Power grid multi-energy complementary scheduling method and device based on wind and light output uncertainty

Abstract

The invention discloses a power grid multi-energy complementary scheduling method and device based on wind and light output uncertainty, and the method comprises the steps: obtaining the actual wind and light output data and predicted wind and light output data corresponding to a target power grid at a plurality of historical time points in a historical time period, and obtaining the wind and light output uncertainty of the target power grid based on the difference value between the actual wind and light output data and the predicted wind and light output data; constructing a covariance matrix; based on the covariance matrix, determining a wind and light prediction output error, obtaining a plurality of initial wind and light prediction output scene sets, and correcting each initial wind and light prediction output scene set by using the wind and light prediction output error to obtain a plurality of wind and light prediction output scene sets, each wind and light prediction output scene set comprises wind and light prediction output data of each time point in a future time period; and inputting each wind-solar prediction output scene set into a preset wind-solar complementary scheduling model to select an output scene set, obtaining a target wind-solar prediction output scene set, and performing electric energy scheduling according to the target wind-solar prediction output scene set.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system dispatching, in particular to a power grid multi-energy complementary dispatching method and device based on wind and light output uncertainty. BACKGROUND

[0002] The rapid growth of new energy installed capacity combined with low power prediction accuracy leads to insufficient flexible regulation resources in the power system, which is particularly serious in power systems dominated by thermal power, posing a threat to the safe and stable operation of the power grid. Therefore, it is necessary to complementarily dispatch the power of the power grid based on wind and light output uncertainty.

[0003] Currently, power dispatch is usually directly based on the predicted wind and light output capacity. However, this method only focuses on the static capacity of wind and light output, and since wind and light output has time-varying characteristics, the static capacity cannot represent the actual output capacity of wind and light, resulting in poor dispatching effect and affecting the stable operation of the power grid. SUMMARY

[0004] The present application provides a power grid multi-energy complementary dispatching method and device based on wind and light output uncertainty, which can improve the accuracy of power grid multi-energy complementary dispatching based on wind and light output uncertainty and ensure the stable operation of the power grid.

[0005] According to a first aspect of the present application, a power grid multi-energy complementary dispatching method based on wind and light output uncertainty is provided, comprising:

[0006] Obtaining actual wind and light output data and predicted wind and light output data corresponding to a target power grid at a plurality of historical time points in a historical time period, and constructing a covariance matrix based on the difference between the actual wind and light output data and the predicted wind and light output data at the same historical time point;

[0007] Based on the covariance matrix, determine the wind and light prediction output error, and obtain a plurality of initial wind and light prediction output scene sets, and correct each initial wind and light prediction output scene set using the wind and light prediction output error to obtain a plurality of wind and light prediction output scene sets, wherein each wind and light prediction output scene set includes wind and light prediction output data at each time point in a future time period;

[0008] Each wind and light prediction output scene set is input into a preset wind and light complementary dispatching model to select the output scene set, obtain a target wind and light prediction output scene set, and perform power dispatch according to the target wind and light prediction output scene set.

[0009] According to a second aspect of the present application, a power grid multi-energy complementary dispatching device based on wind and light output uncertainty is provided, comprising:

[0010] The acquisition unit is configured to acquire actual wind-solar output data and predicted wind-solar output data corresponding to a target power grid at a plurality of historical time points in a historical time period, and construct a covariance matrix based on a difference between the actual wind-solar output data and the predicted wind-solar output data at the same historical time point.

[0011] The correction unit is configured to determine wind-solar prediction output errors based on the covariance matrix, acquire a plurality of initial wind-solar prediction output scenario sets, and correct each of the initial wind-solar prediction output scenario sets by using the wind-solar prediction output errors to obtain a plurality of wind-solar prediction output scenario sets, wherein each of the wind-solar prediction output scenario sets includes wind-solar prediction output data at each time point in a future time period.

[0012] The selection unit is configured to input each of the wind-solar prediction output scenario sets into a preset wind-solar complementary scheduling model to select an output scenario set, obtain a target wind-solar prediction output scenario set, and perform power scheduling according to the target wind-solar prediction output scenario set.

[0013] According to a third aspect of the present application, a computer readable storage medium is provided, which stores a computer program, and the program is executed by a processor to implement the above power grid multi-energy complementary scheduling method based on wind-solar output uncertainty.

[0014] According to a fourth aspect of the present application, a computer device is provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the above power grid multi-energy complementary scheduling method based on wind-solar output uncertainty when executing the program.

[0015] According to the power grid multi-energy complementary scheduling method and device based on wind and light output uncertainty provided by the application, compared with the current method of directly scheduling electric energy based on predicted wind and light output capacity, the application obtains actual wind and light output data and predicted wind and light output data corresponding to a target power grid at a plurality of historical time points in a historical time period, constructs a covariance matrix based on the difference between the actual wind and light output data and the predicted wind and light output data at the same historical time point, determines wind and light prediction output errors based on the covariance matrix, and obtains a plurality of initial wind and light prediction output scene sets, respectively corrects each initial wind and light prediction output scene set by using the wind and light prediction output errors to obtain a plurality of wind and light prediction output scene sets, wherein each wind and light prediction output scene set includes wind and light prediction output data at each time point in a future time period, finally inputs each wind and light prediction output scene set into a preset wind and light complementary scheduling model to select an output scene set, obtains a target wind and light prediction output scene set, and performs electric energy scheduling according to the target wind and light prediction output scene set. Thus, the covariance matrix is constructed by using the actual wind and light output data and the predicted wind and light output data with time sequence, the wind and light prediction output errors with time-varying characteristics are determined based on the covariance matrix, the wind and light prediction output data in the initial wind and light prediction output scene set is corrected by using the wind and light prediction output errors, and finally the wind and light prediction output data after error correction is used to realize the complementary scheduling of the power grid, so that the scheduling accuracy of electric energy in the power grid can be improved, and the stable operation of the power grid can be ensured. BRIEF DESCRIPTION OF DRAWINGS

[0016] The drawings described herein are used to provide further understanding of the application, form a part of the application, and the illustrative embodiments of the application and the description thereof are used to explain the application, and do not constitute improper limitations on the application. In the drawings:

[0017] Figure 1 A flow chart of a power grid multi-energy complementary scheduling method based on wind and light output uncertainty provided by an embodiment of the application is shown;

[0018] Figure 2 A flow chart of another power grid multi-energy complementary scheduling method based on wind and light output uncertainty provided by an embodiment of the application is shown;

[0019] Figure 3 A wind power actual output, prediction output and generation scene schematic diagram provided by an embodiment of the application is shown;

[0020] Figure 4 A photovoltaic actual output, prediction output and generation scene schematic diagram provided by an embodiment of the application is shown;

[0021] Figure 5A schematic diagram of a regulation range for thermal power processing under a certain existing power grid dispatching mode is shown.

[0022] Figure 6 A schematic diagram of power curtailment and power shortage under a certain existing power grid dispatching mode is shown.

[0023] Figure 7 A schematic diagram of a thermal power regulation range in a multi-energy complementary dispatching process of a power grid based on wind and light output uncertainty is shown.

[0024] Figure 8 A schematic diagram of power curtailment and power shortage in a multi-energy complementary dispatching process of a power grid based on wind and light output uncertainty is shown.

[0025] Figure 9 A schematic diagram of a structure of a multi-energy complementary dispatching device of a power grid based on wind and light output uncertainty is shown.

[0026] Figure 10 A schematic diagram of a structure of another multi-energy complementary dispatching device of a power grid based on wind and light output uncertainty is shown.

[0027] Figure 11 A schematic diagram of an entity structure of a computer device is shown. DETAILED DESCRIPTION

[0028] Hereinafter, the present application will be described in detail with reference to the accompanying drawings and in conjunction with embodiments. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0029] At present, the method of directly scheduling electric energy based on predicted wind and light output capacity cannot represent the actual output capacity of wind and light due to the time-varying characteristics of wind and light output, thereby resulting in poor scheduling effect and affecting the stable operation of the power grid.

[0030] In order to solve the above problems, the embodiments of the present application provide a multi-energy complementary dispatching method of a power grid based on wind and light output uncertainty, as shown in Figure 1 The method comprises the following steps.

[0031] 101, obtaining actual wind and light output data and predicted wind and light output data corresponding to a target power grid at a plurality of historical time points in a historical time period, and constructing a covariance matrix based on a difference between the actual wind and light output data and the predicted wind and light output data at the same historical time point.

[0032] The historical time period can be selected according to actual needs, such as the past several months or the past year; the historical time point can be each day within the past month; the actual wind and solar power output data includes the sum of the actual processing data of wind power generation equipment and the actual processing data of photovoltaic power generation equipment.

[0033] In this embodiment of the invention, the uncertainties of wind and solar power (wind power generation equipment and photovoltaic power generation equipment) are considered. Actual wind and solar power output data of the target power grid at multiple time points within a historical period are obtained through the power grid monitoring system, and predicted wind and solar power output data for the corresponding time points are obtained through the wind and solar power output prediction system. The acquired actual and predicted data are cleaned to remove outliers, missing values, or duplicate data. For example, interpolation or filtering methods can be used to handle missing data or outliers. Data alignment: Ensure that the actual and predicted data at the same time point correspond. Since the actual and predicted data may have inconsistent timestamps, time alignment processing is required, for example, using nearest neighbor interpolation or linear interpolation methods. Data standardization: Standardize the data to eliminate the influence of dimensions and orders of magnitude. For example, the Z-score standardization method can be used to convert the data into a distribution with a mean of 0 and a standard deviation of 1. For each historical time point, the difference between the actual and predicted wind and solar power output data is calculated. Assuming the actual wind and solar power output data is a random variable, a covariance matrix is ​​constructed. Each element of the covariance matrix represents the covariance between the differences at two different time points. For example, if the difference vector is X = (X1, X2, ..., X...), then... t ) T , where x i Let Σ be the difference between the actual and predicted wind and solar power output at the i-th historical time point, and t be the number of historical time points. Then the covariance matrix Σ is:

[0034]

[0035] 102. Based on the covariance matrix, determine the wind and solar power prediction error and obtain multiple initial wind and solar power prediction scenario sets. Use the wind and solar power prediction error to correct each initial wind and solar power prediction scenario set to obtain multiple wind and solar power prediction scenario sets. Each of the wind and solar power prediction scenario sets includes wind and solar power prediction data for each time point in the future time period.

[0036] For the embodiment of the present application, after the covariance matrix is determined, it is also necessary to determine the wind and light prediction output error. Based on this, step 102 specifically comprises: determining the installed capacity of the wind and light power generation equipment corresponding to the target power grid, and based on the installed capacity, performing standardization processing on the covariance matrix to obtain a standardized covariance matrix; performing decomposition of a lower triangular matrix on the standardized covariance matrix to obtain a lower triangular matrix; generating a plurality of uniformly distributed random numbers by using a pseudo-random number generator, converting each random number into a standard normal distribution to correspondingly obtain a plurality of standard normal random vectors, wherein the dimension of each standard normal random vector is the same as the vector dimension in the standardized covariance matrix; performing linear transformation on each standard normal random vector by using the lower triangular matrix, and based on the transformation result, determining a sample wind and light prediction output error, and based on the sample wind and light prediction output error, determining the wind and light prediction output error. The method for determining the wind and light prediction output error based on the sample wind and light prediction output error comprises: based on the sample wind and light prediction output error, constructing an error feature matrix; determining the element mean of each element in the error feature matrix, and subtracting each element in the error feature matrix from the element mean to obtain a centralized error matrix; determining the error covariance matrix corresponding to the centralized error matrix; performing eigenvalue decomposition on the error covariance matrix to obtain error matrix eigenvalues and error matrix eigenvectors; based on the size of the error matrix eigenvalues, selecting a preset number of error matrix eigenvectors in the error matrix eigenvectors, and determining the preset number of error matrix eigenvectors as principal component error eigenvectors, and based on the principal component error eigenvectors, determining the wind and light prediction output error.

[0037] Specifically, the covariance matrix is standardized according to the following formula:

[0038]

[0039] wherein, is the standardized covariance matrix, Cov[X] is the matrix element in the covariance matrix, and V[X] is the installed capacity. Then, the covariance matrix is decomposed into a lower triangular matrix:

[0040]

[0041] wherein, L is the lower triangular matrix, and at the same time, n standard normal random vectors Z (1) , (2) , (n) which are independent of the covariance matrix and have the same distribution are generated, and further, the standard normal samples are mapped into sample wind and light prediction output errors X (i) by using radial transformation:

[0042] X (i) = μ + LZ (i)

[0043] Wherein, μ is a radio conversion coefficient set according to actual demand, Z (i) is the i-th standard normal random vector. Further, the error feature matrix is constructed according to column stacking based on the sample wind light prediction output error, and the centralized error matrix of the error feature matrix is determined, the covariance matrix (error covariance matrix) of the centralized error matrix is calculated, and the eigenvalue decomposition of the error covariance matrix is as follows:

[0044] C e v i = λ i v i

[0045] Wherein, C e is the error covariance matrix, v i is the eigenvalue, and v i is the eigenvector. The eigenvalues are arranged in descending order, and the cumulative variance contribution rate z of the eigenvalues is calculated as follows:

[0046]

[0047] Wherein, i is the i-th eigenvalue, k is a preset number, and m is the total number of eigenvalues. The cumulative variance contribution rate of the largest k value is selected as the preset number, for example, if the cumulative contribution rate of the first three eigenvalues is 90%, and the first four is 96%, then k = 4. And the eigenvectors corresponding to the first four eigenvalues are determined as the principal component error feature vectors, and the error corresponding to the principal component error feature vectors in the error feature matrix is determined, and the error mean is taken as the wind light prediction output error. The present application embodiment fully considers the time correlation characteristics of the prediction error, so as to improve the determination accuracy of the error.

[0048] Further, a neural network model is used to predict the predicted wind and light output data of each wind power generation device and photovoltaic power generation device at each future time point. From the predicted wind and light output data of each wind power generation device and each photovoltaic power generation device at each time point, a plurality of initial wind and light prediction output scene sets are formed, wherein the initial wind and light prediction output scene set contains the predicted wind and light output data of different positions at each time point in the future time period. The predicted error is superimposed on the predicted output to obtain a wind and light prediction output scene set. The wind and light prediction output error is used to correct the wind and light prediction output data in the initial wind and light prediction output scene set in the embodiment of the present application. Finally, the error corrected wind and light prediction output data is used to realize the complementary scheduling of the power grid, so as to improve the scheduling accuracy of the electric energy in the power grid and ensure the stable operation of the power grid.

[0049] 103. Input each wind and solar predicted power output scenario set into the preset wind and solar complementary scheduling model to select the power output scenario set, obtain the target wind and solar predicted power output scenario set, and perform power scheduling according to the target wind and solar predicted power output scenario set.

[0050] In this embodiment of the invention, to improve the scheduling accuracy of the preset wind-solar complementary scheduling model, it is first necessary to train and construct the preset wind-solar complementary scheduling model. Based on this, the method includes: constructing a preset initial wind-solar complementary scheduling model; obtaining a sample dataset, wherein the sample dataset includes multiple sample wind-solar predicted output scene sets labeled with wind-solar output scene sets whose scheduling effects meet the requirements; dividing the sample dataset into a training set and a test set, using the training set to train the preset initial wind-solar complementary scheduling model, and using the test set to test the trained preset initial wind-solar complementary scheduling model, and finally using the trained preset initial wind-solar complementary scheduling model that meets the test conditions as the preset wind-solar complementary scheduling model.

[0051] Specifically, during model training, a pre-defined initial wind-solar hybrid scheduling model is first constructed, followed by the acquisition of a sample dataset. The dataset is ensured to contain all necessary files, and the data is converted to a format understandable by the pre-defined initial wind-solar hybrid scheduling model. Finally, the model is trained and tested. Specifically, the dataset can be divided first: using randomness or a specific strategy (such as stratified sampling), the sample dataset is divided into a training set and a test set. The model is then trained using the training set, and tested using the test set to evaluate its performance on unseen data. Metrics such as mCP, precision, and recall on the test set are calculated and recorded. If the model performance does not meet the requirements, the training phase can be returned for further iterations or adjustments. This process yields a pre-defined wind-solar hybrid scheduling model that meets the requirements.

[0052] Further, the preset wind and light complementary scheduling model includes a power grid power shortage target function, a power grid power abandonment target function, and a thermal power operation cost target function. After the preset wind and light complementary scheduling model is constructed, the model needs to be used to select a target wind and light prediction output scene set. Based on this, step 103 specifically includes: converting the power grid power shortage target function, the power grid power abandonment target function, and the thermal power operation cost target function into a single-target scene set selection function by using a hierarchical weight optimization method; obtaining scene set selection constraint conditions and determining scene set attribute data of each wind and light prediction output scene set, generation attribute data of remaining power generation devices, unit attribute data of thermal power generation devices, and regional power attribute information of a target power grid belonging to a region, wherein the scene set attribute data includes a wind and light power generation device quantity in each wind and light prediction output scene set, wind and light prediction output data at each time point in a future time period, the generation attribute data includes output data of the remaining power generation devices, the regional power attribute data includes mutual aid power and load, and the unit attribute data includes a unit quantity of the thermal power generation devices, a thermal power combustion cost coefficient, unit output data, and unit start-stop cost; based on the scene set selection constraint conditions, the scene set attribute data of each wind and light prediction output scene set, the generation attribute data, the unit attribute data, and the regional power attribute information are substituted into the single-target scene set selection function respectively to obtain an evaluation value corresponding to each wind and light prediction output scene set, and based on the evaluation value, the target wind and light prediction output scene set is determined in each wind and light prediction output scene set. The method for converting the power grid power shortage target function, the power grid power abandonment target function, and the thermal power operation cost target function into the single-target scene set selection function by using the hierarchical weight optimization method includes: the power grid power shortage target function is set as wherein, the power grid power abandonment target function is set as wherein, the thermal power operation cost target function is set as wherein, MinF1 is power shortage data, MinF2 is power abandonment data, MinF3 is thermal power operation cost, n is a wind and light power generation device identifier in a wind and light prediction output scene set, N is a total quantity of wind and light power generation devices, t is a future time period identifier, T is a total length of a future time period, p is a region identifier of a target power grid, M is a total quantity of regions of the target power grid, loe n,t,p , foe n,t,p are power shortage and power abandonment of the p region at time point t under the nth scene set respectively, is wind and light output of the p region at time point t under the scene set n, are hydropower output, thermal power output, nuclear power output, pumped storage output, and direct current output of the p region at time point t respectively, is total output of other power sources of the p region at time point t, The mutual aid electricity amount fed into the p area at the time point t for the nth scene set, L t,p The load of the p area at the time point t, k and K are the serial number and the number of thermal power units respectively, a, b, and c are the thermal power combustion cost coefficients, P k The output of the k unit of the thermal power equipment, S k The start-stop cost of the k unit; the power grid power shortage objective function, the power grid power abandonment objective function, and the thermal power operation cost objective function are processed by a range algorithm, and the processed power grid power shortage objective function, the processed power grid power abandonment objective function, and the processed thermal power operation cost objective function are obtained; the processed power grid power shortage objective function, the processed power grid power abandonment objective function, and the processed thermal power operation cost objective function are divided into different levels, and the same level weight coefficient of each objective function in the same level and the level weight coefficient of the objective function between different levels are determined; based on the same level weight coefficient and the level weight coefficient, the processed power grid power shortage objective function, the processed power grid power abandonment objective function, and the processed thermal power operation cost objective function are weighted and aggregated, and the single-target scene set selection function is obtained.

[0053] Specifically, a preset wind-solar complementary scheduling model is constructed with the minimum square sum of power grid power shortage and power abandonment and the minimum thermal power operation cost as the target, and the preset wind-solar complementary scheduling model is as follows:

[0054] The objective function 1 (the preset wind-solar complementary scheduling model) is as follows:

[0055]

[0056] The objective function 2 (the power grid power abandonment objective function) is as follows:

[0057]

[0058] Among them,

[0059]

[0060] The objective function 3 (the thermal power operation cost objective function) is as follows:

[0061]

[0062] The scene set selection constraint condition is as follows:

[0063] The water level storage capacity constraint is as follows:

[0064]

[0065] Among them, Z i,tV i,t is the storage capacity of reservoir i on day t, are the lower and upper limits of the water level of reservoir i on day t, are the lower and upper limits of the storage capacity of reservoir i on day t.

[0066] The water balance equation is:

[0067] V i,t+1 = V i,t + (q i,t + R i-1 - R i,t ) Δt t

[0068] wherein R i,t = Q i,t + S i,t . q i,t , R i,t , Q i,t , S i,t are the interval flow, outflow, power generation flow and abandoned water flow of reservoir i on day t, V i,t+1 is the water volume of reservoir i on day t+1, V i,t is the water volume of reservoir i on day t, and Δt is the time change.

[0069] The power generation flow constraint is:

[0070]

[0071] wherein are the lower and upper limits of the power generation flow of reservoir i on day t, Q i,t is the power generation of reservoir i on day t.

[0072] The outflow constraint is:

[0073]

[0074] wherein are the lower and upper limits of the outflow of reservoir i on day t, R i,t is the outflow of reservoir i on day t.

[0075] The hydropower output constraint is:

[0076]

[0077] wherein and are the lower and upper limits of the output of reservoir i, is the output data of reservoir i.

[0078] The thermal power unit output constraint is:

[0079]

[0080] where, and are the minimum and maximum output of thermal power unit i, respectively, is the output data of thermal power unit i.

[0081] Thermal power total output constraint:

[0082] The total thermal power output in the future time should be limited by the fuel, and the consumption should not exceed the supply plan:

[0083]

[0084] where, is the output of thermal power unit i on day t, I th is the number of thermal power units; is the upper limit of the total thermal power output in T period.

[0085] Minimum start-stop time of thermal power unit:

[0086] The start-stop operation of large thermal power units involves the complex dynamic characteristics of the boiler thermal system. It usually takes 24 to 48 hours from the cold standby state to grid-connected operation, and this process may be further extended. In addition, a single start-stop operation of a thermal power unit will generate significant economic costs. From the perspective of economic operation of the power system, the frequent start-stop operation mode is not feasible. Based on the physical characteristics and actual operation requirements of thermal power units, the minimum continuous operation time constraint of units in a day must be considered in the scheduling model:

[0087]

[0088] where, U i,k is the state variable of the i-th unit on the k-th day. If the unit i is on, U i,k is 1, otherwise U i,k is 0; Y i,t is the start variable of the i-th unit on the t-th day. If the unit i is from shutdown to startup, Y i,t is 1, otherwise Y i,t is 0; Z i,t is the shutdown variable of the i-th unit on the t-th day. If the unit i is from startup to shutdown, Z i,t is 1, otherwise Z i,t is 0, and T is the total period length of the future time period.

[0089] Minimum number of thermal power units in operation:

[0090] In order to maintain the safe operation of the power grid, the minimum number of units in operation requirement must be met on the generation side:

[0091]

[0092] wherein, N th,min is the minimum number of thermal power units to be started, U i,t is the number of i-th thermal power units at t time, I th is the total number of thermal power units.

[0093] Nuclear power output constraints:

[0094]

[0095] wherein, is the output of nuclear power sent to the region on the t day, is the actual output process of the grid regulating nuclear power.

[0096] Pumped storage output constraints:

[0097]

[0098] wherein, is the output of pumped storage sent to the p region on the t day, is the actual output process of the grid regulating pumped storage.

[0099] Inter-regional mutual constraints:

[0100] The gradual increase of wind and solar power generation ratio is necessary for solving the future new energy consumption and supply problems, and full coordination of regional power grid and inter-regional mutual aid is necessary.

[0101]

[0102] wherein, is the p region mutual electricity on the t day, is the p region mutual electricity sent to the s region on the t day, S is other regions except the sending end, is the p region power generation on the t day, is the p region power generation on the t day under the nth scenario set, is the pumped storage output sent to the p region on the t day, is the pumped storage output sent to the p region on the t day, is the p region wind and solar output on the t day, L t,p is the error output.

[0103] Power station distribution constraints:

[0104] All grid regulating power sources should meet the regional power grid distribution requirements

[0105]

[0106] where P i,t is the power output of the grid regulation power, P i,t,p is the power output of the i-th power plant to the p-th region on the t-th day, k is the grid regulation power distribution ratio specified by the grid, and M is the total number of regions.

[0107] The multi-objective problem is converted into a single objective problem by using a hierarchical weight optimization method, and the specific steps are as follows:

[0108] The multi-objective optimization problem is modeled as minF(x)=[f1(x),f2(x),...,f m (x)] T , where minF(x) is a single objective scenario selection function, f1(x) is a grid power shortage objective function, f2(x) is a grid power curtailment objective function, and f m (x) is a thermal power operation cost objective function.

[0109] The range method is used to carry out dimensionless processing on each objective function where f ic (x) is each function after processing, f i max (x) is the maximum value of the f i (x) function, and f i min (x) is the minimum value of the f i (x) function. The priority of the objective function is determined by expert experience or production requirements. Let the objective priority be P={p1,p2,...,p m}, where P i is the priority of the i-th objective function, p1 has the highest priority, and p m has the lowest priority. According to the priority P, the m objective functions are divided into L levels (L≤m), and the resulting levels are F={F 1 ,F 2 ,...,F L}. F is a different level, F i is the i-th objective function in the level F, the priority of the high-level objective is strictly higher than that of the bottom-level objective, and the priority difference of the objectives in the same level is small. For each objective function in different levels F, an exponential weight distribution is used to reflect the priority through the order of magnitude difference. The specific rules are as follows: high priority objective (F1): weight coefficient ω1=M, M is a large positive number, which ensures that its optimization contribution far exceeds that of other objectives, and the next priority objective (F 2 ~F L ): weight coefficient ω j =ε·M, ε is an exponential decay coefficient (a small positive number), which ensures that its contribution is only ε times that of the high priority. For each objective function in the same level F, the relative weight is determined by using the layer analysis method The same level weight coefficients are same level weight coefficients, wherein, The same level weight coefficients are same level weight coefficients of different target functions respectively, n is the total number of target functions, and the weight coefficients between different levels, i.e., inter-level weight coefficients, can also be determined according to actual requirements. Then, an aggregation function is constructed to ensure the priority of key target functions and the coordination of multiple target functions, and a single-target scene set selection function Φ(x) is obtained:

[0110]

[0111] The inter-level weight coefficients are inter-level weight coefficients of functions f1(x), f2(x),..., f L The inter-level weight coefficients are inter-level weight coefficients of functions f1(x), f2(x),..., f m The inter-level weight coefficients are inter-level weight coefficients of functions f1(x), f2(x),..., f The same level weight coefficients are same level weight coefficients of corresponding functions. Further, based on scene set selection constraint conditions, scene set attribute data, power generation attribute data, unit attribute data and regional power attribute information of each wind-solar predicted output scene set are respectively substituted into the single-target scene set selection function to obtain an evaluation value corresponding to each wind-solar predicted output scene set, and finally a wind-solar predicted output scene set corresponding to the maximum evaluation value is determined as a target wind-solar predicted output scene set. Then, power dispatching in the power grid is performed according to wind-solar output data of each time point in the target wind-solar predicted output scene set. In the power grid dispatching process, the present application fully considers multiple factors such as curtailed power, power shortage, cost and the like, and can further improve the dispatching accuracy of the power grid and ensure the collaborative and stable operation of the power grid.

[0112] According to the power grid multi-energy complementary scheduling method based on wind and light output uncertainty provided by the application, compared with the current method of directly scheduling electric energy based on predicted wind and light output capacity, the application obtains actual wind and light output data and predicted wind and light output data corresponding to a target power grid at multiple historical time points in a historical time period, constructs a covariance matrix based on the difference between the actual wind and light output data and the predicted wind and light output data at the same historical time point, determines wind and light prediction output errors based on the covariance matrix, and obtains multiple initial wind and light prediction output scene sets, respectively corrects each initial wind and light prediction output scene set by using the wind and light prediction output errors to obtain multiple wind and light prediction output scene sets, wherein each wind and light prediction output scene set includes wind and light prediction output data at each time point in a future time period. Finally, each wind and light prediction output scene set is input into a preset wind and light complementary scheduling model for output scene set selection to obtain a target wind and light prediction output scene set, and electric energy is scheduled according to the target wind and light prediction output scene set. Thus, the covariance matrix is constructed by using actual wind and light output data and predicted wind and light output data with time sequence, the wind and light prediction output errors with time-varying characteristics are determined based on the covariance matrix, the wind and light prediction output data in the initial wind and light prediction output scene set is corrected by using the wind and light prediction output errors, and finally the wind and light prediction output data after error correction is used to realize the complementary scheduling of the power grid, so that the scheduling accuracy of electric energy in the power grid can be improved, and the stable operation of the power grid can be ensured.

[0113] Further, in order to better illustrate the above process of the power grid multi-energy complementary scheduling based on wind and light output uncertainty, as a refinement and expansion of the above embodiment, the embodiment of the application provides another power grid multi-energy complementary scheduling method based on wind and light output uncertainty, as shown in Figure 2 The method comprises the following steps.

[0114] 201. Obtain actual wind and light output data and predicted wind and light output data corresponding to a target power grid at multiple historical time points in a historical time period, and construct a covariance matrix based on the difference between the actual wind and light output data and the predicted wind and light output data at the same historical time point.

[0115] 202. Determine wind and light prediction output errors based on the covariance matrix, obtain multiple initial wind and light prediction output scene sets, and correct each initial wind and light prediction output scene set by using the wind and light prediction output errors to obtain multiple wind and light prediction output scene sets, wherein each wind and light prediction output scene set includes wind and light prediction output data at each time point in a future time period.

[0116] Specifically, a wind and light typical output scene considering time-varying characteristics of wind and light is generated, and the specific steps are as follows: according to the actual output data of wind and light in different prediction periods (future time periods) of the regional power grid and the predicted output data , the wind and light processing prediction error X of each regional power grid is calculated i,m ; the wind and light prediction error vector X=(X1, X2,..., X t ) T , X t is the prediction error random variable on the tth day in the prediction period, and t is the length of the prediction period; the wind and light prediction output error covariance matrix is calculated; since the new energy prediction error is still affected by the installed capacity, the covariance matrix is standardized; according to the standardized covariance matrix and the multivariate normal sampling method, the wind and light prediction output error is sampled; the wind and light prediction output error is superimposed on the wind and light prediction output to generate an output scene set.

[0117] 203, respectively determine the wind power generation limit of the wind power generation equipment, the photovoltaic power generation limit of the photovoltaic power generation equipment, and the thermal power generation limit of the thermal power generation equipment.

[0118] 204, based on the wind power generation limit, the photovoltaic power generation limit, and the thermal power generation limit, scene set screening is performed on each wind and light prediction output scene set to obtain a wind and light prediction output scene set that meets the output demand.

[0119] Specifically, based on the wind power generation limit, the photovoltaic power generation limit, and the thermal power generation limit, scene set screening is performed on each wind and light prediction output scene set, that is, scene sets exceeding the limit are removed.

[0120] 205, input the wind and light prediction output scene set that meets the output demand into a preset wind and light complementary dispatching model to select the output scene set, obtain a target wind and light prediction output scene set, and perform electric energy dispatching according to the target wind and light prediction output scene set.

[0121] Specifically, the wind and light prediction output scene set that meets the output demand is input into the preset wind and light complementary dispatching model, and the model can directly output a target wind and light prediction output scene set that meets the dispatching demand, and finally the target power grid is dispatched according to the wind power output data and the photovoltaic output data in the target wind and light prediction output scene set. Dispatching of dispatching power sources such as thermal power, hydropower, nuclear power, and pumped storage.

[0122] For example, in recent years, with the rapid growth of new energy installation in the power grid area, the fluctuation of new energy output during the day has increasingly highlighted the impact on the safe and stable operation of the power grid, and higher requirements have been put forward for the regulation capacity of the grid dispatching power. The embodiments of the present application perform multi-energy complementary dispatching on the regional grid dispatching power (thermal power, hydropower, nuclear power, pumped storage) and the remaining regional wind and light power stations. The remaining regional grid dispatching power does not participate in the optimization calculation, and the typical operation mode is given according to the actual operation data in 2024, and the basic parameters are shown in Table 1. Other parameters are mainly the combustion cost, start-stop cost and shortest start-stop time. The combustion cost coefficients are a=0.02, b=15 and c=350 respectively; different start-stop costs are set for different installed capacity units, and the start-stop cost of 1000MW unit is 3 million yuan, the start-stop cost of 700MW unit is 1.5 million yuan, and the start-stop cost of the remaining units is 1.2 million yuan; the shortest start-stop time is 10 days.

[0123] Table 1 Basic parameters of power stations

[0124]

[0125] Figure 3 、 Figure 4 For the generated 5 groups of new energy (wind and light output) scenes and actual output and predicted output, in a short prediction period, the wind and light output prediction accuracy is high, and the scene fluctuation range is small; with the extension of the prediction period, the fluctuation range of the scene gradually increases, resulting in greater balancing pressure for the power grid. The current prediction result presents a high output overestimation trend on the whole, reflecting the deficiency of prediction accuracy. If the power grid dispatching plan is only based on the predicted output, the system power shortage risk will be significantly increased. Case 1: 360 days from January 1, 2024 to December 25, 2024 are divided into 36 time periods, numbered as time periods 1-36 respectively, and rolling calculation is performed, and the results are shown in Tables 2 and 3:

[0126] Table 2 Abandonment of each region under two dispatching modes (January 1 to January 10)

[0127]

[0128] Table 3 Number of better performance of abandonment of each region under two dispatching modes

[0129]

[0130] Figure 5 、 Figure 6 The adjustment range and abandonment under the traditional method are shown, Figure 7 , Figure 8The adjustment range and rejection condition under the method of the present application are shown. After further adjustment on the basis of the plan from January 1 to January 10, the maximum power shortage of the traditional method is 120 million kWh, the total power shortage is 670 million kWh, the maximum power shortage of the embodiment of the present application is 48 million kWh, and the total power shortage is 180 million kWh; in the foreseeable period, the proportion of completely balanced days is 20% for the traditional model and 50% for the embodiment of the present application; after adjustment by both methods, there is no power rejection, and new energy is further consumed. It can be seen that the embodiment of the present application still performs significantly better than the traditional method, and has guiding significance for the safe operation of the power grid.

[0131] According to the power grid multi-energy complementary dispatching method based on wind and light output uncertainty provided by the present application, compared with the current method of directly dispatching electric energy based on the predicted wind and light output capacity, the present application obtains actual wind and light output data and predicted wind and light output data corresponding to a target power grid at a plurality of historical time points in a historical time period, constructs a covariance matrix based on the difference between the actual wind and light output data and the predicted wind and light output data at the same historical time point, determines wind and light prediction output errors based on the covariance matrix, and obtains a plurality of initial wind and light prediction output scene sets, respectively corrects each initial wind and light prediction output scene set by using the wind and light prediction output errors to obtain a plurality of wind and light prediction output scene sets, wherein each wind and light prediction output scene set includes wind and light prediction output data at each time point in a future time period; finally, each wind and light prediction output scene set is input into a preset wind and light complementary dispatching model to select an output scene set, obtain a target wind and light prediction output scene set, and perform electric energy dispatching according to the target wind and light prediction output scene set. Thus, the covariance matrix is constructed by using the actual wind and light output data and the predicted wind and light output data with time sequence, the wind and light prediction output errors with time-varying characteristics are determined based on the covariance matrix, the wind and light prediction output data in the initial wind and light prediction output scene set is corrected by using the wind and light prediction output errors, and finally the wind and light prediction output data after error correction is used to realize the complementary dispatching of the power grid, so as to improve the dispatching accuracy of electric energy in the power grid and ensure the stable operation of the power grid.

[0132] Further, as a specific implementation of Figure 1 , the embodiment of the present application provides a power grid multi-energy complementary dispatching device based on wind and light output uncertainty, as shown in Figure 9 , the device comprises an acquisition unit 31, a correction unit 32, and a selection unit 33.

[0133] The acquisition unit 31 can be configured to acquire actual wind and light output data and predicted wind and light output data corresponding to a target power grid at a plurality of historical time points in a historical time period, and construct a covariance matrix based on a difference between the actual wind and light output data and the predicted wind and light output data at the same historical time point.

[0134] The correction unit 32 can be configured to determine wind and light prediction output errors based on the covariance matrix, acquire a plurality of initial wind and light prediction output scene sets, and correct each of the initial wind and light prediction output scene sets using the wind and light prediction output errors to obtain a plurality of wind and light prediction output scene sets, wherein each of the wind and light prediction output scene sets includes wind and light prediction output data at each time point in a future time period.

[0135] The selection unit 33 can be configured to input each of the wind and light prediction output scene sets into a preset wind and light complementary scheduling model to select an output scene set, obtain a target wind and light prediction output scene set, and perform power scheduling according to the target wind and light prediction output scene set.

[0136] In a specific application scenario, in order to determine the wind and light prediction output errors, as shown in the following formula (1), the correction unit 32 can be configured to determine the wind and light prediction output errors based on the covariance matrix and the actual wind and light output data. Figure 10 The correction unit 32 includes a standardization module 321, a decomposition module 322, a generation module 323, and a first determination module 324.

[0137] The standardization module 321 can be configured to determine the installed capacity of the wind and light power generation equipment corresponding to the target power grid, and perform standardization processing on the covariance matrix based on the installed capacity to obtain a standardized covariance matrix.

[0138] The decomposition module 322 can be configured to perform lower triangular matrix decomposition on the standardized covariance matrix to obtain a lower triangular matrix.

[0139] The generation module 323 can be configured to generate a plurality of uniformly distributed random numbers using a pseudo-random number generator, convert each of the random numbers into a standard normal distribution, and correspondingly obtain a plurality of standard normal random vectors, wherein the dimension of each standard normal random vector is the same as the dimension of a vector in the standardized covariance matrix.

[0140] The first determination module 324 can be configured to perform linear transformation on each of the standard normal random vectors using the lower triangular matrix, determine sample wind and light prediction output errors based on the transformation results, and determine wind and light prediction output errors based on the sample wind and light prediction output errors.

[0141] In a specific application scenario, in order to determine the wind and light prediction output error, the first determination module 324 can be specifically used for constructing an error feature matrix based on the sample wind and light prediction output error; determining the element mean of each element in the error feature matrix, and subtracting each element in the error feature matrix from the element mean to obtain a centralized error matrix; determining the error covariance matrix corresponding to the centralized error matrix; performing eigenvalue decomposition on the error covariance matrix to obtain error matrix eigenvalues and error matrix eigenvectors; based on the size of the error matrix eigenvalues, selecting a preset number of error matrix eigenvectors in the error matrix eigenvectors, and determining the preset number of error matrix eigenvectors as principal component error eigenvectors, and determining the wind and light prediction output error based on the principal component error eigenvectors.

[0142] In a specific application scenario, in order to construct a preset wind and light complementary scheduling model, the device further includes a construction unit 34.

[0143] The construction unit 34 can be used to construct a preset initial wind and light complementary scheduling model; obtain a sample data set, wherein the sample data set includes a plurality of sample wind and light prediction output scene sets with scheduling effect meeting the requirements of wind and light output scene set labels; divide the sample data set into a training set and a test set, train the preset initial wind and light complementary scheduling model using the training set, and test the trained preset initial wind and light complementary scheduling model using the test set, and finally use the trained preset initial wind and light complementary scheduling model that meets the test conditions as the preset wind and light complementary scheduling model.

[0144] In a specific application scenario, the preset wind and light complementary scheduling model includes a power grid power shortage target function, a power grid power abandonment target function, and a thermal power operation cost target function; in order to obtain a target wind and light prediction output scene set, the selection unit 33 includes a conversion module 331, a second determination module 332, and a prediction module 333.

[0145] The conversion module 331 can be used to convert the power grid power shortage target function, the power grid power abandonment target function, and the thermal power operation cost target function into a single-target scene set selection function using a hierarchical weight optimization method.

[0146] The second determination module 332 can be used to obtain a scenario set selection constraint condition, and determine scenario set attribute data of each wind-solar predicted output scenario set, power generation attribute data of the remaining power generation devices, unit attribute data of the thermal power generation devices, and regional power attribute information of a region to which the target power grid belongs. The scenario set attribute data includes the number of wind-solar power generation devices in each wind-solar predicted output scenario set, wind-solar predicted output data at each time point in a future time period. The power generation attribute data includes output data of the remaining power generation devices. The regional power attribute data includes mutual aid power and load. The unit attribute data includes the number of thermal power generation units, thermal power combustion cost coefficients, unit output data, and unit start-stop costs.

[0147] The prediction module 333 can be used to substitute the scenario set attribute data of each wind-solar predicted output scenario set, the power generation attribute data, the unit attribute data, and the regional power attribute information into the single-target scenario set selection function based on the scenario set selection constraint condition, to obtain an evaluation value corresponding to each wind-solar predicted output scenario set, and determine the target wind-solar predicted output scenario set in each wind-solar predicted output scenario set based on the evaluation value. In a specific application scenario, in order to determine the single-target scenario set selection function, the conversion module 331 can be specifically used to set the power grid power shortage target function as wherein,

[0148] The power grid power curtailment target function is set as wherein, The thermal power operation cost target function is set as wherein, MinF1 is power shortage data, MinF2 is power curtailment data, MinF3 is thermal power operation cost, n is a wind-solar power generation device identifier in a wind-solar predicted output scenario set, N is the total number of wind-solar power generation devices, t is a future time period identifier, T is the total length of the future time period, p is a region identifier of the target power grid, M is the total number of regions of the target power grid, loe n,t,p , foe n,t,p are power shortage and power curtailment of the pth region at the nth scenario set and time point t, respectively, is wind-solar output of the pth region at the nth scenario set and time point t, are hydropower output, thermal power output, nuclear power output, pumped storage output, and direct current output of the pth region at time point t, respectively, is total output of other power sources of the pth region at time point t, is mutual aid power fed into the pth region at the nth scenario set and time point t, L t,p is load of the pth region at time point t, k and K are thermal power unit serial numbers and the number of units, respectively, a, b, and c are thermal power combustion cost coefficients, and P is power.k is the output of the kth thermal power generating unit, S k is the start-stop cost of the kth thermal power generating unit, the power grid power shortage target function, the power grid curtailment target function, and the thermal power operation cost target function are processed by a range algorithm, respectively, to obtain a processed power grid power shortage target function, a processed power grid curtailment target function, and a processed thermal power operation cost target function, the processed power grid power shortage target function, the processed power grid curtailment target function, and the processed thermal power operation cost target function are divided into different levels, respectively, and a same-level weight coefficient of each target function in the same level and a level-to-level weight coefficient of the target functions between different levels are determined, and the processed power grid power shortage target function, the processed power grid curtailment target function, and the processed thermal power operation cost target function are aggregated by weighting based on the same-level weight coefficient and the level-to-level weight coefficient, to obtain the single-target scenario set selection function.

[0149] In a specific application scenario, in order to screen the wind-solar predicted output scenario set, the device further includes a screening unit 35.

[0150] The screening unit 35 can be used to determine a wind power generation limit value of a wind power generating device, a photovoltaic power generation limit value of a photovoltaic power generating device, and a thermal power generation limit value of a thermal power generating device, respectively, and perform scenario set screening on each wind-solar predicted output scenario set based on the wind power generation limit value, the photovoltaic power generation limit value, and the thermal power generation limit value, to obtain a wind-solar predicted output scenario set meeting the output requirement.

[0151] The selection unit 33 can also be used to input the wind-solar predicted output scenario set meeting the output requirement into a preset wind-solar complementary dispatching model to select the output scenario set, to obtain a target wind-solar predicted output scenario set.

[0152] It should be noted that other corresponding descriptions of the functions of the device for multi-energy complementary dispatching of a power grid based on wind-solar output uncertainty provided in the embodiments of the present application can be referred to the corresponding descriptions of the method shown in Figure 1 and will not be described here.

[0153] Based on the above, as Figure 1According to the method, the embodiment of the present application also provides a computer readable storage medium, which stores a computer program, and the program is executed by a processor to realize the following steps: obtaining actual wind and light output data and predicted wind and light output data corresponding to a target power grid at a plurality of historical time points in a historical time period, constructing a covariance matrix based on a difference between the actual wind and light output data and the predicted wind and light output data at the same historical time point, determining wind and light prediction output errors based on the covariance matrix, and obtaining a plurality of initial wind and light prediction output scene sets, correcting each of the initial wind and light prediction output scene sets by using the wind and light prediction output errors to obtain a plurality of wind and light prediction output scene sets, wherein each of the wind and light prediction output scene sets includes wind and light prediction output data at each time point in a future time period, inputting each of the wind and light prediction output scene sets into a preset wind and light complementary scheduling model to select an output scene set, obtaining a target wind and light prediction output scene set, and performing electric energy scheduling according to the target wind and light prediction output scene set.

[0154] According to the method and the device, the embodiment of the present application also provides a computer device, which includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the memory and the processor are arranged on a bus, and the processor realizes the following steps when executing the program: obtaining actual wind and light output data and predicted wind and light output data corresponding to a target power grid at a plurality of historical time points in a historical time period, constructing a covariance matrix based on a difference between the actual wind and light output data and the predicted wind and light output data at the same historical time point, determining wind and light prediction output errors based on the covariance matrix, and obtaining a plurality of initial wind and light prediction output scene sets, correcting each of the initial wind and light prediction output scene sets by using the wind and light prediction output errors to obtain a plurality of wind and light prediction output scene sets, wherein each of the wind and light prediction output scene sets includes wind and light prediction output data at each time point in a future time period, inputting each of the wind and light prediction output scene sets into a preset wind and light complementary scheduling model to select an output scene set, obtaining a target wind and light prediction output scene set, and performing electric energy scheduling according to the target wind and light prediction output scene set. Figure 1 Figure 9 According to the method and the device, the embodiment of the present application also provides a computer device, which includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the memory and the processor are arranged on a bus, and the processor realizes the following steps when executing the program: obtaining actual wind and light output data and predicted wind and light output data corresponding to a target power grid at a plurality of historical time points in a historical time period, constructing a covariance matrix based on a difference between the actual wind and light output data and the predicted wind and light output data at the same historical time point, determining wind and light prediction output errors based on the covariance matrix, and obtaining a plurality of initial wind and light prediction output scene sets, correcting each of the initial wind and light prediction output scene sets by using the wind and light prediction output errors to obtain a plurality of wind and light prediction output scene sets, wherein each of the wind and light prediction output scene sets includes wind and light prediction output data at each time point in a future time period, inputting each of the wind and light prediction output scene sets into a preset wind and light complementary scheduling model to select an output scene set, obtaining a target wind and light prediction output scene set, and performing electric energy scheduling according to the target wind and light prediction output scene set. Figure 11

[0155] ​​By the technical scheme of the present application, the actual wind and light output data and the predicted wind and light output data corresponding to the target power grid at multiple historical time points in a historical time period are acquired, a covariance matrix is constructed based on the difference between the actual wind and light output data and the predicted wind and light output data at the same historical time point, the wind and light prediction output error is determined based on the covariance matrix, and multiple initial wind and light prediction output scene sets are acquired, the wind and light prediction output error is used to correct each initial wind and light prediction output scene set respectively, and multiple wind and light prediction output scene sets are obtained, wherein each wind and light prediction output scene set includes the wind and light prediction output data at each time point in a future time period; finally, each wind and light prediction output scene set is input into a preset wind and light complementary scheduling model for output scene set selection to obtain a target wind and light prediction output scene set, and electric energy scheduling is performed according to the target wind and light prediction output scene set. Thus, the covariance matrix is constructed by using the actual wind and light output data and the predicted wind and light output data with time sequence, the wind and light prediction output error with time-varying characteristics is determined based on the covariance matrix, the wind and light prediction output data in the initial wind and light prediction output scene set is corrected by using the wind and light prediction output error, and finally the wind and light prediction output data after error correction is used to realize the complementary scheduling of the power grid, so that the scheduling accuracy of the electric energy in the power grid can be improved, and the stable operation of the power grid can be ensured.

[0156] Obviously, those skilled in the art should understand that the modules or steps of the present application described above can be realized by general computing devices, which can be concentrated on a single computing device or distributed on a network composed of multiple computing devices, and alternatively, they can be realized by program codes executable by computing devices, so that they can be stored in storage devices and executed by computing devices, and in some cases, the steps shown or described can be executed in different order, or they can be manufactured into individual integrated circuit modules or multiple modules or steps into a single integrated circuit module. Thus, the present application is not limited to any specific combination of hardware and software.

[0157] The above only describes the preferred embodiments of the present application and is not used to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A grid multi-energy complementary dispatch method based on the uncertainty of wind and solar power output, characterized in that, include: Obtain actual and predicted wind and solar power output data for the target power grid at multiple historical time points within a historical period. Construct a covariance matrix based on the difference between the actual and predicted wind and solar power output data at the same historical time point. Based on the covariance matrix, the wind and solar power prediction error is determined, and multiple initial wind and solar power prediction scenario sets are obtained. The wind and solar power prediction error is used to correct each initial wind and solar power prediction scenario set to obtain multiple wind and solar power prediction scenario sets. Each wind and solar power prediction scenario set includes wind and solar power prediction data for each time point in a future time period. Each of the predicted wind and solar power output scenarios is input into a preset wind-solar complementary scheduling model to select the power output scenario set, thereby obtaining the target predicted wind and solar power output scenario set, and power scheduling is performed according to the target predicted wind and solar power output scenario set.

2. The method according to claim 1, characterized in that, Based on the covariance matrix, the wind and solar power prediction output error is determined, including: Determine the installed capacity of wind and solar power generation equipment corresponding to the target power grid, and based on the installed capacity, standardize the covariance matrix to obtain a standardized covariance matrix; The standardized covariance matrix is ​​decomposed into a lower triangular matrix to obtain the lower triangular matrix; Multiple uniformly distributed random numbers are generated using a pseudo-random number generator. Each of the random numbers is then converted into a standard normal distribution, resulting in multiple standard normal random vectors. The dimension of each standard normal random vector is the same as the vector dimension in the standardized covariance matrix. The lower triangular matrix is ​​used to perform a linear transformation on each of the standard normal random vectors, and the sample wind and solar power prediction output error is determined based on the transformation results. Based on the sample wind and solar power prediction output error, the wind and solar power prediction output error is determined.

3. The method according to claim 2, characterized in that, Based on the sample wind and solar power prediction output error, the wind and solar power prediction output error is determined, including: Based on the wind and solar power prediction error of the sample, an error feature matrix is ​​constructed; Determine the mean of each element in the error feature matrix, and subtract the mean from each element in the error feature matrix to obtain the centered error matrix; Determine the error covariance matrix corresponding to the centered error matrix; The error covariance matrix is ​​decomposed into eigenvalues ​​to obtain the eigenvalues ​​and eigenvectors of the error matrix. Based on the magnitude of the eigenvalues ​​of the error matrix, a preset number of error matrix eigenvectors are selected from the error matrix eigenvectors, and the preset number of error matrix eigenvectors are determined as principal component error eigenvectors. The wind and solar power prediction output error is determined based on the principal component error eigenvectors.

4. The method according to claim 1, characterized in that, Before inputting each of the predicted wind and solar power output scenarios into a preset wind-solar complementary scheduling model to select the power output scenario set and obtaining the target predicted wind and solar power output scenario set, the method further includes: Construct a pre-defined initial wind-solar hybrid scheduling model; Obtain a sample dataset, wherein the sample dataset includes multiple sample wind and solar power output scene sets labeled with wind and solar power output scene sets whose scheduling effect meets the requirements; The sample dataset is divided into a training set and a test set. The preset initial wind-solar complementary scheduling model is trained using the training set, and the trained preset initial wind-solar complementary scheduling model is tested using the test set. Finally, the trained preset initial wind-solar complementary scheduling model that meets the test conditions is taken as the preset wind-solar complementary scheduling model.

5. The method according to claim 1, characterized in that, The preset wind-solar hybrid dispatch model includes a power grid shortage objective function, a power grid curtailment objective function, and a thermal power operation cost objective function. Each of the predicted wind and solar power output scenarios is input into a preset wind-solar complementary scheduling model to select the power output scenario set, thereby obtaining the target predicted wind and solar power output scenario set, including: The objective functions of power grid shortage, power grid curtailment, and thermal power operation cost are transformed into a single-objective scenario set selection function using a hierarchical weight optimization method. Obtain the scenario set selection constraints and determine the scenario set attribute data for each wind and solar power predicted output scenario set, the power generation attribute data of other power generation equipment, the unit attribute data of thermal power generation equipment, and the regional power attribute information of the target power grid area. The scenario set attribute data includes the number of wind and solar power generation equipment in each wind and solar power predicted output scenario set and the wind and solar power predicted output data at each time point in the future time period. The power generation attribute data includes the output data of other power generation equipment. The regional power attribute data includes mutual assistance power and load. The unit attribute data includes the number of thermal power generation units, thermal power combustion cost coefficient, unit output data, and unit start-up and shutdown costs. Based on the scenario set selection constraints, the scenario set attribute data, the power generation attribute data, the generator set attribute data, and the regional power attribute information of each wind and solar power predicted output scenario set are substituted into the single-objective scenario set selection function to obtain the evaluation value corresponding to each wind and solar power predicted output scenario set. Based on the evaluation value, the target wind and solar power predicted output scenario set is determined in each wind and solar power predicted output scenario set.

6. The method according to claim 5, characterized in that, The objective functions for power grid shortage, power grid curtailment, and thermal power operation cost are transformed into a single-objective scenario set selection function using a hierarchical weighted optimization approach, including: The objective function for power grid shortage is set as follows: in, The objective function for power grid curtailment is set as follows: in, The objective function for thermal power plant operating costs is set as follows: Wherein, MinF1 represents power shortage data, MinF2 represents power curtailment data, MinF3 represents thermal power operating costs, n represents the identifier of wind and solar power generation equipment in the wind and solar power prediction output scenario set, N represents the total number of wind and solar power generation equipment, t represents the identifier of the future time period, T represents the total length of the future time period, p represents the regional identifier of the target power grid, M represents the total number of regions in the target power grid, and loe n,t,p foe n,t,p These represent the power shortage and power abandonment in region p at time point t in the nth scenario set. Contribute to the scenery of region p at time point t under scene set n. These represent the hydropower output, thermal power output, nuclear power output, pumped storage power output, and DC power output in region p at time point t. The total output of other power sources in region p at time point t. The mutual assistance power input to region p at time point t in the nth scenario set, L t,p Let P be the load in region p at time point t, k and K be the thermal power unit serial number and number of units respectively, a, b, and c be the thermal power combustion cost coefficients respectively, and P be the load in region p. k For the output of unit k of the thermal power generation equipment, S k The start-up and shutdown cost of unit k; The range algorithm is used to perform dimensionless processing on the target functions of power grid shortage, power grid curtailment, and thermal power operation cost, respectively, to obtain the processed target functions of power grid shortage, power grid curtailment, and thermal power operation cost. The processed target functions for power grid shortage, power grid curtailment, and thermal power operation cost are divided into different levels, and the weight coefficients of each target function in the same level and the weight coefficients of the target functions in different levels are determined. Based on the same-level weight coefficient and the inter-level weight coefficient, the processed power grid shortage objective function, the processed power grid curtailment objective function, and the processed thermal power operation cost objective function are weighted and aggregated to obtain the single-objective scenario set selection function.

7. The method according to claim 1, characterized in that, Before inputting each of the predicted wind and solar power output scenarios into a preset wind-solar complementary scheduling model to select the power output scenario set and obtaining the target predicted wind and solar power output scenario set, the method further includes: Determine the limits for wind power generation of wind power generation equipment, the limits for photovoltaic power generation of photovoltaic power generation equipment, and the limits for thermal power generation of thermal power generation equipment respectively; Based on the wind power generation limit, the photovoltaic power generation limit, and the thermal power generation limit, a scenario set is selected for each wind and solar power generation prediction scenario set to obtain a wind and solar power generation prediction scenario set that meets the power output requirements. Each of the predicted wind and solar power output scenarios is input into a preset wind-solar complementary scheduling model to select the power output scenario set, thereby obtaining the target predicted wind and solar power output scenario set, including: The wind and solar power output scenario set that meets the power output requirements is input into the preset wind and solar complementary scheduling model to select the power output scenario set and obtain the target wind and solar power output scenario set.

8. A power grid multi-energy complementary dispatching device based on the uncertainty of wind and solar power output, characterized in that, include: The acquisition unit is used to acquire the actual wind and solar power output data and the predicted wind and solar power output data corresponding to the target power grid at multiple historical time points within a historical time period, and to construct a covariance matrix based on the difference between the actual wind and solar power output data and the predicted wind and solar power output data at the same historical time point. The correction unit is used to determine the wind and solar power prediction error based on the covariance matrix, and to obtain multiple initial wind and solar power prediction scene sets. The wind and solar power prediction error is used to correct each initial wind and solar power prediction scene set to obtain multiple wind and solar power prediction scene sets. Each wind and solar power prediction scene set includes wind and solar power prediction data for each time point in a future time period. The selection unit is used to input each of the wind and solar predicted power output scenario sets into a preset wind and solar complementary scheduling model to select the power output scenario set, obtain the target wind and solar predicted power output scenario set, and perform power scheduling according to the target wind and solar predicted power output scenario set.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images