Drainage basin water-wind-light dynamic adaptability aggregation decomposition scheduling rule compilation method

Abstract

The invention relates to the field of multi-energy coupling complementary scheduling, and discloses a watershed water-wind-light dynamic adaptability aggregation decomposition scheduling rule compilation method, which comprises the following steps: S1, constructing a multivariable joint stochastic simulation model, and jointly reconstructing a water-wind-light non-consistent combination scene; the method comprises the following steps: S1, solving an optimal scheduling trajectory by adopting a successive approximation dynamic programming method, S3, based on an aggregation and decomposition theory, respectively selecting appropriate independent variable factors and scheduling decision factors for an aggregation system (an integrated system) and a decomposition system (a water-wind-light-bundled subsystem in the integrated system), and S4, based on a water-wind-light scene under a non-consistency condition, carrying out scheduling on the water-wind-light-bundled subsystem in the integrated system. S5, taking the watershed water-wind-light non-consistency scene combination as the input of the model, and applying the EnKF-KNN process in the S4 on the basis of the static scheduling function type determined in the step S3, taking a static scheduling function parameter as an initial value, and applying a data assimilation and machine learning coupling model.

Description

Technical Field

[0001] The present invention belongs to the field of multi-energy coupling complementary scheduling, and specifically relates to a method for compiling dynamic adaptive aggregation and decomposition scheduling rules for a watershed water-wind-solar integrated system. Background Art

[0002] The development and utilization of renewable energy is an important measure to ensure energy security and achieve the dual carbon goals. The integrated development of renewable energy, mainly water, wind and light, integrates the rich hydropower resources in the basin and the rich wind and light resources in its surrounding areas in space, achieves complementary advantages through integrated and large-scale development, and improves the absorption and storage capacity of renewable energy. This is the only way for the high-quality leapfrog development of renewable energy in the new era. Therefore, how to safely and efficiently utilize the water, wind and light resources in the basin is a consensus problem that the global energy field is seeking to break through.

[0003] However, under the changing environment, the water, wind and light resources in the basin show non-consistent characteristics, which are specifically manifested in the changes in statistical characteristics (such as mean and variance) or probability distribution patterns, resulting in the original static scheduling rules based on fixed parameters derived from historical data no longer being applicable. On the one hand, climate change affects the meteorological and hydrological elements of the basin, resulting in changes in water, wind and light resources, such as rainfall and evaporation affecting runoff. The ever-increasing human activities further affect its generation and development, such as changes in land use types change the surface albedo and roughness, thereby affecting radiation and surface wind speed; on the other hand, the temporal and spatial complementarity between resources in the basin's integrated water, wind and light system is complex, the hydropower and electricity between power stations are closely connected, and it is difficult to take into account multiple control targets. The scheduling and operation modes between power stations in each bundled subsystem are mutually restrained, which makes it more difficult to understand the multi-energy complementary scheduling mechanism at the basin level and implement efficient and accurate adaptive management.

[0004] The following problems exist in the existing technology: (1) The integrated water, wind and solar power system in the river basin uses historical water, wind and solar power sequence data to derive static dispatch rules (i.e., the dispatch rule parameters are fixed values) for each water, wind and solar power bundled subsystem one by one to guide the dispatch operation. However, under a changing environment, resources show inconsistent characteristics. If static dispatch rules are still used, the dispatch decisions made cannot adapt to the inconsistent water, wind and solar power change scenario sequence, resulting in an underestimated dispatch benefit. At the same time, under a changing environment, the controlling factors that dominate the changes in the dispatch rules of the complementary system are unclear; (2) The dispatch rules of the water, wind and solar power bundled subsystems in the integrated water, wind and solar power system in the river basin are complexly connected, and the relationship between the energy and water volume of each subsystem, as well as the mapping relationship between the subsystem and the total energy and total water volume of the integrated system are still unclear. If these problems are not solved, the integrated system will not be able to cope with the impact of resource inconsistency on the safety and stability of its dispatch operation under a changing environment, and will not be able to ensure the maximization of the power generation benefit of the complementary system. Summary of the invention

[0005] The purpose of the present invention is to provide a method for compiling dynamic adaptive aggregation and decomposition scheduling rules for a water-wind-solar integrated system in a river basin under inconsistent conditions. The method fully considers the inconsistency of water, wind and light sequences under a changing environment; comprehensively considers the requirements of new energy grid connection and comprehensive utilization of water resources in the river basin, constructs a multi-objective complementary optimization scheduling model, analyzes the basic form of the scheduling function of the integrated system, and reveals the multi-energy complementary scheduling mechanism at the river basin level; identifies the adaptive scheduling rules of the water-wind-solar integrated system in the river basin under inconsistent conditions, obtains the time-varying scheduling function parameter sequence, and formulates an adaptive scheduling method that is dynamically adjusted with resource variations, thereby providing a scientific basis and technical support for the safe and efficient operation of the water-wind-solar integrated system in the river basin under a changing environment.

[0006] To achieve the above object, the present invention provides the following technical solution: a method for compiling dynamic adaptive aggregation and decomposition scheduling rules for water, wind and solar power in a river basin, the method comprising the following steps:

[0007] S1: Construct a multivariable joint random simulation model to jointly reconstruct the inconsistent combination scenario of water, wind and light;

[0008] S2: Construct a complementary optimization scheduling model and use the successive approximation dynamic programming method to solve the optimal scheduling trajectory;

[0009] S3: Based on the aggregation and decomposition theory, appropriate independent variable factors and scheduling decision factors are selected for the aggregation system (integrated system) and the decomposition system (the water, wind and solar power bundling subsystem within the integrated system), and a static scheduling rule function based on fixed parameters is condensed;

[0010] S4: Based on the hydro-wind-solar scenario under inconsistent conditions, the static dispatch function parameters are used as the initial values, and the data assimilation and machine learning coupling model is used to select the optimal observed variables using the machine learning method. The optimal observed variables are assimilated through the data assimilation algorithm, so as to update the parameters of the dispatch rule, identify the time-varying parameter sequence of the dispatch function of each bundled subsystem power station, and form a dynamic adaptive dispatch rule based on time-varying parameters;

[0011] S5: Take the inconsistent scenario combination of water, wind and solar power in the river basin as the input of the model. Based on the static scheduling function type determined in step S3, use the EnKF-KNN process in S4 to identify the scheduling process of the integrated water, wind and solar power system in the river basin under inconsistent conditions. Compare the scheduling benefits based on the static scheduling function and the dynamic adaptive scheduling function to demonstrate the superiority of the dynamic adaptive scheduling rules.

[0012] Preferably, in step S1, the water, wind and solar scenes in the basin are jointly simulated under inconsistent conditions, ① analyzing the probability statistical characteristics of the water, wind and solar resource sequence (runoff, wind power output, photovoltaic output), and selecting the optimal distribution profile; ② diagnosing the correlation between different variables, and selecting the autoregressive model for random simulation if the variables are independent, and if there is correlation between the variables, using the copula function to derive the joint probability distribution function;

[0013]

[0014] Where: {x(t)} represents the time series of the variable; p represents the autoregressive order; a j represents the autoregressive coefficient; ε(t) represents the random disturbance term (white noise).

[0015] F(x1,x2)=C(U,V;θ)

[0016] Where: θ represents the parameter in the copula function, which describes the dependence between two variables; F(x1,x2) represents the joint distribution function of the copula of two variables, the marginal distribution of one variable is U=F1(x1), and the marginal distribution of the other variable is V=F2(x2).

[0017] ③Introduce variation factors to adjust statistical characteristic parameters and generate water-wind-light inconsistency scenarios. There are three types of inconsistency scenarios (mean variation, C v variability, seasonal variability).

[0018] For mean variation, while keeping other parameters unchanged, the simulated values ​​are enlarged and reduced at the variation level (assuming 5%):

[0019]

[0020] For C v The variation is obtained by multiplying the simulated value and the mean value by the variation factor:

[0021]

[0022] For seasonal variation, seasonal factors are introduced and the time scale is varied:

[0023]

[0024]

[0025]

[0026] In the formula, and represents the mean of the non-consistent scenario, simulation value, and sequence; k represents the change factor; TDF j (t) represents the cumulative distribution function; Y represents the overall number of simulation years; Cumulative distribution function representing the non-consistent scenario; and Represents the cumulative distribution function before and after the inconsistency change.

[0027] Preferably, in step S2: Construction and solution of the optimization scheduling model of the integrated water, wind and photovoltaic system in the basin. ① Systematically summarize the characteristics of water, wind, photovoltaic and multiple scheduling requirements, construct a multi-objective complementary optimization scheduling model, and clarify the objective function, constraints and decision variables.

[0028] (1) Objective function

[0029]

[0030] Where: E, F and G represent the total power generation, water shortage rate and total energy storage of the integrated water, wind and solar system in the basin respectively; R and r represent the total number and serial number of hydropower stations; T and t represent the number and serial number of the dispatching period; ΔT t Indicates the number of hours in the period; and W respectively represent the hydropower, wind power and photovoltaic output of the rth hydropower, wind power and photovoltaic power station in the tth period; r (t) and Q r (t) represent the downstream water demand and outflow of the reservoir group respectively; K r represents the comprehensive output coefficient of the rth hydropower station; I r (t) represents the inflow runoff of the rth hydropower station in the tth period; H j (t) represents the generating head of the hydropower station; N and D represent the output matrix and power flow matrix of each hydropower station, both of which are decision variables.

[0031] (3) Constraints

[0032] S(t)=S(t-1)+[I(t)-Q(t)-Q e (t)-Q lose (t)]′DT t

[0033]

[0034]

[0035]

[0036]

[0037]

[0038]

[0039]

[0040] Where: S(t-1) and S(t) represent the initial storage capacity and final storage capacity of the reservoir at the tth period; I(t) and Q(t) represent the inflow and outflow of the reservoir at the tth period; Q e (t) and Q lose (t) represents the amount of evaporation loss and leakage of the reservoir in period t; S (t) and Q (t) and Indicates the upper and lower limits of the reservoir capacity and outflow at time period t; It represents the upper limit of hydropower output, that is, the installed capacity of the hydropower station; r+1 (t) represents the inflow runoff of the r+1th reservoir in the tth period; Q r (t) and represents the outflow of the rth reservoir in the tth period and the interval inflow between the rth reservoir and the r+1th reservoir; represents the output of the rth decomposition system; It represents the upper limit of the output of the rth decomposition system, that is, the maximum transmission capacity; Indicates the guaranteed output of the integrated system; N hws (t) represents the total output of the integrated system.

[0041] ② Based on the historical water-wind-solar sequence, the successive approximation dynamic programming algorithm (DPSA) is used to solve the problem, obtain the optimal dispatching trajectory of each power station, and form a "historical optimal dispatching information database". The solution idea of ​​DPSA is to first assume that the operating status of other reservoirs in the cascade reservoirs remains unchanged during the solution process, and only perform DP solution for one reservoir. After updating the optimal dispatching rule of the target reservoir, the dispatching strategy of the reservoir is fixed and the next reservoir is updated in sequence. In this way, all reservoirs are iterated and approximated one by one until the iteration accuracy is met.

[0042] Preferably, in step S3: based on the aggregation and decomposition theory, suitable independent variable factors and scheduling decision factors are selected for the aggregation system (watershed water-wind-solar integrated system) and the decomposition system (various sub-complementary systems within the watershed water-wind-solar integrated system), and a static scheduling function is condensed. Specifically, the watershed water-wind-solar integrated system is regarded as an aggregation system, and individual water-wind-solar complementary systems are regarded as decomposition systems. The scheduling rules of the aggregation system are first extracted, and then the total output of the aggregation system is allocated according to the proportion of each decomposition system.

[0043] For the total scheduling period T (t = 1, 2, ..., T), the function expression of the static scheduling rule of the aggregation system is:

[0044] Y a (t) = f a (I a (t),ES a (t),EI a (t),EA a (t); p a )

[0045] Where Y t a represents the scheduling decision of the aggregate system in period t; represents the stored energy, input energy and available energy of the aggregate system at time period t; Represents the parameters of the scheduling function.

[0046] The functional expression of the static scheduling rule of the decomposition system is:

[0047] Y d,r (t) = f d,r (S d,r (t-1),I d,r (t),EA d,r (t),EWS d,r (t),EI d,r (t),Y a (t); p d,r )

[0048] Where Y d,r (t) represents the scheduling decision of the rth decomposition system in the tth period; EWS d,r (t) represents the sum of bundled wind and solar power outputs of the rth decomposition system in the tth period; Represents the parameters of the scheduling function of the rth decomposition system.

[0049] Preferably, in step S4: formulation of dynamic adaptive dispatching rules for the integrated water, wind and solar system in the watershed based on EnKF-KNN, where the EnKF-KNN method is used to update the parameters of the static dispatching rules so that the dispatching rules change from static to dynamic.

[0050] ①Use the scheduling function parameters as state variables and the static scheduling function parameter set as the initial value of the state variables. According to the seasonal and periodic characteristics of water, wind and light resources, form the state transfer equation and observation equation, and construct the data assimilation model.

[0051] For T (t = 1, 2, ..., T) scheduling periods (consisting of m years, each year has n periods), the state transition equations of the aggregate system and the decomposed system are:

[0052]

[0053]

[0054] In the formula, represents the kth value of the set of predicted state variables of the aggregate system and the rth decomposed system; represents the kth value of the set of updated state variables of the aggregate system and the rth decomposition system; Represents the error term, which follows a Gaussian distribution with a mean of 0 and a variance of U a (i,j) and U d,r (i,j).

[0055] The selection of observation values ​​is very important for the assimilation effect. Here, the "historical optimal scheduling information library" is constructed, which consists of the optimal scheduling trajectory of the historical period in step 2. When it is necessary to select the observed variable of the current period, the KNN algorithm is used, that is, the Euclidean distance between all optional combinations in the "historical optimal scheduling information library" and the current observed variable is calculated according to the factors contained in the discrimination condition. The scheduling decision information with the closest distance is the best observed variable of the current period.

[0056]

[0057] Where: d k(t) represents the Euclidean distance between the elements of the kth candidate combination and the tth scheduling period in the “historical optimal scheduling information database”; e kn represents the nth element of the kth candidate combination of the “historical optimal scheduling information library”; e tn is the nth element of the tth scheduling period; N is the number of factors included in the judgment condition. In particular, in order to avoid the influence of different numerical values ​​on the distance calculation, the values ​​of all elements should be normalized first.

[0058] After selecting the observation values, add the observation errors to form the observation equation.

[0059]

[0060]

[0061] In the formula, and The kth value of the set of observed variables of the aggregate system and the rth decomposition system is obtained by perturbing the optimal scheduling decision and generated; and Represents the observation error term, which obeys Gaussian distribution, has a mean of 0 and a variance of Wa (i,j) and W d,r (i,j).

[0062] ② Based on the state transfer equation and the observation equation, the values ​​of the state variables are updated and adjusted by assimilating the observation values. Here, the forecast state variables of the aggregate system are taken as an example (the assimilation and update process of the forecast state variables of the decomposed system is not repeated here). The state variables are updated through the following equations.

[0063]

[0064] Where KG(i,j) represents the Kalman gain matrix, which represents the weight between the forecast scheduling decision and the observation value.

[0065] Calculate according to the following formula:

[0066]

[0067] In the formula, represents the covariance between the forecast state variables and the forecast dispatch decision; represents the autocovariance of the forecast scheduling decision. Specifically,

[0068]

[0069]

[0070] In the formula, K represents the set size; represents the mean of the set of forecast state variables; Represents the mean of the forecast scheduling decision set.

[0071] ③After updating the state variables (dispatch function parameters) through the above equation, calculate the mean of the updated state variable set. Then input the input data of the current period (runoff, wind power output and photovoltaic output values), and substitute them into the dispatch function to obtain the dispatch decision of the current period.

[0072] In addition, due to the hydraulic connection within the integrated water-wind-solar system of the basin, the dispatch function parameters of the aggregate system should be adjusted first to obtain the dispatch decision of the aggregate system. Then, the dispatch function parameters of each decomposition system are adjusted from upstream to downstream to obtain the dispatch decision of each decomposition system. In this way, all dispatch decision processes for the current period are completed.

[0073] Preferably, in step S5: the water-wind-solar inconsistency scenario combination of the river basin simulated in step 1 is used as the input of the scheduling model, based on the static scheduling function form determined in step S3, the EnKF-KNN process in step S4 is used to identify the time-varying parameters of the dynamic adaptive aggregation decomposition scheduling rules of the river basin water-wind-solar integrated system under inconsistent conditions, thereby obtaining the scheduling process of the river basin water-wind-solar integrated system under inconsistent conditions, and comparing it with the scheduling results under the static scheduling rules to demonstrate the superiority of the dynamic adaptive scheduling rules.

[0074] Compared with the prior art, the present invention has the following beneficial effects:

[0075] (1) The present invention identifies the variation characteristics and complementary properties of runoff, wind power output, and photovoltaic output in a river basin under a changing environment, and accurately simulates non-consistent scenarios, providing a basic premise for implementing coordinated regulation of the integrated water, wind, and photovoltaic systems in the river basin.

[0076] (2) Based on the aggregation-decomposition concept, the present invention determines the dispatching rule type of each bundled subsystem in the integrated water-wind-solar system of the river basin, thereby accurately describing the dispatching and operating rules between different resources and different power stations in the energy system at the river basin level. It is an important tool for realizing the coordinated regulation of the integrated water-wind-solar system of the river basin.

[0077] (3) The method for compiling dynamic adaptive dispatching rules for the integrated water, wind and solar power system in a river basin under inconsistency conditions proposed in this invention uses data assimilation methods to identify the time-varying dispatching function parameter sequence. It realizes the adaptive dynamic adjustment of the dispatching function parameters of the entire river basin as the input inconsistency changes, and reveals the response mechanism of the dispatching rules to resource inconsistency. It makes up for the defects of traditional static dispatching rules based on fixed parameters, and provides important theoretical guarantees for the energy system to cope with the changing environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 A flow chart of a method for compiling dynamic adaptive dispatching rules for a watershed water-wind-solar integrated system under non-consistent conditions in an embodiment of the present invention;

[0079] Figure 2 It is a flow chart of a method for compiling static dispatching rules of a watershed water-wind-solar integrated system based on aggregation decomposition in an embodiment of the present invention;

[0080] Figure 3 Schematic diagram of the calculation principle of EnkF-KNN based on the acceptance / rejection method in an embodiment of the present invention;

[0081] Figure 4 This is a graph of the inconsistent input scene of water, wind and light resources under the condition of mean value increase in an embodiment of the present invention;

[0082] Figure 5 This is an inconsistency input scene diagram of water, wind and light resources under the condition of mean value reduction in an embodiment of the present invention;

[0083] Figure 6 C in the embodiment of the present invention v Non-uniform input scenario diagram of water, wind and light resources under rising conditions;

[0084] Figure 7 C in the embodiment of the present invention v Non-consistent input scenario diagram of water, wind and light resources under falling conditions;

[0085] Figure 8 It is an inconsistent input scene diagram of water, wind and light resources under seasonal variation in an embodiment of the present invention;

[0086] Fig. 9 The change process of water level, inflow and outflow in the DPSA result in the embodiment of the present invention. (a) is the first decomposition system diagram, (b) is the second decomposition system diagram;

[0087] Fig.10 This is a static scheduling rule diagram of the aggregation system in an embodiment of the present invention. DETAILED DESCRIPTION

[0088] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0089] The method for compiling dynamic adaptive aggregation and decomposition scheduling rules for a watershed water-wind-solar integrated system under inconsistent conditions provided in this embodiment includes the following steps:

[0090] Step 1. Joint simulation of water, wind and light scenarios in the basin under inconsistent conditions:

[0091] ①Select the optimal distribution model of water, wind and light resources in the basin

[0092] Eight different probability distributions (Table 1) were selected to test the goodness of fit of the probability distribution process of runoff, wind power output, and photovoltaic output sequences in the integrated water-wind-solar system of the basin, including the Kolmogorov-Smirnov (KS) test, the root mean square error criterion (RMSE), and the Akaike information criterion (AIC) method to optimize the probability distribution function, and finally the best probability distribution model was selected, and the results are shown in Table 2. Overall, the inflow runoff is highly consistent with the generalized extreme value distribution in most months of the year, especially in the flood season; the wind power output sequence is highly consistent with the Weibull distribution in most months of the year; photovoltaic output 1 is highly consistent with the gamma distribution and lognormal distribution in the year, and photovoltaic output 2 is highly consistent with the generalized extreme value distribution in the year.

[0093] Table 1 Probability distribution function to be adopted

[0094]

[0095] D=max|F exp (x)-F obs |

[0096] Where: D represents the KS test statistic; F exp (x) represents the cumulative empirical distribution function of the sequence {x}; Fobs represents the empirical frequency. When the test statistic meets the given significance level, the null hypothesis is accepted.

[0097]

[0098] Where: p represents the empirical frequency; p i represents the theoretical frequency; n represents the number of samples. The smaller the RMSE, the better the fitting accuracy.

[0099] AIC=mln(MSE)+2k

[0100]

[0101] Where: MSE represents mean square error; k represents the number of model parameters. The smaller the AIC, the better the fitting accuracy.

[0102] Table 2 Optimal probability distribution of water-wind-light sequence at monthly scale

[0103]

[0104] ②Construction of a joint random simulation model for water, wind and light scenarios in a river basin under non-consistent conditions

[0105] Diagnose the correlation between the runoff, wind power output, and photovoltaic output variable sequences. For independent variables, AR is used for random simulation; for multidimensional variables with correlation, a joint probability distribution function is constructed based on the Copula function, and a variable is selected as the main variable. A long series of samples is randomly simulated, and then other slave variables are simulated based on the Bayesian conditional probability formula.

[0106] Through correlation diagnosis, there is a certain correlation between the runoff sequence into the reservoir and the wind and solar power output sequence. Therefore, the AR model is used to simulate the runoff, and then the Frank Copula function (the following formula) is used to couple the wind power output and photovoltaic output.

[0107]

[0108] Where C(·) represents the Copula function; θ represents the parameter of the Copula function.

[0109] The simulation results are shown in Table 3-5. The results show that the error between the simulated sequence and the measured sequence is small, and the overall characteristics of the two sequences are relatively close.

[0110] Table 3 Comparison of monthly runoff simulation results and measured data

[0111]

[0112] Table 4 Comparison of wind power output simulation results and measured data on a monthly basis

[0113]

[0114] Table 5 Comparison of monthly simulation results and measured data of photovoltaic output

[0115]

[0116]

[0117] ③ Representative water and landscape scenes in the basin under non-consistent conditions

[0118] After verification, the simulation accuracy of the joint random simulation model is good and it can be used to generate inconsistent scenarios. The statistical characteristic factor adjustment method is further used to generate inconsistent scenarios. The statistical characteristic factor adjustment method focuses on statistical characteristic parameters (such as mean and coefficient of variation) and focuses on the impact of the changing environment on the statistical characteristic parameters of the variables. v , seasonal variation is the variation type, simulating the inconsistent scenarios under 5 changing environments (as shown below Figure 1-5): mean increase, mean decrease, increase, decrease, and seasonal variation. Finally, 100 sets of inconsistent input scenarios of water, wind and light resources were obtained. From the statistical parameters in the figure, it can be seen that the parameter changes obtained by simulation are consistent with the expected variation type.

[0119] Step 2. Construction and solution of the optimal dispatching model of the integrated water, wind and solar system in the basin

[0120] ①Construction of complementary optimization scheduling model for integrated water, wind and solar power systems in river basins

[0121] The multiple dispatching objectives (such as power generation, water supply and flood control) of the integrated water, wind and solar power system in the basin are systematically sorted out, and the constraints of each power station and power grid as well as the hydraulic and power connections between upstream and downstream power stations are considered to establish a multi-objective complementary optimization dispatching model as follows:

[0122] (1) Objective function

[0123]

[0124] Where: E, F and G represent the total power generation, water shortage rate and total energy storage of the integrated water, wind and solar system in the basin respectively; R and r represent the total number and serial number of hydropower stations; T and t represent the number and serial number of the dispatching period; ΔT t Indicates the number of hours in the period; and W respectively represent the hydropower, wind power and photovoltaic output of the rth hydropower, wind power and photovoltaic power station in the tth period; r (t) and Q r (t) represent the downstream water demand and outflow of the reservoir group respectively; K r represents the comprehensive output coefficient of the rth hydropower station; I r (t) represents the inflow runoff of the rth hydropower station in the tth period; H j (t) represents the generating head of the hydropower station; N and D represent the output matrix and power flow matrix of each hydropower station, both of which are decision variables.

[0125] (4) Constraints

[0126] S(t)=S(t-1)+[I(t)-Q(t)-Q e (t)-Q lose (t)]′DT t

[0127]

[0128]

[0129]

[0130]

[0131]

[0132]

[0133]

[0134] Where: S(t-1) and S(t) represent the initial storage capacity and final storage capacity of the reservoir at the tth period; I(t) and Q(t) represent the inflow and outflow of the reservoir at the tth period; Q e (t) and Q lose (t) represents the amount of evaporation loss and leakage of the reservoir in period t; S (t) and Q (t) and Indicates the upper and lower limits of the reservoir capacity and outflow at time period t; It represents the upper limit of hydropower output, that is, the installed capacity of the hydropower station; r+1 (t) represents the inflow runoff of the (r+1)th reservoir in the tth period; Q r (t) and It represents the outflow of the rth reservoir in the tth period and the interval inflow between the rth reservoir and the r+1th reservoir. represents the output of the rth decomposition system; It represents the upper limit of the output of the rth decomposition system, that is, the maximum transmission capacity; Indicates the guaranteed output of the integrated system; N hws (t) represents the total output of the integrated system.

[0135] ②Solving the optimal dispatching trajectory of the integrated water, wind and solar system in the basin under historical scenarios

[0136] Taking the historical series of water, wind and solar power in the basin from 1959 to 2019 as input, the dynamic programming method by successive approximation (DPSA) is used to solve the complementary optimization scheduling model, obtain a long series of scheduling samples, and clarify the optimal scheduling trajectory of each power station in the basin.

[0137] The optimized output results of DPSA are shown in Table 6. It can be seen that: ① The average output of the Cihaxia Hydropower Station in the cascade hydropower station system is the largest (989.80MW), accounting for 39.73% of the average output of the cascade hydropower stations; Longyangxia Hydropower Station is second (696.93MW), accounting for 27.97%; Yangqu Hydropower Station is in the middle (621.66MW), accounting for 24.95%; Banduo Hydropower Station is the least (182.99MW), accounting for 7.35%. ② The average total output of the basin water-wind-solar integrated system under optimized scheduling is 3802.70MW, of which sub-complementary system 1 contributes 2945.30MW, accounting for 77.45%, and sub-complementary system 2 contributes 857.40MW, accounting for 22.55%. ③ The average annual output of cascade hydropower stations under optimized dispatch is 2491.39MW, accounting for 65.52% of the average annual output of the integrated water-wind-solar system. Hydropower output is the main bearer of the complementary output of water, wind and solar. In addition, the "minimum water shortage rate" in the objective function is converted into a "penalty function" for solution during optimization. In the optimized dispatch results, the outflow in each period can meet the water demand, so it will not be discussed separately.

[0138] Table 6 Output results of water-wind-solar integrated system under optimal dispatch

[0139]

[0140] In the optimal dispatching trajectory, the changes in water level and flow of each reservoir are as follows: Figure 6 . It can be seen that the inflow of each decomposition system increases from January to July and decreases from August to December. The change trend of power generation flow is the same as that of inflow. In particular, the amplitude of the change trend of decomposition system 2 is smaller than that of decomposition system 1. The amount of abandoned water is mainly concentrated in June-October, among which the amount of abandoned water in decomposition system 2 is greater than that in decomposition system 1, and it is earlier in time. The water level distribution of each decomposition system is more concentrated in September-November.

[0141] Step 3. Extraction of static dispatching rules of the integrated water-wind-solar system in the basin based on aggregation decomposition:

[0142] ① Static scheduling rules of the aggregation system

[0143] The variables set initially selected for the aggregated virtual reservoir are: total output, total outflow, total energy storage, total inflow, total input energy, and total available energy. Based on the selected variable set and the proposed long-term linear dispatch function type, the correlation coefficient method is used to calculate the correlation between each decision variable and the state variable, and the factor with the strongest correlation is selected to extract the dispatch function of the aggregated reservoir. Finally, the total output of the aggregated system is selected as the decision variable and the total input energy is selected as the independent variable to construct the static dispatch rule of the aggregated system. The results are as follows: Figure 7 .

[0144] ②Decomposition system static scheduling rules

[0145] The variable set of the decomposition system includes: initial storage capacity, final storage capacity, hydropower output, inflow, outflow, available energy, total output of the aggregated energy system, supporting wind power output, supporting photovoltaic output and comprehensive wind and solar output.

[0146] According to the aggregation and decomposition theory, the total output of the aggregation system is allocated to each decomposition system (such as attached Figure 2 ). The stepwise regression method is used to extract the static operation law of the decomposition system, and the results are shown in Table 7. It can be seen that the dependent variable of decomposition system 1 is the outbound flow, and the dependent variable of decomposition system 2 is the storage capacity at the end of the period.

[0147] Table 7 Parameter matrix of static scheduling rules of decomposition system

[0148]

[0149]

[0150] Note: From top to bottom, they are January to December. Q1(t) represents the outflow of decomposition system 1, and S2(t) represents the storage capacity of decomposition system 2 at the end of the period.

[0151] Step 4. Formulate adaptive dynamic scheduling rules for the integrated water, wind and solar power system in the river basin.

[0152] The scheduling function parameters of each scheduling period are regarded as state variables to be assimilated and updated, and the optimal scheduling decision selected from the "historical optimal scheduling information library" using the KNN algorithm is regarded as the observation variable, and a data assimilation model based on the ensemble Kalman filter algorithm (EnKF) is constructed.

[0153] Specifically, in the state transfer equation, the scheduling function parameter update value matrix of the same period of the previous year is disturbed as the parameter prediction value matrix of the current period (the static scheduling function parameter matrix is ​​the initial matrix value of the first period). For T (t = 1, 2, ..., T) scheduling periods (from m years, each year has n periods), the state transfer equations of the aggregate system and the decomposed system are:

[0154]

[0155]

[0156] In the formula, represents the kth value of the set of predicted state variables of the aggregate system and the rth decomposed system; represents the kth value of the set of updated state variables of the aggregate system and the rth decomposition system; Represents the error term, which follows a Gaussian distribution with a mean of 0 and a variance of U a (i,j) and U d,r (i,j).

[0157] The optimal scheduling decision matrix selected by disturbance in the observation equation is used as the observation variable matrix of the current period. For the aggregate system and the decomposed system, they are:

[0158]

[0159]

[0160] In the formula, and The kth value of the set of observed variables of the aggregate system and the rth decomposition system is obtained by perturbing the optimal scheduling decision and generated; and Represents the observation error term, which obeys Gaussian distribution, has a mean of 0 and a variance of W a (i,j) and W d,r (i,j).

[0161] The Kalman gain matrix is ​​calculated by the following equation to obtain the updated values ​​of the scheduling function parameters (taking the scheduling function parameters of the aggregated system as an example, the scheduling function parameters of the decomposed system will not be repeated).

[0162]

[0163] Where KG(i,j) represents the Kalman gain matrix, which represents the weight between the forecast scheduling decision and the observation value; represents the covariance between the forecast state variables and the forecast dispatch decision; represents the autocovariance of forecast scheduling decision; K represents the set size; represents the mean of the set of forecast state variables; Represents the mean of the forecast scheduling decision set.

[0164] After updating the scheduling function parameters of the aggregate system, update the scheduling function parameters of each decomposition system one by one (such as the attached Figure 4 ).

[0165] Step 5. Formulate adaptive dynamic aggregation and decomposition scheduling rules for the integrated water-wind-solar system in the basin under non-consistent conditions.

[0166] The water-wind-solar non-consistency scenario combination of the basin simulated in step 1 is used as the input of the scheduling model. Based on the static scheduling function form determined in step 3, the EnKF-KNN process in step 4 is used. The time-varying parameters of the dynamic adaptive aggregation decomposition scheduling rule of the basin water-wind-solar integrated system under the condition of inconsistency are identified, so as to obtain the scheduling process of the basin water-wind-solar integrated system under the condition of inconsistency, and compared with the scheduling results under the static scheduling rules to demonstrate the superiority of the dynamic adaptive scheduling rules. The results are shown in Table 8 below.

[0167] Table 8 Scheduling results in non-consistent scenarios

[0168]

[0169] In summary, the method for compiling dynamic adaptive aggregation and decomposition scheduling rules for the integrated water-wind-solar system in a river basin under inconsistent conditions proposed in the present invention can effectively respond to the impact of the changing environment on the scheduling and operation safety and stability of the complementary energy system, improve the resilience of the complementary energy system, and ensure the maximization of its power generation efficiency.

[0170] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0171] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for compiling dynamic adaptive aggregation and decomposition scheduling rules for water, wind and solar power in a river basin, characterized by: The method comprises the following steps: S1: Construct a multivariable joint random simulation model to jointly reconstruct the inconsistent combination scenario of water, wind and light; S2: Construct a complementary optimization scheduling model and use the successive approximation dynamic programming method to solve the optimal scheduling trajectory; S3: Based on the aggregation and decomposition theory, appropriate independent variable factors and scheduling decision factors are selected for the aggregation system (integrated system) and the decomposition system (the water, wind and solar power bundling subsystem within the integrated system), and a static scheduling rule function based on fixed parameters is condensed; S4: Based on the hydro-wind-solar scenario under inconsistent conditions, the static dispatch function parameters are used as the initial values, and the data assimilation and machine learning coupling model is used to select the optimal observed variables using the machine learning method. The optimal observed variables are assimilated through the data assimilation algorithm, so as to update the parameters of the dispatch rule, identify the time-varying parameter sequence of the dispatch function of each bundled subsystem power station, and form a dynamic adaptive dispatch rule based on time-varying parameters; S5: Take the inconsistent scenario combination of water, wind and solar power in the river basin as the input of the model. Based on the static scheduling function type determined in step S3, use the EnKF-KNN process in S4 to identify the scheduling process of the integrated water, wind and solar power system in the river basin under inconsistent conditions. Compare the scheduling benefits based on the static scheduling function and the dynamic adaptive scheduling function to demonstrate the superiority of the dynamic adaptive scheduling rules.

2. According to claim 1, a method for compiling basin water, wind and solar dynamic adaptive aggregation and decomposition scheduling rules is characterized by: In step S1, the water, wind and solar scenes in the basin are jointly simulated under inconsistent conditions. ① The probability statistical characteristics of the water, wind and solar resource sequence (runoff, wind power output, photovoltaic output) are analyzed, and the optimal distribution profile is selected; ② The correlation between different variables is diagnosed. If the variables are independent, the autoregressive model is selected for random simulation. If there is a correlation between the variables, the copula function is used to derive the joint probability distribution function; Where: {x(t)} represents the time series of the variable; p represents the autoregressive order; a j represents the autoregressive coefficient; ε(t) represents the random disturbance term (white noise). F(x1,x2)=C(U,V;θ) Where: θ represents the parameter in the copula function, which describes the dependence between two variables; F(x1,x2) represents the joint distribution function of the copula of two variables, the marginal distribution of one variable is U=F1(x1), and the marginal distribution of the other variable is V=F2(x2). ③Introduce variation factors to adjust statistical characteristic parameters and generate water-wind-light inconsistency scenarios. There are three types of inconsistency scenarios (mean variation, C v variability, seasonal variability). For mean variation, while keeping other parameters unchanged, the simulated values ​​are enlarged and reduced at the variation level (assuming 5%): For C v The variation is obtained by multiplying the simulated value and the mean value by the variation factor: For seasonal variation, seasonal factors are introduced and the time scale is varied: In the formula, and represents the mean of the non-consistent scenario, simulation value, and sequence; k represents the change factor; TDF j (t) represents the cumulative distribution function; Y represents the overall number of simulation years; Cumulative distribution function representing the non-consistent scenario; and Represents the cumulative distribution function before and after the inconsistency change.

3. According to claim 1, a method for compiling basin water, wind and solar dynamic adaptive aggregation and decomposition scheduling rules is characterized by: In step S2: Construction and solution of the optimal dispatching model of the integrated water, wind and photovoltaic system in the basin. ① The system summarizes the characteristics of water, wind and photovoltaic and multiple dispatching requirements, constructs a multi-objective complementary optimal dispatching model, and clarifies the objective function, constraints and decision variables. (1) Objective function Where: E, F and G represent the total power generation, water shortage rate and total energy storage of the integrated water, wind and solar system in the basin respectively; R and r represent the total number and serial number of hydropower stations; T and t represent the number and serial number of the dispatching period; ΔT t Indicates the number of hours in the period; and W respectively represent the hydropower, wind power and photovoltaic output of the rth hydropower, wind power and photovoltaic power station in the tth period; r (t) and Q r (t) represent the downstream water demand and outflow of the reservoir group respectively; K r represents the comprehensive output coefficient of the rth hydropower station; I r (t) represents the inflow runoff of the rth hydropower station in the tth period; H j (t) represents the generating head of the hydropower station; N and D represent the output matrix and power flow matrix of each hydropower station, both of which are decision variables. (2) Constraints S(t)=S(t-1)+[I(t)-Q(t)-Q e (t)-Q lose (t)]′DT t Where: S(t-1) and S(t) represent the initial storage capacity and final storage capacity of the reservoir at the tth period; I(t) and Q(t) represent the inflow and outflow of the reservoir at the tth period; Q e (t) and Q lose (t) represents the amount of evaporation loss and leakage of the reservoir in period t; and and Indicates the upper and lower limits of the reservoir capacity and outflow at time period t; It represents the upper limit of hydropower output, that is, the installed capacity of the hydropower station; r+1 (t) represents the inflow runoff of the r+1th reservoir in the tth period; Q r (t) and represents the outflow of the rth reservoir in the tth period and the interval inflow between the rth reservoir and the r+1th reservoir; represents the output of the rth decomposition system; It represents the upper limit of the output of the rth decomposition system, that is, the maximum transmission capacity; Indicates the guaranteed output of the integrated system; N hws (t) represents the total output of the integrated system. ② Based on the historical water-wind-solar sequence, the successive approximation dynamic programming algorithm (DPSA) is used to solve the problem, obtain the optimal dispatch trajectory of each power station, and form a "historical optimal dispatch information database". The solution idea of ​​DPSA is to assume that the operating status of other reservoirs in the cascade reservoirs remains unchanged during the solution process, and only perform DP solution for one reservoir. After updating the optimal dispatch rule of the target reservoir, the dispatch strategy of the reservoir is fixed and the next reservoir is updated in sequence. In this way, all reservoirs are iterated and approximated one by one until the iteration accuracy is met.

4. According to claim 1, a method for compiling basin water, wind and solar dynamic adaptive aggregation and decomposition scheduling rules is characterized by: In step S3: based on the aggregation and decomposition theory, appropriate independent variable factors and scheduling decision factors are selected for the aggregation system (watershed water-wind-solar integrated system) and the decomposition system (various sub-complementary systems within the watershed water-wind-solar integrated system), and a static scheduling function is condensed. Specifically, the watershed water-wind-solar integrated system is regarded as an aggregation system, and the individual water-wind-solar complementary systems are regarded as decomposition systems. The scheduling rules of the aggregation system are first extracted, and then the total output of the aggregation system is allocated according to the proportion of each decomposition system. For the total scheduling period T (t = 1, 2, ..., T), the function expression of the static scheduling rule of the aggregation system is: Y a (t)=f a (I a (t),ES a (t),EI a (t),EA a (t);p a ) In the formula, represents the scheduling decision of the aggregate system in period t; represents the stored energy, input energy and available energy of the aggregate system at time period t; Represents the parameters of the scheduling function. The functional expression of the static scheduling rule of the decomposition system is: Y d,r (t)=f d,r (S d,r (t-1),I d,r (t),EA d,r (t),EWS d,r (t),EI d,r (t),Y a (t);p d,r ) Where Y d,r (t) represents the scheduling decision of the rth decomposition system in the tth period; EWS d,r (t) represents the sum of bundled wind and solar power outputs of the rth decomposition system in the tth period; Represents the parameters of the scheduling function of the rth decomposition system.

5. According to claim 1, a method for compiling basin water, wind and solar dynamic adaptive aggregation and decomposition scheduling rules is characterized by: In step S4: the dynamic adaptive dispatching rules of the integrated water, wind and solar power system in the river basin are formulated based on EnKF-KNN. Here, the EnKF-KNN method is used to update the parameters of the static dispatching rules, so that the dispatching rules change from static to dynamic. ① Taking the scheduling function parameters as state variables and the static scheduling function parameter set as the initial value of the state variables, the state transfer equation and observation equation are formed according to the seasonal and periodic characteristics of water, wind and light resources, and the data assimilation model is constructed. For T (t = 1, 2, ..., T) scheduling periods (consisting of m years, each year has n periods), the state transition equations of the aggregate system and the decomposed system are: In the formula, represents the kth value of the set of predicted state variables of the aggregate system and the rth decomposed system; represents the kth value of the set of updated state variables of the aggregate system and the rth decomposition system; Represents the error term, which follows a Gaussian distribution with a mean of 0 and a variance of U a (i,j) and U d,r (i,j). The selection of observations is very important for the assimilation effect. Here, the "historical optimal scheduling information library" is constructed, which consists of the optimal scheduling trajectory of the historical period in step 2. When it is necessary to select the observed variable of the current period, the KNN algorithm is used, that is, the Euclidean distance between all optional combinations in the "historical optimal scheduling information library" and the current observed variable is calculated according to the factors contained in the discrimination condition. The scheduling decision information with the closest distance is the best observed variable of the current period. Where: d k(t) represents the Euclidean distance between the elements of the kth candidate combination and the tth scheduling period in the "historical optimal scheduling information database"; e kn represents the nth element of the kth candidate combination of the "historical optimal scheduling information library"; e tn is the nth element of the tth scheduling period; N is the number of factors included in the judgment condition. In particular, in order to avoid the influence of different numerical values ​​on the distance calculation, the values ​​of all elements should be normalized first. After selecting the observation values, add the observation errors to form the observation equation. In the formula, and The kth value of the set of observed variables of the aggregate system and the rth decomposition system is obtained by perturbing the optimal scheduling decision and generated; and Represents the observation error term, which obeys Gaussian distribution, has a mean of 0 and a variance of W a (i,j) and W d,r (i,j). ② Based on the state transfer equation and the observation equation, the values ​​of the state variables are updated and adjusted by assimilating the observation values. Here, the forecast state variables of the aggregate system are taken as an example (the assimilation and update process of the forecast state variables of the decomposed system is not repeated here). The state variables are updated through the following equations. Where KG(i,j) represents the Kalman gain matrix, which represents the weight between the forecast scheduling decision and the observation value. Calculate according to the following formula: In the formula, represents the covariance between the forecast state variables and the forecast dispatch decision; represents the autocovariance of the forecast scheduling decision. Specifically, In the formula, K represents the set size; represents the mean of the set of forecast state variables; Represents the mean of the forecast scheduling decision set. ③ After updating the state variables (dispatching function parameters) through the above equation, calculate the mean of the updated state variable set, and then input the input data of the current period (runoff, wind power output and photovoltaic output values), and substitute them into the dispatching function to obtain the dispatching decision of the current period. In addition, due to the hydraulic connection within the integrated water-wind-solar system in the river basin, the scheduling function parameters of the aggregation system should be adjusted first to obtain the scheduling decision of the aggregation system. Then, the scheduling function parameters of each decomposition system should be adjusted from upstream to downstream to obtain the scheduling decision of each decomposition system. In this way, all scheduling decision processes in the current period are completed.

6. The method for compiling basin water, wind and solar dynamic adaptive aggregation and decomposition scheduling rules according to claim 1 is characterized by: In step S5: the water-wind-solar inconsistency scenario combination of the river basin simulated in step 1 is used as the input of the scheduling model, and based on the static scheduling function form determined in step S3, the EnKF-KNN process in step S4 is used to identify the time-varying parameters of the dynamic adaptive aggregation decomposition scheduling rules of the river basin water-wind-solar integrated system under inconsistent conditions, thereby obtaining the scheduling process of the river basin water-wind-solar integrated system under inconsistent conditions, and comparing it with the scheduling results under the static scheduling rules to demonstrate the superiority of the dynamic adaptive scheduling rules.

Images