Wind and light pumped storage system capacity optimization configuration method considering seasonal and meteorological characteristics, medium and equipment
Through technical means such as conditional TGAN and adaptive K-M feature clustering method, the capacity configuration of the wind and light pumping storage system is optimized, and the problem of insufficient consideration of seasonal and meteorological characteristics is solved, and more efficient wind and light energy utilization and grid stability are achieved.
Patent Information
- Application Number
- CN202411017639.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-26
- Publication Date
- 2024-11-15
AI Technical Summary
The prior art is difficult to effectively consider seasonal and meteorological characteristics in the capacity configuration of the wind and light pumping storage system, resulting in high costs, increased power shortage rate of the power grid and serious energy disposal, affecting the system's operating efficiency.
Conditional TGAN is used to generate wind and light power generation scenarios in different seasons, combining adaptive K-M feature clustering method and particle swarm reduction optimization algorithm, a double-layer optimization model for wind and light pumping and storage combined system capacity is constructed, and wind power and photoelectric output data are optimized, reducing planning time costs are reduced, and improving New energy consumption capacity.
It significantly improves the optimal configuration effect of the wind and light pumping storage system capacity, reduces the phenomenon of energy disposal, and improves the efficiency of new energy utilization and overall system benefits.
Smart Images

Figure CN118971184A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of renewable energy power generation, and specifically relates to a method, medium and equipment for optimizing the configuration of wind-solar pumped storage system capacity taking into account seasonal and meteorological characteristics. Background Art
[0002] With the continuous expansion and extension of the scale of the power grid, the construction of a wind-solar-pumped-storage combined system can effectively smooth out the fluctuations in wind-solar (wind power and photovoltaic) power generation and improve the utilization efficiency of wind-solar energy. Unreasonable capacity configuration will lead to excessive costs or increased power shortages in the power grid, and the uncertainty of meteorological factors will cause fluctuations in wind power and photovoltaic output, increasing the difficulty of wind power and photovoltaic prediction, thereby affecting the evaluation of the overall benefits of the combined system operation and causing errors in the optimization results of capacity configuration. Therefore, it is necessary to take into account the output differences of wind power and photovoltaic power under different seasonal meteorological characteristics, and study the capacity optimization configuration problem and control strategy of the wind-solar-pumped-storage combined system.
[0003] However, on the one hand, it takes several years or even decades to collect a large amount of system operation-related data before the system is put into operation, which will incur a huge planning time cost. This method does not conform to the actual engineering practice. On the other hand, the phenomenon of energy abandonment is becoming increasingly serious. In the first half of 2017 alone, the national wind abandonment reached 23.5 billion kW·h and the solar abandonment reached 3.7 billion kW·h. Therefore, it is necessary to estimate the operating benefits of wind, solar and pumped storage systems under the condition of less data on power generation scenarios and improve the new energy consumption capacity. Summary of the invention
[0004] In view of the deficiencies in the prior art, the present invention provides a method, medium and equipment for optimizing the capacity configuration of a wind-solar-pumped-storage system taking into account seasonal and meteorological characteristics, taking into account the randomness of wind power and photovoltaic power in different seasons under multiple meteorological combinations, and effectively improving the capacity optimization configuration effect of the wind-solar-pumped-storage combined system.
[0005] To achieve the above object, the present invention adopts the following technical solutions:
[0006] The method for optimizing the capacity of a wind-solar pumped storage system considering seasonal and meteorological characteristics is characterized by comprising the following steps:
[0007] Step 1: Collect meteorological data and historical wind and solar power output data of different seasons in a certain area, and use conditional TGAN to generate wind and solar power generation scenarios in different seasons considering multiple meteorological combinations;
[0008] Step 2: Construct a two-layer optimization model for the capacity of a wind-solar-pumped-storage combined system. The upper model uses historical wind-solar output data as input, with the goal of maximizing annual economic benefits and minimizing wind-solar fluctuations. The lower model uses an adaptive KM feature clustering method to reduce wind-solar power generation scenarios in different seasons, obtains a subset of wind-solar power generation scenarios, and calculates the time-series wind and solar power abandonment power of the corresponding scenarios.
[0009] Step 3: The particle swarm reduction optimization algorithm is used to solve the upper model. The lower model combines the calculated time-series wind and solar power curtailment to optimize the control of the pumped storage power station energy storage capacity of the upper model and feed it back to the upper model. When the iterative solution of the upper and lower models reaches the maximum number of iterations, the optimal configuration plan for the wind-solar-pumped storage system capacity is obtained.
[0010] To optimize the above technical solutions, the specific measures taken also include:
[0011] Further, in step 1, the meteorological data includes wind speed, wind direction, air pressure, temperature, relative humidity and total irradiance at different heights in different seasons; and the collected data is processed as follows:
[0012]
[0013] In the formula, x t is the original data at time t, x t ′ is the processed data at time t, x max is the maximum value of the data information, x min It is the minimum value of data information.
[0014] Furthermore, in step 1, the conditional TGAN is used to generate wind and solar power generation scenarios in different seasons taking into account multiple meteorological combinations. The specific process is as follows:
[0015] The historical wind and solar power output data is used as a training sample set;
[0016] Define wind and solar power generation scenarios in different seasons as static features m , m is different power generation scenarios, and meteorological data is used as dynamic feature c i , i represents the number of training sample sets, set the number of training times N, input the data in the training sample set and its dynamic feature seasonal sequence into the conditional TGAN to complete the network training, and input the seasonal scene feature condition s m and c i Input the generator to generate results for wind and solar power generation scenarios in spring, summer, autumn and winter.
[0017] Furthermore, the network training process of the conditional TGAN includes:
[0018] 1) Embed the time information into the training sample set to generate the training sample time series set (sm , x 1:T ) and the joint distribution P, T is the length of the time series, x 1:T is a dynamic feature vector of time series length 1 to T. The training sample time series set is used as the input of the embedding function. The embedding function calculation formula is:
[0019]
[0020] In the formula, s m is the real high-dimensional static feature vector, h S is the low-dimensional static feature vector after mapping, h t represents the low-dimensional dynamic feature vector at time t, x t represents the high-dimensional dynamic feature vector at time t, e S , e are the embedding functions corresponding to the static eigenvector and the dynamic eigenvector respectively;
[0021] 2) For h S 、h t Reproduce and reconstruct the original high-dimensional dynamic and static characteristics. The specific calculation formula is:
[0022]
[0023] In the formula, r S , r are the recurrence functions of static eigenvector and dynamic eigenvector respectively, To reproduce the decoded high-dimensional static feature vector, To recover the high-dimensional dynamic feature vector at time t;
[0024] 3) m With c i As a supervision factor, the random input with known distribution is used as the input of the generator. The specific calculation formula of the generator function is:
[0025]
[0026] In the formula, g S , g are the generator functions of static feature vector and dynamic feature vector respectively, z S 、z t Random sampling of static and dynamic random vector spaces with known distributions, is the generated static feature vector, To generate the dynamic feature vector at the,th moment;
[0027] 4) The output of the TGAN generator and the output of the embedded function are passed through a joint encoder to obtain a synthesized output, and the synthesized output and the seasonal scene feature condition s to be generated are combined. m ′ and c i′ is used as the input of the discriminator to obtain the classification result, and the output result of the discriminator is compared with the real scenery time series. The discriminator function formula is:
[0028]
[0029] In the formula, are the discriminator functions of static feature vectors and dynamic feature vectors respectively, Indicates the static feature generation classification result, Generate classification results for dynamic features at time t, is the output result of the joint encoding at the,th moment;
[0030] 5) Optimize the global loss function V. The specific formula is:
[0031]
[0032] In the formula, the global loss function V includes V(e, r) and λ, η are weight parameters, L er The joint loss function is the output result of the embedding function and the recurrence function, is distributed (s m , x 1:T )~P expectation operator, L gd is the joint loss function of the generator and the discriminator, L eg is the supervision loss function,
[0033] Furthermore, in step 2, the upper model uses the historical output data of wind and solar as input, with the goal of maximizing the annual economic benefit and minimizing the wind and solar fluctuation rate, as follows:
[0034] The objective function F that maximizes the annual economic benefit 1 The specific formula is:
[0035]
[0036] The objective function F with the minimum wind and solar fluctuation rate 2 The specific formula is:
[0037]
[0038] Where a is the number of days, A is the average annual cost parameter of the pumped storage power station, and C w,t is the unit price of electricity sold by the wind farm to the grid during the period, C so,t is the unit price of electricity sold by the photovoltaic power plant to the power grid during the period, C f,t is the unit price of electricity sold by the pumped storage power station to the grid during period t, C p,tis the pumping fee for period t, is the average annual construction and operation cost of the reservoir unit, is the average annual installation and operation cost of the pump unit, P wt is the power transmitted to the grid by the wind farm during period t, P so,t is the power delivered to the grid by the photovoltaic power station during period t, P f,t is the power generated by the turbine unit during period t, P p,t is the pumping power of the water pump during the period, R max is the maximum capacity of the reservoir, Q max is the maximum power generation capacity of the turbine unit, They are the installed capacities of wind power and photovoltaic power respectively.
[0039] Furthermore, the constraints of the upper model include:
[0040] The grid is subject to power balance constraints:
[0041] P min ≤P w,t +P so,t +P ft ≤P max ;
[0042] Where P min , P max is the minimum and maximum power that the grid can receive;
[0043] Reservoir energy balance constraints:
[0044]
[0045] Where η p is the pumping efficiency of the pump, η f is the power generation efficiency of the turbine, R dump is the capacity not stored in the reservoir;
[0046] Reservoir capacity constraints:
[0047] 0≤R t ≤R max ;
[0048] In the formula, R t is the capacity of the reservoir at time t, R max to its maximum capacity.
[0049] Furthermore, in step 2, the lower model uses an adaptive KM feature clustering method to reduce the wind and solar power generation scenarios in different seasons, specifically including the following steps:
[0050] 1) Take the wind and solar power generation scenes generated in the same season in step 1 as the data set Y = [Y 1, Y 2 , Y 3 , ..., Y l ] T Input, define the number of wind and solar power generation scene clusters in the same season as k, and randomly select the cluster center U at the beginning 1 , U 2 …U k ;
[0051] 2) Calculate the effective distance between the data in each data set and each cluster center, and divide the data into the cluster closest to it. The calculation formula for the effective distance is:
[0052]
[0053] In the formula, H αβ is the αth cluster center U α With sparse data Y β The effective distance between αβ is the αth cluster center U α In the case of sparse data Y β The previous coefficients are ω α is the cluster center U α The weight coefficient, χ is the data set excluding the cluster center in the data set, and δ is the sparse parameter;
[0054] 3) Recalculate the cluster center U of each cluster α ′, the calculation formula is:
[0055]
[0056] In the formula, ε is the εth data of the original cluster, and φ is the number of data contained in the original cluster;
[0057] 4) Repeat steps 2) and 3) until the cluster center no longer changes;
[0058] 5) Return the cluster center U 1 , U 2 …U k , which represents a subset of wind and solar power generation scenarios.
[0059] Furthermore, in step 3, the lower model optimizes and controls the energy storage capacity of the pumped storage power station of the upper model in combination with the calculated time-series abandoned wind and solar power and feeds back to the upper model, specifically:
[0060] The following judgments are made based on the sequential wind and solar power abandonment power of the subset of wind and solar power generation scenarios:
[0061] When the scheduled wind and solar curtailment power is higher than the working power of the wind-solar pumped storage power generation system, the wind-solar pumped storage power generation system still operates at the working power;
[0062] When the sequential wind and solar power abandonment power is not higher than the working power of the wind-solar and pumped-storage power generation system, further judgment is made: when the sequential wind and solar power abandonment power is higher than the rated power, the wind-solar and pumped-storage power generation system operates with the sequential wind and solar power abandonment power; when the sequential wind and solar power abandonment power is not higher than the rated power, the energy storage capacity of the pumped-storage power station is the sum of the pumping volume and the power generation volume.
[0063] Accordingly, the present invention proposes a computer-readable storage medium storing a computer program, characterized in that the computer program enables a computer to execute the above-mentioned method for optimizing the capacity configuration of a wind-solar-pumped storage system taking into account seasonal and meteorological characteristics.
[0064] Accordingly, the present invention proposes an electronic device, characterized in that it includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it implements the method for optimizing the capacity configuration of a wind-solar-pumped storage system taking into account seasonal and meteorological characteristics as described above.
[0065] The beneficial effects of the present invention are as follows: the present invention uses conditional TGAN to generate a large number of wind and solar power generation scenes in different seasons, deeply explores the dynamic and static characteristics contained in the wind power and photovoltaic output data in different seasons under the multivariate meteorological combination, takes the dynamic and static characteristics of the wind power and photovoltaic output data as conditional supervision items, combines unsupervised adversarial training with supervised training, and can better capture the step-by-step dynamic probability distribution of the new energy time series under the multivariate meteorological combination, and significantly improves the accuracy of time series generation; the present invention establishes a two-layer optimization model for the capacity of a wind-solar-pumped-storage combined system, the upper layer considers the overall economic benefits and new energy volatility of the combined system, the lower layer takes the abandoned wind and abandoned photovoltaic power as the control target, and increases the upper limit power of the combined system by increasing the capacity of the pumped-storage power station, and further optimizes and controls the capacity of the wind-solar-pumped-storage combined system; the adaptive KM feature clustering method is used to reduce the scene, and a new effective distance calculation method is defined to improve the accuracy of the clustering algorithm; the present invention uses a particle swarm reduction optimization algorithm to solve the upper model, and the algorithm introduces a step-by-step file reduction method in the late iteration of the multi-objective particle swarm optimization algorithm, and uses the radius reduction technology to optimize it, thereby reducing memory usage and accelerating convergence speed. BRIEF DESCRIPTION OF THE DRAWINGS
[0066] Figure 1 The present invention is a flow chart of the method for optimizing the capacity configuration of a wind-solar-pumped storage system taking into account seasonal and meteorological characteristics.
[0067] Figure 2 This is the training flow chart of the conditional TGAN of the present invention.
[0068] Figure 3 The figure is a flow chart of the particle swarm reduction optimization algorithm of the present invention. DETAILED DESCRIPTION
[0069] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.
[0070] In one embodiment, the present invention proposes a method for optimizing the capacity configuration of a wind-solar pumped storage system taking into account seasonal and meteorological characteristics, such as Figure 1 As shown, the following steps are included.
[0071] Step 1: Collect meteorological data and historical wind and solar power output data of different seasons in a certain area, normalize the original data, use conditional TGAN to generate a large number of wind and solar power generation scenarios in different seasons, deeply explore the dynamic and static characteristics contained in the wind power and photovoltaic output data of different seasons based on multi-meteorological combinations, use the dynamic and static characteristics of wind power and photovoltaic output data as conditional supervision items, combine unsupervised adversarial training with supervised training, construct a global loss function, optimize the embedding function, reproduction function, generator function, and discriminator function, which can better capture the step-by-step dynamic probability distribution of the time series of new energy output under multi-meteorological combinations, and significantly improve the accuracy of time series generation.
[0072] In this embodiment, the collected meteorological data include: wind speed, wind direction, air pressure, temperature, relative humidity, and total irradiance at different heights in different seasons, and the data is processed. The processing process is as follows:
[0073]
[0074] In the formula, x t is the original data at time t, x t ′ is the processed data at time, x max is the maximum value of the data information, x min It is the minimum value of data information.
[0075] like Figure 2 As shown in Figure 2, the specific process of using conditional TGAN to generate wind and solar power generation scenarios in different seasons considering multiple meteorological combinations is as follows:
[0076] Step 1.1: Use the processed original data of wind power and photovoltaic output as the training sample set.
[0077] Step 1.2: Define wind and solar power generation scenarios in different seasons as static features m , m is different power generation scenarios, and the normalized meteorological data is used as the dynamic feature c i, i represents the number of training sample sets, set the number of training times N, input the data in the training sample set and its dynamic feature seasonal sequence into the conditional TGAN to complete the network training, and generate the scene feature condition s m and c i Input the generator to generate results for wind power and photovoltaic scenarios in spring, summer, autumn and winter.
[0078] In this embodiment, the network training is completed by conditional TGAN in step 1.2, and the specific process is as follows:
[0079] a. Embed the time information into the training sample set to generate the training sample time series set (s m , x 1:T ) and the joint distribution P, T is the length of the time series, x 1:T is a dynamic feature vector of time series length 1 to T. The training sample time series set data is used as the input of the embedding function e. The calculation formula of the embedding function e is:
[0080]
[0081] In the formula, s m is the real high-dimensional static feature vector, h S is the low-dimensional static feature vector after mapping, h t represents the low-dimensional dynamic feature vector at time t, x t represents the high-dimensional dynamic feature vector at time t, e S , e are the embedding functions corresponding to the static feature vector and dynamic feature vector respectively.
[0082] b. for h s 、h t Reproduce and reconstruct the original high-dimensional dynamic and static characteristics. The specific calculation formula is:
[0083]
[0084] In the formula, r S , r are the recurrence functions of static eigenvector and dynamic eigenvector respectively, To reproduce the decoded high-dimensional static feature vector, To restore the high-dimensional dynamic feature vector at time t.
[0085] c. m With c i As a supervision factor, the random input with known distribution is used as the input of the generator. The specific calculation formula of the generator function g is:
[0086]
[0087] In the formula, g S, g are the generator functions of static feature vector and dynamic feature vector respectively, z S 、z t Random sampling of static and dynamic random vector spaces with known distributions, is the generated static feature vector, To generate the dynamic feature vector at the moment,.
[0088] d. The output of the TGAN generator and the output of the embedded function are combined through a joint encoder to obtain a synthesized output. The synthesized output and the seasonal scene feature condition s to be generated are combined. m ′ and c i ′ is used as the input of the discriminator to obtain the classification result, and the output result of the discriminator is compared with the real wind and solar power time series. The discriminator function The formula is:
[0089]
[0090] In the formula, Indicates the static feature generation classification result, The classification results of dynamic features at time t, static features and dynamic features are the wind and solar power generation scenarios in different seasons under multi-meteorological combinations. Output result for joint encoding.
[0091] e. Optimize the global loss function V. The specific formula is:
[0092]
[0093] In the formula, the global loss function V includes V(e, r) and λ, η are weight parameters, L er The joint loss function is the output result of the embedding function and the recurrence function, is distributed (s m , x 1:T )~P expectation operator, L gd is the joint loss function of the generator and the discriminator, L eg is the supervision loss function,
[0094] Step 2: Construct a two-layer optimization model for the capacity of a wind-solar-pumped-storage combined system. The upper layer uses the wind-solar power plant data under the characteristics of each season as input, and takes the maximum annual economic benefit and the minimum wind power and photovoltaic fluctuation rate as the goal to construct a wind-solar-pumped-storage combined system capacity optimization model. The lower model uses the adaptive KM feature clustering method to reduce a large number of wind-solar power generation scenes in different seasons, defines a new effective distance calculation method, improves the accuracy of the clustering algorithm, obtains a subset of wind-solar seasonal meteorological characteristic scenes, and calculates the time series of wind and photovoltaic power abandonment for the corresponding scenes. The two-layer model of the present invention effectively takes into account the annual benefits, wind-solar fluctuations, and wind and photovoltaic power abandonment, and the upper and lower layers are iteratively solved to make the results more accurate.
[0095] In this embodiment, in the two-layer optimization model of the wind-solar-pumped-storage combined system capacity, the upper model specifically uses the wind-solar power plant data under the characteristics of each season as input, with the goal of maximizing the annual economic benefits and minimizing the volatility of wind power and photovoltaic power.
[0096] The objective function F that maximizes the annual economic benefit 1 The specific formula is:
[0097]
[0098] The objective function F of minimizing the volatility of wind power and photovoltaic power 2 The specific formula is:
[0099]
[0100] Where a is the number of days, A is the average annual cost parameter of the pumped storage power station, and C w,t is the unit price of electricity sold by the wind farm to the grid during period t, C so,t is the unit price of electricity sold by the photovoltaic power plant to the power grid during period t, C f,t is the unit price of electricity sold by the pumped storage power station to the grid during period t, C p,t is the pumping fee for period t, is the average annual construction and operation cost of the reservoir unit, is the average annual installation and operation cost of the pump unit, P w,t is the power transmitted to the grid by the wind farm during period t, P so,t is the power delivered to the grid by the photovoltaic power station during the period, P f,t is the power generated by the turbine unit during period t, P p,t is the pumping power of the water pump during period t, R max is the maximum capacity of the reservoir, Q max is the maximum power generation capacity of the turbine unit, They are the installed capacities of wind power and photovoltaic power respectively.
[0101] The upper model constraints are as follows:
[0102] The grid is subject to power balance constraints:
[0103] P min ≤P w,t +P so,t +P f,t ≤P max ;
[0104] Where P min , P max is the minimum and maximum power that the grid can receive.
[0105] Reservoir energy balance constraints:
[0106]
[0107] Where η p is the pumping efficiency of the pump, η f is the power generation efficiency of the turbine, R dump It is the capacity not stored in the reservoir.
[0108] Reservoir capacity constraints:
[0109] 0≤R t ≤R max ;
[0110] In the formula, R t is the capacity of the reservoir at the moment, R max to its maximum capacity.
[0111] In this embodiment, the adaptive KM feature clustering method is specifically as follows:
[0112] Step 2.1: Take the large number of wind and solar power generation scenes in the same season generated in step 1 as the data set Y = [Y 1 , Y 2 , Y 3 , ..., T l ] T Input, define the number of scenery scene clusters in the same season as k, and randomly select the cluster center U at the beginning 1 , U 2 …U k .
[0113] Step 2.2: Calculate the effective distance between the data in each data set and each cluster center, and divide the data into the cluster closest to it. The specific calculation formula for the effective distance is:
[0114]
[0115] In the formula, H αβ is the αth cluster center U α With sparse data Y βThe effective distance between αβ is the αth cluster center U α In the sparse data Y β The previous coefficients are ω α is the cluster center U α The weight coefficient of , χ is the data set excluding the cluster center in the data set, and δ is the sparse parameter.
[0116] Step 2.3: Recalculate the new cluster center U of each cluster α ′, the calculation formula is:
[0117]
[0118] In the formula, ε is the εth data of the original cluster, and φ is the number of data contained in the original cluster.
[0119] Step 2.4: Repeat steps 2.2 and 2.3 until the cluster center point no longer changes.
[0120] Step 2.5: Return the cluster center U 1 , U 2 …U k , that is, a subset of seasonal meteorological characteristic scenes.
[0121] Step 3: The upper model is solved by particle swarm reduction optimization algorithm. The algorithm introduces stepwise file reduction method in the later iteration of multi-objective particle swarm optimization algorithm, and optimizes it by reducing radius technology, so as to reduce memory usage and accelerate convergence speed. The lower model further optimizes and controls the energy storage capacity of the pumped storage power station of the upper model and feeds it back to the upper layer. The lower layer takes the abandoned wind and solar power as the control target, and increases the upper limit power of the joint system by increasing the capacity of the pumped storage power station. According to the time series abandoned wind and solar power of the subset of seasonal meteorological characteristic scenes of wind and solar power, when the abandoned wind and solar power is higher than the working power of the wind-solar pumped storage power generation system, the system still operates at the working power; when the abandoned wind and solar power is not higher than the working power but higher than the rated power, the wind-solar pumped storage power generation system operates at the abandoned wind and solar power; when the abandoned wind and solar power is not higher than the rated power, the energy storage capacity of the pumped storage power station is the sum of the pumping amount and the power generation amount. When the upper and lower iterative solutions reach the maximum number of iterations, the optimal configuration scheme of the wind-solar pumped storage system capacity is obtained.
[0122] like Figure 3 As shown, in this embodiment, the specific steps of the particle swarm reduction optimization algorithm are:
[0123] Step 3.1: Initialize the position and velocity of the group particles and reduce the radius D share , calculate the fitness value F of each particle 1 、F 2 .
[0124] Step 3.2: According to the Pareto dominance principle, obtain the archive set to store the current non-inferior solution.
[0125] Step 3.3: Determine the historical optimal pbest and global optimal gbest of each particle in the archive set.
[0126] Step 3.4: Update the particle's speed and position, calculate the corresponding fitness value and update it. The update formula for the particle's speed and position is:
[0127]
[0128] Where σ is the inertia weight coefficient, and is a uniform random number in the range [0,1], c 1 and c 2 They are self-learning factor and social learning factor. is the individual optimal value of j particles, is the global optimal value.
[0129] Step 3.5: Process the file set and make a judgment. If the number n of solutions in the file set is greater than the maximum number of solutions in the file set, go to step 3.6, otherwise go to step 3.7.
[0130] Step 3.6: Perform file optimization and reduction until the file no longer overflows, then go to step 3.7. The specific steps of file optimization and reduction are:
[0131] 1) Calculate the objective function value of the archive solution F = [F 1 (Archive), F 2 (Archive)] T , and sort F in ascending order of a certain dimension value in the target, let n = size(F, 2), n is the number of columns of F, which is also the number of corresponding solutions;
[0132] 2) Select the bth point as the base point;
[0133] 3) Calculate the distance between base point b and point b+1 and point b+2 respectively, and record them as d 1 d 2 (d 1 <d 2 ), and compare their size relationships:
[0134] If d 1 >D share , then point b+1 becomes the new base point, return to 3);
[0135] If d 2 <Dshare , then delete point b+1 and point b+2 in Archive and F, update Archive and F as well as the number n, and return to 3);
[0136] If d 1 ≤D share ≤d 2 , then delete point b+1 in Archive and F, use point b+2 as the new base point, update Archive and F as well as the number n, and return to 3).
[0137] Step 3.7: Check whether the maximum number of iterations has been reached. If so, terminate the process and output the Pareto optimal solution set in the archive, i.e., the pumped storage capacity configuration plan. Otherwise, go to step 3.3.
[0138] In another embodiment, the present invention proposes a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the method for optimizing the capacity configuration of a wind-solar-pumped storage system taking into account seasonal and meteorological characteristics as described in Example 1.
[0139] In another embodiment, the present invention proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the method for optimizing the capacity configuration of a wind-solar-pumped storage system taking into account seasonal and meteorological characteristics as described in Example 1.
[0140] In the embodiments disclosed in the present application, the computer storage medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. The computer storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the above. More specific examples of computer storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.
[0141] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.
[0142] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.
Claims
1. Capacity optimization configuration method of wind-solar pumped storage system considering seasonal and meteorological characteristics, It is characterized in that The steps include: Step 1: Collect meteorological data and historical wind and solar power output data of different seasons in a certain area, and use conditional TGAN to generate wind and solar power generation scenarios in different seasons considering multiple meteorological combinations; Step 2: Construct a two-layer optimization model for the capacity of a wind-solar-pumped-storage combined system. The upper model uses historical wind-solar output data as input, with the goal of maximizing annual economic benefits and minimizing wind-solar fluctuations. The lower model uses an adaptive KM feature clustering method to reduce wind-solar power generation scenarios in different seasons, obtains a subset of wind-solar power generation scenarios, and calculates the time-series wind and solar power abandonment power of the corresponding scenarios. Step 3: The particle swarm reduction optimization algorithm is used to solve the upper model. The lower model combines the calculated time-series wind and solar power curtailment to optimize the control of the pumped storage power station energy storage capacity of the upper model and feed it back to the upper model. When the iterative solution of the upper and lower models reaches the maximum number of iterations, the optimal configuration plan for the wind-solar-pumped storage system capacity is obtained.
2. The method for optimizing the capacity of a wind-solar-pumped storage system considering seasonal and meteorological characteristics as claimed in claim 1, Features: In step 1, the meteorological data includes wind speed, wind direction, air pressure, temperature, relative humidity and total irradiance at different heights in different seasons; and the collected data is processed as follows: In the formula, x t is the original data at time t, x t ′ is the processed data at time t, x max is the maximum value of the data information, x min It is the minimum value of data information.
3. The method for optimizing the capacity of a wind-solar-pumped storage system considering seasonal and meteorological characteristics as claimed in claim 1, Features: In step 1, the conditional TGAN is used to generate wind and solar power generation scenarios in different seasons considering multiple meteorological combinations. The specific process is as follows: The historical wind and solar power output data is used as a training sample set; Define wind and solar power generation scenarios in different seasons as static features m , m is different power generation scenarios, and meteorological data is used as dynamic feature c i , i represents the number of training sample sets, set the number of training times N, input the data in the training sample set and its dynamic feature seasonal sequence into the conditional TGAN to complete the network training, and input the seasonal scene feature condition s m and c i Input the generator to generate results for wind and solar power generation scenarios in spring, summer, autumn and winter.
4. The method for optimizing the capacity of a wind-solar-pumped storage system considering seasonal and meteorological characteristics as claimed in claim 3, Features: The network training process of the conditional TGAN includes: 1) Embed the time information into the training sample set to generate the training sample time series set (s m , x 1:T ) and the joint distribution P, T is the length of the time series, x 1:T is a dynamic feature vector of time series length 1 to T. The training sample time series set is used as the input of the embedding function. The embedding function calculation formula is: In the formula, s m is the real high-dimensional static feature vector, h S is the low-dimensional static feature vector after mapping, h t represents the low-dimensional dynamic feature vector at time t, x t represents the high-dimensional dynamic feature vector at time t, e S , e are the embedding functions corresponding to the static eigenvector and the dynamic eigenvector respectively; 2) For h S 、h t Reproduce and reconstruct the original high-dimensional dynamic and static characteristics. The specific calculation formula is: In the formula, rS and r are the recurrence functions of the static eigenvector and the dynamic eigenvector respectively. To reproduce the decoded high-dimensional static feature vector, To recover the high-dimensional dynamic feature vector at time t; 3) m With c i As a supervision factor, the random input with known distribution is used as the input of the generator. The specific calculation formula of the generator function is: In the formula, g S , g are the generator functions of static feature vector and dynamic feature vector respectively, z S 、z t Random sampling of static and dynamic random vector spaces with known distributions, is the generated static feature vector, To generate the dynamic feature vector at the tth moment; 4) The output of the TGAN generator and the output of the embedded function are passed through a joint encoder to obtain a synthesized output, and the synthesized output and the seasonal scene feature condition s to be generated are combined. m ′ and c i ′ is used as the input of the discriminator to obtain the classification result, and the output result of the discriminator is compared with the real scenery time series. The discriminator function formula is: In the formula, are the discriminator functions of static feature vectors and dynamic feature vectors respectively, Indicates the static feature generation classification result, Generate classification results for dynamic features at time t, is the output result of joint coding at time t; 5) Optimize the global loss function V. The specific formula is: In the formula, λ and η are weight parameters, L er The joint loss function is the output result of the embedding function and the recurrence function, is distributed (s m , x 1:T )~P expectation operator, L gd is the joint loss function of the generator and the discriminator, L eg is the supervision loss function, 5. The method for optimizing the capacity of a wind-solar-pumped storage system considering seasonal and meteorological characteristics as claimed in claim 1, Features: In step 2, the upper model uses the historical output data of wind and solar power as input, with the goal of maximizing annual economic benefits and minimizing wind and solar power fluctuations, as follows: The objective function F that maximizes the annual economic benefit 1 The specific formula is: The objective function F with the minimum wind and solar fluctuation rate 2 The specific formula is: In the formula, a is the number of days, A is the average annual cost parameter of the pumped storage power station, and C w,t is the unit price of electricity sold by the wind farm to the grid during period t, C so,t is the unit price of electricity sold by the photovoltaic power plant to the power grid during period t, C f,t is the unit price of electricity sold by the pumped storage power station to the grid during period t, C p,t is the pumping fee for period t, is the average annual construction and operation cost of the reservoir unit, is the average annual installation and operation cost of the pump unit, P w,t is the power transmitted to the grid by the wind farm during period t, P so,t is the power delivered to the grid by the photovoltaic power station during period t, P f,t is the power generation of the turbine unit during period t, P p,t is the pumping power of the water pump during period t, R max is the maximum capacity of the reservoir, Q max is the maximum power generation capacity of the turbine unit, They are the installed capacities of wind power and photovoltaic power respectively.
6. The method for optimizing the capacity configuration of a wind-solar-pumped storage system taking into account seasonal and meteorological characteristics as claimed in claim 5, Features: The constraints of the upper model include: The grid is subject to power balance constraints: P min ≤P w,t +P so,t +P f,t ≤P max ; Where P min , P max is the minimum and maximum power that the grid can receive; Reservoir energy balance constraints: Where N p is the pumping efficiency of the pump, η f is the power generation efficiency of the turbine, R dump is the capacity not stored in the reservoir; Reservoir capacity constraints: 0≤R t ≤R max ; In the formula, R t is the capacity of the reservoir at time t, R max to its maximum capacity.
7. The method for optimizing the capacity of a wind-solar-pumped storage system considering seasonal and meteorological characteristics as claimed in claim 1, Features: In step 2, the lower model uses an adaptive KM feature clustering method to reduce the wind and solar power generation scenarios in different seasons, which specifically includes the following steps: 1) Take the l wind and solar power generation scenes in the same season generated in step 1 as the data set Y = [Y 1 , Y 2 , Y 3 , ..., Y l ] T Input, define the number of wind and solar power generation scene clusters in the same season as k, and randomly select the cluster center U at the beginning 1 , U 2 …U k ; 2) Calculate the effective distance between the data in each data set and each cluster center, and divide the data into the cluster closest to it. The calculation formula for the effective distance is: In the formula, H αβ is the αth cluster center U α With sparse data Y β The effective distance between αβ is the αth cluster center U α In the sparse data Y β The previous coefficients are ω α is the cluster center U α The weight coefficient, X is the data set excluding the cluster center in the data set, and δ is the sparse parameter; 3) Recalculate the cluster center U of each cluster α ′, the calculation formula is: In the formula, ε is the εth data of the original cluster, and φ is the number of data contained in the original cluster; 4) Repeat steps 2) and 3) until the cluster center no longer changes; 5) Return the cluster center U 1 , U 2 …U k , which represents a subset of wind and solar power generation scenarios.
8. The method for optimizing the capacity of a wind-solar-pumped storage system considering seasonal and meteorological characteristics as claimed in claim 1, Features: In step 3, the lower model optimizes and controls the energy storage capacity of the pumped storage power station of the upper model in combination with the calculated time-series abandoned wind and solar power and feeds back to the upper model, specifically: The following judgments are made based on the time-series abandoned wind and solar power of the subset of wind and solar power generation scenarios: When the scheduled wind and solar curtailment power is higher than the working power of the wind-solar pumped storage power generation system, the wind-solar pumped storage power generation system still operates at the working power; When the sequential wind and solar power abandonment power is not higher than the working power of the wind-solar and pumped-storage power generation system, further judgment is made: when the sequential wind and solar power abandonment power is higher than the rated power, the wind-solar and pumped-storage power generation system operates with the sequential wind and solar power abandonment power; when the sequential wind and solar power abandonment power is not higher than the rated power, the energy storage capacity of the pumped-storage power station is the sum of the pumping volume and the power generation volume.
9. A computer-readable storage medium storing a computer program, It is characterized in that The computer program enables the computer to execute the method for optimizing the capacity configuration of a wind-solar-pumped storage system taking into account seasonal and meteorological characteristics as described in any one of claims 1 to 8.
10. An electronic device, It is characterized in that include: A memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for optimizing the configuration of the capacity of a wind-solar-pumped storage system taking into account seasonal and meteorological characteristics as described in any one of claims 1 to 8 is implemented.
