimization dispatch, CHP units often use a simplified algebraic equation constraint model (also known as a static model):

[0163]

[0164] Among them, T is the scheduling period; △T is the ramp time; It is the upper spare capacity and lower spare capacity of the unit; The maximum and minimum power output allowed by the unit during the T period; The upward and downward climbing rate of the CHP unit is generally 2%P e / min, P e The rated power of the unit, after the optimization of the control system transformation, the CHP unit climbing rate can be increased to 4%P e / min or above.

[0165] The static model shown in formula (1) cannot characterize the nonlinear characteristics of the changes in variables such as pressure, flow, and temperature during the CHP unit regulation process. When the static model is used to consider the unit's rapid ramping capability, the unit's power output ramping can achieve similar effects to the dynamic model, but the static model cannot constrain the unit's short-term ramping shortage, safety, and stability caused by changes in variables such as unit pressure and flow. If the ramping constraint of a single CHP unit is too loose, it will make it difficult to implement actual operation regulation or cause safety accidents.

[0166] The fast ramping capability constraint of the extraction steam CHP unit is established by respectively establishing the differential algebraic equation of the electric-thermal coupling dynamic relationship of the CHP unit, the fast ramping capability optimization control algorithm model, and the thermal fluctuation model of the thermal system.

[0167] (1) Mathematical model of the dynamic relationship between electric and thermal coupling of CHP units:

[0168]

[0169] Where: u B is the fuel quantity of the unit; u T u is the opening of the steam inlet regulating valve of the high-pressure cylinder of the steam turbine; H Adjust the butterfly valve opening for heating extraction steam; m x is the circulating water flow rate of the unit; τ H is the circulating water return temperature of the unit; p t is the pressure before the turbine; P H is the generating power of the unit; p z Extraction steam pressure for heating turbine; m H is the heating extraction steam flow; p1 is the first stage pressure of the steam turbine; r m is the actual amount of coal entering the coal mill; r B is the boiler combustion rate. K1, K2, K3, K4, K5, K6, K7 are static parameters; ε is the delay time of the milling process, T s Milling inertia time, C b Boiler heat storage coefficient, T t Turbine inertia time, C h Heat storage coefficient of unit heater.

[0170] (2) Rapid climbing capability optimization control algorithm model:

[0171] According to the above model, the optimization control method is used to achieve the rapid ramping capacity adjustment of the CHP unit and enhance the short-term flexibility of the unit. T (t), boiler u B (t) and heating u H (t) Three variable PID controllers realize three key functions: accurate tracking of electric power, rapid recovery of thermal power and safe and stable system operation:

[0172]

[0173] Among them, K PT , K IT , K DT , K PB , K IB , K DB , K PH , K IH , K DH for u T (t),u B (t),u H (t)PID parameters. During the control optimization cycle, is the main steam setpoint, is the set value of the unit's power generation. Set the extraction steam flow rate for heating.

[0174] like Figure 2 As shown, during the period T, the main steam pressure of the unit Extraction steam flow is the fixed value input, with the unit power output in adjacent dispatching periods The step signal is formed as the variable load disturbance input, and the dynamic constraint model of rapid ramping capability is constructed together with the CHP unit model. Among them, the main steam pressure is set according to the unit's electrical and thermal output range, and the relationship between the unit's thermal output and the extraction steam flow rate is shown in formula (7).

[0175] Considering the safety and stability of the system, the adjustment process needs to consider the impact of output variable fluctuations and should meet the following path constraints:

[0176]

[0177] At the same time, the steady-state value of each output variable cannot exceed the allowable error range and must also meet the following final value constraints:

[0178]

[0179] Among them, M pt The fluctuation range of main steam pressure is ±5% of the set value; M PH is the fluctuation range of electric power, taking 1MW; M mH is the fluctuation range of extraction steam flow, which is 50t / h; δ pt The error range of the main steam pressure is ±2.5% of the set value, δ PH is the error range of electric power, taking 0.1MW; δ mH The error range of extraction steam flow is 5t / h.

[0180] (3) Thermal fluctuation model of thermal system:

[0181] The dynamic model of the CHP unit's rapid ramping capability is modeled and simulated. Given a step signal of variable load disturbance, the change process of the main steam pressure, extraction steam flow rate, and unit electric power within the optimization control cycle is analyzed. The impact of the heat change caused by the extraction steam flow fluctuation on the heat load demand is calculated as follows:

[0182]

[0183] Among them, Q Load is the heat load demand, m L is the circulating water flow rate of the heating network, c w is the specific heat capacity of water, τ out Heating network water supply temperature, τ in Heat network return water temperature; Q H,equ is the thermal power of the CHP unit, △h is the enthalpy drop of the heating extraction steam, m H (t) is the flow regulation process value, m H,equ is the average equivalent flow rate of the regulation process, t0 and t e The starting and ending time points of the unit optimization control; C sys is the heat capacity of the heating area, τ indoor is the temperature of the heating area, dτ indoor / dt is the rate of temperature change.

[0184] The static model without considering the rapid climbing ability is adopted, and the climbing rate is 2%P e / min. Static model considering fast climbing capability (climbing rate 4.5%P e / min) and dynamic models are used to simulate the CHP units. Figure 3(1) to Figure 3(3) As shown, the operating point For example, the variable load capacity of the three modeling methods can reach The results show that the CHP unit has a faster ramp rate and a larger ramp capacity when considering the rapid ramp capability. The main steam pressure and extraction steam flow of the dynamic model return to a steady state after fluctuations. The maximum fluctuation of the main steam pressure is 15.73, which is within the ±5% fluctuation range of the set value of 15, which can ensure the safe operation of the unit. The equivalent average flow rate within 300s is 393.19t / h, which is about 4471.33kJ less than the set value of 400t / h, which is converted into a temperature change of 4.21×10 -3 ℃, proving that the temperature change caused by steam flow fluctuation has little effect on the heat load. A step disturbance was formed and input into the dynamic model of the CHP unit for inspection. The data showed that the main steam pressure peak of 15.82 exceeded the maximum upper limit, there were safety hazards in the operation of the unit, and the steam flow fluctuation also increased, which was not conducive to the stable heating of the unit.

[0185] Step 2: Construct a CEHS optimal scheduling model considering the fast ramping capability constraints of CHP units:

[0186] (1) Objective function:

[0187] The optimization scheduling of the present invention aims to minimize the operating cost of CEHS:

[0188]

[0189] In the formula, f i G , f l W The function form is as follows:

[0190]

[0191] Among them, f G is the operating cost of the thermal power unit, cost coefficients a2, a1, a0; P G Output for thermal power units; f H is the operating cost of the CHP unit, power generation and heating cost coefficients b2, b1, b0, c2, c1, c0; P H , Q H is the electrical and thermal output of the CHP unit; f W is the penalty cost of wind power, cost coefficient θ; P W To generate power for wind turbines, is the actual wind power; Ω G ,Ω H ,Ω W are respectively the thermal power unit, CHP unit and wind power unit. The decision variables are The scheduling cycle is 2 hours, the interval △T is 5 minutes, and the scheduling period T∈M (M=1,2,…,24).

[0192] (2) Constraints:

[0193] 1) Thermal power unit operation constraints, including unit output constraints and unit ramp constraints; wind turbine unit operation constraints; CEHS electric power balance and thermal power balance constraints; power system network security constraints.

[0194]

[0195] Among them, R G is the electrical ramp rate of the thermal power unit; They are the minimum and maximum output values ​​of thermal power units respectively; For upper and lower standby of thermal power units; P Load Electric load demand; Q LoadHeat load demand; is the transmission power of branch s in period T; are the lower and upper bounds of the maximum transmission power of branch s, respectively, and Ω B For all transmission lines in the power system.

[0196] 2) Dynamic constraints on the rapid ramping capability of the CHP unit, as shown in equations (2)-(5). Written in the standard form of the DAE model:

[0197]

[0198] In the model, u is the control variable, y is the algebraic state variable, and x is the differential state variable, where:

[0199]

[0200] Where: f is the differential algebraic equation dynamic model of the CHP unit, corresponding to equation (2); g(t) is the system path constraint, corresponding to equations (3)-(4); h(t e ) is the system at time t e The final value constraint corresponds to equation (5). Inequality constraints can be converted into equality constraints by adding slack variables greater than zero and added to the system.

[0201] 3) The thermal system calculates the heat fluctuation and temperature change of the CHP unit through equations (6)-(9). According to the production process and the heating demand for life, the temperature change within the scheduling period (5 minutes) is less than 0.5°C, and the temperature change within 2 hours is less than 1-2°C, which can be converted into the heat change constraint by equation (9).

[0202]

[0203] Among them, △q is the allowable heat change value within the scheduling period (5 minutes); △Q is the allowable heat change value within the scheduling cycle (2 hours).

[0204] Step 3: Design a two-layer dynamic optimization scheduling strategy for the electric-thermal coupling system for hierarchical solution. Use the quasi-sequential method based on radial basis function to solve the dynamic optimization problem of the inner CHP unit control process, and use the dynamic adaptive particle swarm algorithm to generate the outer system scheduling steady-state optimization plan:

[0205] 1) Outer optimization model:

[0206]

[0207] Where: J1 is the objective function of the dispatch model, corresponding to equations (10)-(11); X = [X1, X2, X3, X4] is the decision vector composed of the optimized dispatch variables, which are the power output of thermal power and wind turbine units and the power and heat output of CHP units respectively; d(X) is the constraint of the dispatch optimization decision variable, corresponding to equations (12)-(23).

[0208] 2) Inner layer optimization model

[0209]

[0210] In the formula,

[0211] X3(t)=x2(t) (26);

[0212] X4(t)=△h·(K7K6m x (96x3(t)-τ H +103)) (27);

[0213] Among them, u * ,y * are the set values ​​of the control variables and output variables obtained by steady-state optimization, u * It is also the feedforward setting for the variable load dynamic process. u(t) is the control variable at the tth moment, y(t) is the measured output of the control object at the t-1th moment, and serves as the feedback of the closed-loop system at the tth moment. Q, R, and S are positive definite weight matrices, and the initial values ​​can be determined by normalizing the variables.

[0214] The objective function J1 dispatches decision variables to achieve the CEHS economic goal, and uses intelligent algorithms to give scheduling plans and iterative optimization; the objective function J2 guides the control variables and output variables to approach the desired steady-state values, so that the unit control system accelerates the convergence speed and makes the manipulated variables as smooth as possible, and the quasi-sequential method is used for solution.

[0215] (1) Outer scheduling steady-state optimization algorithm:

[0216] Considering that the discretized CEHS optimization scheduling is a complex nonlinear dynamic optimization problem, the particle swarm algorithm that can efficiently handle multiple constraint optimization can be used as the optimization algorithm for the optimization calculation layer. Compared with other optimization algorithms, the dynamic adaptive particle swarm algorithm (DAPSO) shows relatively excellent convergence ability and solution accuracy in the CEHS dynamic optimization scheduling problem with DAE constraints in the present invention. Its mathematical description is:

[0217]

[0218] Where: is the velocity of the i-th particle at the k-th iteration; is the position of the ith particle at the kth iteration; c1 and c2 are acceleration coefficients, generally c1=c2=2; r1 and r2 are two random numbers in the range [0, 1]; P i k is the best position searched by the i-th particle in k iterations; is the best position searched by the entire group in k iterations; is the inertia weight of the ith particle at the kth iteration; α and β are selected in [0, 1], and α = β = 0.5 is generally taken; a and b are the thresholds for controlling the aggregation factor, and a = 0.9 and b = 0.5 are generally taken; is the evolution speed factor; s1 is the fitness value aggregation factor; s2 is the spatial aggregation factor.

[0219]

[0220] Where: F(P i k ) is P i k The fitness value of F k is the best fitness value in the kth iteration; is the average fitness value in the kth iteration; N is the population size; L is the longest radius of the search space; D is the dimension of the solution space; g id is the d-th dimension coordinate of the ith particle; is the average value of the d-th dimension coordinates of all particles.

[0221] CEHS two-layer dynamic optimization scheduling strategy process is as follows Figure 4 As shown in the figure, the steps of the outer scheduling steady-state optimization algorithm based on DAPSO are:

[0222] 1) Set DAPSO parameters according to the CEHS dynamic optimization problem and generate decision variables for the scheduling solution;

[0223] 2) Load the data and decision variables of the CEHS scheduling model Substitute the constraints of the algebraic equations of non-CHP units for verification;

[0224] 3) Decision variables Input the DAE model of the CHP unit to perform dynamic optimization calculation of the process and verify whether the decision variables meet the control characteristic constraints of the unit;

[0225] 4) If the decision variables of the scheduling scheme do not meet any of the constraints in step 2) and step 3), the decision variables are multiplied by a larger penalty coefficient and then the output is used to calculate the objective function value of the scheduling model;

[0226] 5) If the decision variables of the scheduling scheme meet all constraints, the calculated objective function value is output and the minimum value of the comparison function is retained;

[0227] 6) The number of iterations k = k + 1, and the next population scheduling scheme is updated using equations (28) and (29) until the set number of iterations termination condition is met.

[0228] (2) Inner process dynamic optimization method:

[0229] The quasi-sequential algorithm combines the advantages of the sequential algorithm and the simultaneous algorithm. It is suitable for solving large-scale strongly nonlinear dynamic optimization problems with path constraints. It is divided into a simulation calculation layer and an optimization calculation layer. In the simulation calculation layer, the control variable is represented as a constant function or linear function in each time unit. The state variable is discretized using the orthogonal collocation method based on the radial basis function (RBF), thereby discretizing the nonlinear DAEs into a set of nonlinear algebraic equations, which can be solved by the Newton iteration method to obtain the state variable values ​​at the discrete points. At the same time, according to the derivative formula of the state variable with respect to the control variable at the discrete point, the sensitivity matrix can be calculated for the optimization calculation layer to obtain the gradient information of the objective function with respect to the control variable. The piecewise function value of the control variable becomes the decision variable of the optimization problem. The constraints no longer contain DAEs equality constraints. The state variable only appears in the inequality constraints in the form of path constraints. The original problem is simplified to an NLP problem, which can be directly solved by the sequential quadratic programming algorithm. The structure of the quasi-sequential algorithm is as follows: Figure 5 shown.

[0230] The steps of the inner process dynamic optimization method based on the RBF quasi-sequential method are as follows:

[0231] 1) Given the initial value:

[0232] Given the initial value of the control variable u0, the initial value of the state variable x0, the optimization time t0 and t e and bounds on all variables.

[0233] 2) Discrete variables:

[0234] The DAE model of discrete formula (20) discretizes each continuous variable along the time axis, so only the univariate numerical approximation method needs to be considered for discretization. Using RBF to implement discretization, the discretized NLP model has a simpler formula description and is easier to implement programming. First, the derivative function shown in formula (20) Make an approximation, and then get the approximation of the original function x(t) by integration.

[0235] For a given CHP unit DAE initial value problem:

[0236] x(t0)=x0(30);

[0237] It is known that {(x i ,f i )|i=1,2,…,n}, for any given x, equation (30) can be used for interpolation, and the condition of equation (31) is satisfied:

[0238]

[0239] in, is a real coefficient, φ(·) is a univariate function (basis function), is the center of the RBF interpolation.

[0240] The CHP unit optimal control time is discretized into a time series {t0, t1, …, t n}, (30) can be approximated by the following formula.

[0241]

[0242] right Performing indefinite integral, we can get

[0243]

[0244] make Then we can get:

[0245]

[0246] The discretization format of the initial value problem is easy to obtain:

[0247]

[0248] Solving equation (36) together, we get Substituting the value of C0 into equation (35) yields the variable value at the discrete point: In the calculation, C0 can be eliminated by taking the difference. i )) is denoted as x i , then the simplified formula (36) is:

[0249]

[0250] The basis function φ(·) is usually in the form of RBF, where the inverse Multi-Quadric function is most suitable for solving DAE. c is a given positive number, and c = γt e / n, for most DAE optimization models, when γ = 2, the solution process can be easily converged.

[0251]

[0252] Taking the indefinite integral of φ(·) we get:

[0253]

[0254] The dynamic optimization problem shown in (25) is that all variables are time t∈[t0,t e ]. Without loss of generality, the present invention assumes that the optimized DAE has continuous second-order partial derivatives for each variable. Only the discretization of equation (25) is considered, and the optimization control period [t0,t e ] is divided equally, and the time series can be obtained And there is t n =t e , t i -t i-1 =t e / n. i The corresponding variables are denoted as x i ,y i and u i , discretize equation (25) according to equation (37), and we can get the following NLP model:

[0255]

[0256] 3) Model calculation:

[0257] In each optimization iteration, a simulation of the DAE model equation is performed. In the simulation layer, the Newton iteration method is used to calculate the model of the discretized nonlinear algebraic equation group to obtain the value of the state variable.

[0258] 4) Sensitivity calculation:

[0259] While the model is being calculated, the sensitivity information matrix of the state variables to the control variables and time variables is obtained.

[0260] 5) Optimization layer calculation:

[0261] In each optimization iteration, steps 4) and 5) are repeated, and the existing NLP solver is used to solve the nonlinear programming problem after the simulation layer processing until the convergence accuracy of the optimization iteration is reached and the optimal solution is found. Specific embodiment:

[0263] The present invention adopts the improved IEEE30 node system to construct the CEHS model, and its structure is as follows: Figure 6 As shown, nodes 1, 2, 8, and 13 are thermal power units, 5 is a CHP unit, and 150MW of wind power is connected at node 11. The penalty cost for wind curtailment is 850 yuan / MWh.

[0264] The static modeling and dynamic modeling of CHP units are used to participate in the CEHS optimization scheduling calculation, and a comparative analysis of the examples is carried out.

[0265] Example 1: Without considering the rapid ramping capability of the CHP unit, the static model of formula (1) is used, and the unit power ramping is 2%P e / min, the feasible domain is modeled using algebraic equations;

[0266] Example 2: CHP dynamic model, considering the electrothermal coupling characteristics of the CHP unit optimization control, which is modeled using differential algebraic equations as shown in Equation (20).

[0267] Under the above calculation conditions, the method of the present invention is applied to perform scheduling optimization, and the results are as follows:

[0268] In order to compare the impact of the rapid ramping capability of CHP units on the dispatching scheme, two different examples are used for dispatching optimization. The output of each unit is as follows: Figure 7 As shown. Compared with Example 1, in the dispatching scheme of Example 2, the thermal power output is basically the same in each period, the CHP unit output is significantly reduced in multiple periods, and the wind power output is significantly increased in more periods. Overall, the optimized dispatching of Example 2 reduces the power output of the CHP unit by 4.3% and increases the wind power output by 9.79%. This is because the rapid downward climbing ability of the CHP unit in multiple periods makes its own power output drop even lower, thus freeing up more space for absorbing wind power.

[0269] Table 1 compares the two examples from the perspective of CEHS operating costs. It can be seen that the dispatch cost of Example 2 is reduced by 19,681 yuan, mainly due to the increase in wind power consumption, a significant reduction in wind abandonment costs of 15,538 yuan, and a slight reduction in CHP and thermal power output costs, thereby reducing the total cost. The simulation results show that the dispatch scheme that considers the dynamic modeling of the rapid ramping capability of CHP units can increase wind power consumption and reduce thermal power and CHP power consumption.

[0270] Table 1 Scheduling scheme costs under different examples

[0271]

[0272] Figure 8 The distribution of CHP unit operating points in 24 time periods under different examples is intuitively displayed. It can be seen that the operating points in Example 1 are more concentrated and far away from the upper and lower boundaries of the feasible domain; while Example 2 has more operating points close to the lower boundary and a larger distribution range. For example: the operating points of the CHP unit in Example 1 in 5 periods are (259.03, 181.58), while those in Example 2 are (259.03, 155.42). Compared with Example 1, Example 2 has 10 operating points closer to the lower boundary and 5 points closer to the upper boundary. This is because the rapid dynamic adjustment of the CHP unit makes the unit's climbing ability stronger and the unit's electric power optimization space larger, thereby verifying that the unit has better operating flexibility.

[0273] Example 3: Considering the rapid ramping capability of the CHP unit, the static model (1) is adopted, the ramping rate is 4.5% Pe / min, and the feasible domain is modeled by algebraic equations.

[0274] Compared with Example 2, the dispatching cost and the output of each unit in Example 3 are basically the same (Example 2, 242,422 yuan; Example 3, 241,411 yuan). In order to verify the influence of the CHP unit model on the accuracy of power output tracking, system safety, and heating stability under the dynamic and static model dispatching schemes considering the rapid ramping capability, the dispatching cost of each unit in Example 3 is calculated. A step disturbance is input into the dynamic model of the CHP unit for verification, and the results are compared with the dynamic model of Example 2.

[0275] Example 2: The power output dispatch value of the CHP unit and the unit optimization control process value, such as Fig. 9 As shown. In the T11 period, the unit power output was adjusted from 230.35MW at 10 o'clock to 217.16MW at 11 o'clock, the adjustment time was about 100s, and the average shortage in the period was 0.076MW / s; in the T12 period, the load changed from 217.16MW at 11 o'clock to 157.16MW at 12 o'clock, the adjustment time was about 150s, and the average shortage was 0.27MW / s. There is a certain climbing shortage between the process value and the dispatch value in the unit optimization control cycle (a dispatch period), which is due to the adjustment time required for the valve action response in the optimization control process. However, the control adjustment speed is fast, and the average shortage is small, which can be compensated by fine-tuning the output of other units, and has little effect on the feasibility of the dispatch scheme. Although the modeling method of Example 3 can achieve a similar climbing ability effect as the dynamic model, it cannot consider the dynamic characteristics of the above-mentioned unit adjustment process, which will lead to actual unit adjustment mismatch and reduce the feasibility of the dispatch scheme.

[0276] During the execution of the dispatching plan, the change in main steam pressure during rapid adjustment of the CHP unit is related to the safe and stable operation of the unit. Fig.10 As shown, when the main steam pressure setting value is 15.00MPa, the maximum upward peak fluctuation is 15.74MPa, and the maximum downward peak fluctuation is 14.39MPa, which does not exceed the fluctuation range of the set value, and the steady state reaches the set value. The main steam pressure setting value for time periods 11, 17, 18, and 24 is 15.50MPa, and the maximum upward and downward peak fluctuations do not exceed the fluctuation range of the set value, and the steady state reaches the set value. Simulation data proves that the pressure fluctuations caused by the rapid dynamic adjustment process of the CHP unit have no effect on the safety of the unit. The modeling method of Example 3 cannot take into account the dynamic adjustment process constraints of the main steam pressure. Excessive variable load adjustment will cause the pressure of the unit adjustment process to exceed the upper limit of the safety constraint. As Fig.10 The time periods of 2, 4, 10, 16 and 23 will affect the safe and stable operation of the units.

[0277] During the dynamic process of CHP unit rapid ramping capability, such as Fig.11 As shown in the figure, the steady-state values ​​of each period reach the set value, and the peak fluctuation of the regulation process does not exceed the deviation range of the set value. The static model flow scheduling value is obtained by the heat output of the unit that meets the heat load demand. The steady-state value of the dynamic process in the figure coincides with it, and the upward and downward fluctuations are calculated. The average flow equivalent value is shown in the red curve. Compared with Example 2, Example 3 has a larger peak fluctuation in the regulation process in multiple periods, and the average flow equivalent value in the period deviates further from the demand steady-state value.

[0278] The heat and temperature changes caused by the thermal output fluctuations of the CHP units in each dispatching period are shown in Table 2, where m H,equ is the equivalent average extraction steam flow rate of the CHP unit (t / h), △q is the heat change (kJ), △τ is the temperature change in 300s (×10 -3 ℃). The data in Table 2 show the temperature changes caused by the increase or decrease of heat in each period of Example 2. Overall, the CHP unit generates an average of 3845.21 kJ more heat during the scheduling period (2 h). According to equations (6) and (9), this heat causes the average increase of 1.07℃ in the circulating water temperature of the heating network, causing the temperature of the heating area to increase by 3.62×10 -3 The data show that the heat fluctuations generated by the dynamic adjustment process of the CHP unit have no effect on the heating temperature demand.

[0279] Table 2 Effect of extraction steam flow fluctuation on unit thermal output and temperature

[0280]

Claims

1. The inner process dynamic optimization method based on RBF quasi-sequential method is characterized by The steps include: Step 1: Given the initial value: Given the initial value of the control variable u0, the initial value of the state variable x0, the optimization time t0 and t e and bounds constraints on all variables; Step 2: Discrete variables: For a given CHP unit DAE initial value problem: Among them, x is the differential state variable of the CHP unit, f is the differential algebraic equation of the CHP unit, t0 is the starting time point of the unit optimization control, and x0 is the initial value of the differential state variable of the CHP unit; It is known that {(x i ,f i )|i=1,2,...,n}, for any given x, equation (31) can be used for interpolation, and the condition of equation (32) is satisfied: Among them, i is the real coefficient of the i-th sampling point, φ(·) is a one-dimensional basis function, i is the sampling point, n is the total number of sampling points in the control period, x i is the differential state variable value of the i-th sampling point; Among them, f k is the differential algebraic equation function value at the kth sampling point, k is the sampling point, x k is the differential state variable value of the kth sampling point; The CHP unit optimal control time is discretized into a time series {t0, t1, ..., t n }, use the following formula to approximate formula (31); Among them, i is the real coefficient of the i-th sampling point, φ(·) is a one-dimensional basis function, t i is the time point of the i-th sampling point for optimal control; right Performing indefinite integral, we can get: Among them, i is the real coefficient of the i-th sampling point, φ(·) is a one-dimensional basis function, and C0 is a constant; make Then we can get: Among them, t i is the time point of the i-th sampling point for optimal control, is the function after integration of φ(·); The discrete format of the initial value problem is easy to obtain: Among them, i is the real coefficient of the i-th sampling point, φ(·) is a one-dimensional basis function, is the function of φ(·), C0 is a constant, t i is the time point of the i-th sampling point of the optimization control, j is the sampling point, t j is the time point of the jth sampling point of the optimal control, x0 is the initial value of the differential state variable; Solving equation (36) together, we get Substituting the value of C0 into equation (35) yields the variable value at the discrete point: In the calculation, C0 can be eliminated by taking the difference; (x(t i )) is denoted as x i , then the simplified formula (36) is: Among them, k is the real coefficient of the kth sampling point, t i is the time point of the i-th sampling point for optimal control; t j is the time point of the jth sampling point of the optimal control; t j-1 is the time point of the j-1th sampling point of the optimized control; t k is the time point of the kth sampling point of the optimization control, φ(·) is a one-dimensional basis function, is the function after integration of φ(·); The basis function φ(·) is often in the form of RBF, among which the inverse Multi-Quadric function is most suitable for solving DAE. Let c = γt e / n, for most DAE optimization models, when γ = 2, the solution process can be easily converged; that is: Among them, c is a given positive number, r is an intermediate variable, and φ(·) is a one-variable basis function; Taking the indefinite integral of φ(·) we get: Where c is a given positive number, t i is the time point of the i-th sampling point for optimal control; is the function after integration of φ(·); The dynamic optimization problem shown in (25) is that all variables are time t∈[t0,t e ]; without loss of generality, assume that the optimized DAE has continuous second-order partial derivatives for each variable; only consider the discretization of equation (25), for the optimization control period [t0,t e ] is divided equally, and the time series can be obtained And there is t n =t e , t i -t i-1 =t e / n; each time point t i The corresponding variables are denoted as x i ,y i and u i , discretize equation (25) according to equation (37), and get the following NLP model: Where: s is the variable dimension of the vector, u i is the control variable of the i-th sampling point, y i is the algebraic state variable of the i-th sampling point, x s,i is the s-dimensional differential state variable of the i-th sampling point, p is the parameter to be optimized that is independent of t, x L 、x U are the upper and lower limits of the differential state variable, y L ,y U are the upper and lower limits of the algebraic state variables, u L 、u U are the upper and lower limits of the control variables, respectively; f is the dynamic model of differential algebraic equations; g(.) is the path constraint; h(.) is the final value constraint; t i is the time point of the i-th sampling point for optimal control; t j is the time point of the jth sampling point of the optimal control; t j-1 is the time point of the j-1th sampling point of the optimized control; t k is the time point of the kth sampling point for optimal control; x s,i is the s-dimensional differential state variable of the ith sampling point, ζ s,k is the real coefficient of the s-dimensional variable at the k-th sampling point, is the function of the integral of the one-dimensional basis function φ(·); x s,0 is the s-th dimension differential state variable value of the initial sampling point; x s,n u is the s-th dimension differential state variable value of the termination sampling point; n is the control variable value of the termination sampling point, y n is the algebraic state variable value of the termination sampling point; x s,j is the s-dimensional differential state variable at the j-th sampling point, x s,j-1 is the s-dimensional differential state variable of the j-1th sampling point; Step 3: Model calculation: In each optimization iteration, a simulation of the DAE model equation is performed. In the simulation layer, the Newton iteration method is used to calculate the model of the discretized nonlinear algebraic equation group to obtain the value of the state variable. Step 4: Sensitivity calculation: While the model is being calculated, the sensitivity information matrix of the state variables to the control variables and time variables is obtained; Step 5: Optimize layer calculation: In each optimization iteration, steps 4 and 5 are repeated, and the NLP solver is used to solve the nonlinear programming problem after the simulation layer processing until the convergence accuracy of the optimization iteration is reached and the optimal solution is found.

Images