A pumped storage planning method considering the regulation capability of power grid under extreme weather

Abstract

The application provides a pumped storage planning method considering the regulation capacity of a power grid under extreme weather, comprising: establishing characteristic models of various types of power sources and loads in the system; obtaining typical scenarios including extreme working conditions; determining pumped storage reservoir volume based on the characteristic models and the typical scenarios. The method provided by the application can effectively analyze the regulation capacity gap of a power system under multiple uncertainties of sources and loads and the power over-limit condition under the worst condition, and provides reliable theoretical and data support for the construction of subsequent pumped storage power stations and other energy storage devices.

Description

Technical Field

[0001] This application relates to the field of pumped storage technology, and in particular to a pumped storage planning method that takes into account the grid regulation capacity under extreme weather conditions. Background Technology

[0002] Currently, the power system is undergoing a major transformation due to the large-scale integration of new energy sources. By the end of 2024, the cumulative installed capacity of new energy reached 1.41 billion kilowatts, accounting for 42% of the national total installed capacity, setting a new historical record. At the same time, this high proportion of grid connection has brought new challenges to the system's regulation capacity and resilience. Pumped storage, as one of the most mature and scalable green regulation methods currently available, can effectively suppress fluctuations in new energy output, store excess energy, and provide effective guarantees for the safe and stable operation of the power grid.

[0003] In recent years, against the backdrop of global warming, extreme weather events, such as those occurring once every ten years or even once every century, have shown an increasing trend of intensity in certain regions, exacerbating the vulnerability of power systems. Therefore, researching the regulation capacity gap of power systems under multiple uncertainties during extreme weather events is of great significance.

[0004] Existing research has proposed solutions for the scheduling of hydropower, wind power, solar power, and energy storage systems under severe weather conditions. By simulating the output changes of different new energy sources under extreme weather conditions, the optimal output model is obtained, which improves the flexibility and regulation capability of the power grid to a certain extent. Another part of the research proposes collaborative planning strategies for power systems with multiple uncertainties. By using optimization algorithms, the overall performance of the energy system is improved from the perspectives of energy storage configuration and operation mode, and the operating cost of the power system is reduced, providing strong technical support for the "dual carbon" goal.

[0005] Current research focuses on the overall operational optimization of power systems after wind and solar power are integrated into them, aiming to reduce overall operating costs and expand into areas such as energy storage. However, there is a lack of effective analysis regarding the system's regulation capacity under extreme weather conditions, which hinders the rational planning and integration of energy storage in subsequent power systems. For example, Du Yongfeng et al. used interval optimization to address the uncertainty of wind and solar power output, analyzed the impact of different load demands and energy storage devices, and incorporated seasonal typical scenario analysis to make the model reflect the more realistic impact of source-load uncertainty on the power grid. However, this method does not consider the impact of extreme weather on renewable energy output, limiting its practical application. Existing technologies address the joint optimization of new power systems and variable-speed pumped storage power stations, considering factors such as reservoir capacity constraints, thermal power unit output, and ramping constraints. A two-stage optimization method enables pumped storage power stations to better absorb wind and solar renewable resources, but it does not address the rational design of pumped storage reservoir capacity, potentially leading to over-construction and resource waste in the modified hybrid pumped storage power stations. Summary of the Invention

[0006] This application provides a pumped storage planning method that considers the grid regulation capacity under extreme weather conditions. To solve the above-mentioned technical problems, this application adopts the following technical method: This application provides a pumped storage planning method that considers the grid regulation capacity under extreme weather conditions, including: Establish characteristic models for various power sources and loads in the system; Acquire typical scenarios including extreme operating conditions; Based on the aforementioned characteristic model and the aforementioned typical scenario, the volume of the pumped storage reservoir is determined.

[0007] Optionally, the characteristic models of various resources and loads in the system include wind power output model, photovoltaic power output model, load demand model and electric vehicle load model.

[0008] Optionally, the process of constructing the wind power output model includes the following steps: Obtain the Weibull marginal wind speed model and the first-order autoregressive model; The first-order autoregressive model is transformed using the standard normal cumulative distribution function to generate a first-order autoregressive uniform distribution model. The first-order autoregressive uniform distribution model and the Weibull marginal wind speed model are inversely transformed to generate a time-related wind speed sequence. Based on the wind speed sequence, a wind power output model is determined.

[0009] Optionally, the process of constructing the photovoltaic power output model includes the following steps: A probability distribution model of light intensity is constructed using the beta distribution; The photovoltaic power output model is determined by performing a first-order approximation on the light intensity probability distribution model.

[0010] Optionally, the process of constructing the load demand model includes the following steps: A typical daily bimodal load curve model was constructed by using a normal distribution and superimposing a first-order autoregressive model. Peak load calibration is used to calibrate the typical daily double-peak load curve model and determine the load demand model.

[0011] Optionally, the process of constructing the electric vehicle load model includes the following steps: Using a normal distribution, we construct a first-time model of electric vehicles arriving at a charging station and a second-time model of electric vehicles leaving a charging station. Based on the first time model and the second time model, the electric vehicle load model is determined.

[0012] Optionally, the process for determining typical scenarios under extreme operating conditions includes the following steps: Monte Carlo sampling is used to generate a sufficient number of scene samples through large-scale pseudo-random sampling. Principal component analysis is used to reduce the dimensionality of the scene samples while retaining the principal component features; After further standardizing the principal component features, different weights are assigned to them to construct a clustering feature matrix; K-Medoids clustering is used to extract typical scenarios from the clustering feature matrix, generating typical scenarios that include extreme working conditions.

[0013] Optionally, determining the pumped storage reservoir volume based on the characteristic model and the typical scenario includes: Using the aforementioned typical scenario, the output of the special effects model is adjusted to determine the net load for different typical days; Based on the net load of the different typical days, the annual expected energy deficit is determined; The volume of the pumped reservoir is determined based on the expected annual energy deficit.

[0014] This application has the following beneficial effects: The method proposed in this application can effectively analyze the power system regulation capacity gap under multiple uncertainties of source and load, as well as the power limit overrun situation under the worst case, and provides reliable theoretical and data support for the construction of subsequent pumped storage power stations and other energy storage devices. Attached Figure Description

[0015] Figure 1 A flowchart illustrating a pumped storage planning method that considers the grid regulation capacity under extreme weather conditions, provided as an embodiment of this application; Figure 2 This is a schematic diagram illustrating the selection of the optimal number of K-Medoid clusters provided in an embodiment of this application. Figure 3 This application provides a schematic diagram illustrating typical source-load uncertainty of the system under normal weather conditions in an embodiment of the present application. Figure 4 A schematic diagram illustrating the load uncertainty of a system under extreme weather conditions, provided in an embodiment of this application; Figure 4 (a) is a schematic diagram of load uncertainty under heavy rain weather; Figure 4 (b) is a schematic diagram of load uncertainty under drought conditions; Figure 5 Schematic diagrams of the maximum energy gap of the system in different typical scenarios without energy storage provided in the embodiments of this application; Figure 6 A schematic diagram illustrating the maximum reactive power gap of the system in different typical scenarios when no energy storage is installed, as provided in the embodiments of this application; Figure 7 A schematic diagram illustrating the maximum energy gap of the system after adding partial energy storage, provided in an embodiment of this application; Figure 8 A schematic diagram of the maximum reactive power gap of the system after adding partial energy storage, provided for an embodiment of this application. Detailed Implementation

[0016] To facilitate understanding by those skilled in the art, the present application will be further described below in conjunction with embodiments and accompanying drawings. The content mentioned in the embodiments is not intended to limit the present application.

[0017] To solve the above technical problems, such as Figure 1 As shown, this application proposes a pumped storage planning method that considers the grid regulation capacity under extreme weather conditions, including: Step S101: Establish characteristic models for various power sources and loads in the system; The characteristic models for various power sources and loads in this step include wind power output model, photovoltaic power output model, load demand model, and electric vehicle load model.

[0018] Wind power and photovoltaic (PV) power, as the most common new energy power sources in new power systems, have been widely promoted and applied throughout China due to their relatively stable output and higher unit installed capacity resource efficiency. However, the output characteristics of wind and PV power are heavily influenced by local meteorological conditions, fluctuating dramatically on intraday and seasonal scales, and exhibiting greater uncertainty compared to thermal power generation. This impacts the traditional deterministic load forecasting and dispatching operation mode of the power system. The construction process of wind power output models and PV power output models is explained in detail below: The process of constructing the wind power output model is as follows: The output of wind power can be indirectly obtained from wind speed. Extensive experimental data shows that the wind speed histogram exhibits a significant right-skew characteristic, meaning that moderate to low wind speeds are more common in daily life, while strong winds are less frequent. This satisfies the non-negativity and right-skewness characteristics required for the Weibull distribution. The specific formula for the Weibull marginal wind speed model is: (1) In the formula, The speed of the wind; For shape parameters, The larger the value, the sharper the wind speed distribution map becomes, and the wind speed changes are more concentrated around the average value; The range parameter determines the magnitude of the mean wind speed. The larger the value, the more the distribution will shift to the right.

[0019] Since the Weibull distribution involves independent sampling, the sampling results will amplify the smoothing effect between data, which will underestimate the volatility of wind power and lead to a large deviation between the sampling results and the actual situation. Therefore, this application adds a first-order autoregressive model, namely the AR(1) time-correlated structure, to the Weibull distribution. Its core formula is: (2) In the formula, This is the autocorrelation coefficient. The larger the value, the stronger the linear correlation between adjacent time points. In this design, it is set to 0.88. The standard normal perturbation term determines the magnitude of the external perturbation. In generating the wind speed sequence, this application first runs the AR(1) model for 100 steps to warm it up, retaining the parameter values ​​for the subsequent 24 hours to avoid errors caused by a cold start. Then, the standard normal cumulative distribution function is used to transform the first-order autoregressive model, generating a first-order autoregressive uniform distribution model, as shown in the following equation: (3) By performing inverse transformations on the Weibull marginal wind speed model in equation (1) and the first-order autoregressive uniform distribution model in equation (3), a wind speed sequence related to time speed is generated. Then, based on this wind speed sequence, the wind power output model can be further determined, as shown in the following equation: (4) In the formula, This refers to the output value of the wind turbine generator; This refers to the rated (maximum) output of the wind turbine generator set; , and These represent the cut-in, rated, and cut-out wind speeds, respectively. During the conversion, this application employs cubic spline interpolation to generate a smoother power curve, further simulating the power output in real-world scenarios. The model shows that wind turbines can only begin generating electricity when the wind speed exceeds a certain value, and that the turbines need to be disconnected when the wind speed is too strong to ensure overall turbine safety. Furthermore, based on the historical patterns of wind speed variations throughout the day in a specific region, variation weights are added to the wind speed at each moment, making the simulation results more closely resemble actual application scenarios and further improving the model's accuracy.

[0020] The process of constructing the photovoltaic power output model is as follows: The power output of photovoltaic systems can be directly obtained from the light intensity. The probability distribution model of light intensity is generally represented by the Beta distribution, and its formula is as follows: (5) in, and These represent the maximum and real-time light intensity during that time period, respectively. and These are the two shape parameters of the distribution.

[0021] By performing a first-order approximation on the above probability distribution model of light intensity, and neglecting temperature and system losses, we obtain the photovoltaic power output model, that is, the magnitude of photovoltaic active power output: (6) In the formula, This is the maximum light intensity during that period. It represents the maximum active power that the photovoltaic system can generate during that period, meaning that the stronger the sunlight, the greater the active power of the photovoltaic system.

[0022] The process of constructing the load demand model is as follows: To address the randomness of the daily load model, this design employs a normal distribution for random sampling and superimposes a first-order autoregressive model AR(1). Simultaneously, a typical daily bimodal load curve model is constructed to simulate actual fluctuations, with the following formula: (7) In the formula, t represents the corresponding time of day.

[0023] Peak load calibration was used to calibrate a typical daily bi-peak load curve model, resulting in the load demand model, as shown in the following equation: (8) In the formula, This represents the actual peak active load on a typical day.

[0024] The process of constructing the electric vehicle load model is as follows: The load power of electric vehicles is related to factors such as the number of trolleybuses in the local area and the daily mileage of trolleybuses. With the rapid development of electric vehicles, their characteristics of high power, concentrated time, and random space have brought a huge impact on the load side of the power system. Therefore, this application considers the electric vehicle load separately and incorporates it into the total load curve.

[0025] The daily driving distance of an electric vehicle follows a log-normal distribution, and its function is: (9) in, This represents the expected daily mileage. The variance of daily mileage traveled.

[0026] When an electric vehicle starts charging, this design assumes that the moment it starts charging is the time when the vehicle arrives at the charging station, and the time it leaves the charging station is the time when it stops charging. Both of these time points follow a normal distribution, and we can construct the first time model of the electric vehicle arriving at the charging station and the second time model of the electric vehicle leaving the charging station, as shown in equations (10) and (11): (10) In the formula, The expected value for the charging start time; This represents the variance of the charging start time.

[0027] (11) In the formula, This represents the expected time for the charging to end. This represents the variance of the charging end time. The entire charging process can be considered as being performed at a constant power.

[0028] Then, based on the first time model and the second time model, the electric vehicle load model is determined as follows: (12) In the formula, The total number of electric vehicles in the region. It refers to the charging power of a single vehicle. and These represent the distribution of charging start times. Distribution of charging stop times The cumulative distribution function of the two can be used to approximate the proportion of vehicles still in the charging state during that period by subtracting the two.

[0029] Step S102: Obtain typical scenarios including extreme working conditions; During the operation of new power systems, extreme weather events significantly impact source-load characteristics, potentially leading to severe inadequacy in system regulation capacity. This application analyzes historical source-load data under extreme events to simulate a sharp drop in renewable energy output during such events, ultimately yielding typical scenarios encompassing extreme operating conditions. The specific process is as follows: The Monte Carlo method, based on the law of large numbers and the central limit theorem, generates a sufficient number of scene samples through large-scale pseudo-random sampling to ensure that extreme states and tail features are fully covered. It can quickly complete tens of thousands of scene simulations and evaluations within an acceptable error range, providing strong support for engineering decision-making.

[0030] For normal weather scenarios, this application primarily focuses on extracting typical representative scenarios from the daily operation characteristics of the system. Specifically, the 24-hour net load curve corresponding to each normal weather scenario is used as the basic analysis object, and features are constructed from three dimensions: capacity pressure, ramp pressure, and curve shape. Capacity pressure is represented by the peak net load to reflect the system's capacity occupancy level; ramp pressure is represented by the maximum upward ramp rate and the maximum downward ramp rate to reflect the intensity of system regulation demand; the curve shape feature is achieved by removing the mean and normalizing the net load curve, followed by dimensionality reduction using principal component analysis, retaining principal component features with a cumulative explained variance of not less than 90%, to characterize the differences in intraday variation patterns across different scenarios. To highlight the importance of capacity and ramp features for system operational safety, this application standardizes various features and assigns different weights, thereby constructing a clustering feature matrix for cluster analysis.

[0031] For the extraction of typical scenarios, this application employs K-Medoids clustering to classify normal weather scenarios. K-Medoids can directly select actual samples as cluster centers, making it more suitable for extracting typical operating scenarios. The number of clusters is determined by silhouette coefficient comparison, and the center sample of each cluster is taken as the typical scenario of that cluster. Simultaneously, the probability of occurrence of a typical scenario is represented by the proportion of each category's sample count to the total number of normal weather scenarios, ultimately outputting the typical normal weather scenarios and their corresponding probabilities.

[0032] For extreme weather scenarios, this application focuses on identifying high-risk scenarios that exert significant pressure on system operation under conditions of heavy rain and drought. Considering the differences in the impact mechanisms of different types of extreme weather on the system, comprehensive pressure indicators are constructed for heavy rain and drought scenarios respectively. For heavy rain scenarios, system pressure is mainly characterized by weighted power shortage, peak gap, number of high-pressure duration periods, and climbing pressure, where climbing pressure is reflected by both the maximum upward and downward climbing rates. For drought scenarios, system operational pressure is mainly characterized by average daily load, peak load, evening peak load, and volatility. After standardization of each pressure indicator, they are weighted and summed according to preset weights to form comprehensive pressure scores for heavy rain and drought scenarios respectively. In the extreme scenario screening process, this paper uses the quantile method to identify high-risk scenarios, that is, scenarios with a comprehensive pressure score not lower than the 95th percentile are selected as the extreme scenario pool. This method can highlight extreme scenarios with truly significant system pressure characteristics while retaining sufficient sample representativeness. Based on this, the scenarios are further sorted from high to low according to the comprehensive stress score. The top 3 scenarios are selected from the extreme scenario pool as representative extreme scenarios, and the scenario with the highest stress score is defined as the most severe extreme scenario used for subsequent gap analysis and system weakness identification.

[0033] Step S103: Determine the volume of the pumped storage reservoir based on the characteristic model and the typical scenario; When conditions including heavy rain and drought are obtained, typical scenarios containing extreme working conditions can be obtained.

[0034] For cities, the precipitation amount during extreme weather events such as rainstorms can be indirectly calculated using the rainstorm intensity formula. The rainstorm intensity formula is an empirical function that quantitatively describes the relationship between rainstorm intensity, rainfall duration, and return period. It can provide guidance for urban planning and sponge city construction. Its expression is: (13) In the formula, Indicates the intensity of the rainstorm. The duration of the rainstorm The return period is set, while other rainfall parameters are related to local hydrological and atmospheric conditions. Based on the rainfall intensity formula, a constant-duration Chicago rainfall pattern is simulated to obtain the average precipitation and instantaneous precipitation at a certain moment during this rainstorm.

[0035] (14) (15) (16) This application statistically analyzes annual precipitation based on historical weather data, arranging daily precipitation into a sequence. It then defines scenarios for extreme weather: if the average precipitation for three consecutive days within a year exceeds 90% of the daily precipitation sequence, it is considered a rainstorm, typically characterized by thick cloud cover and heavy rainfall. In this case, the output of the corresponding characteristic model is adjusted, reducing wind power output to 40% of normal and photovoltaic output to 50%. If the average precipitation for seven consecutive days within a year is below 10% of the daily precipitation sequence, it is considered a drought, typically characterized by few clouds, light winds, and low humidity. In this case, wind power output is reduced to 80%, while photovoltaic output is increased by 10%. Taking Wuhan's weather data as an example, the probability of a rainstorm in a year is approximately 9.78%, and the probability of a drought is approximately 2.74%. By predicting changes in processing under extreme weather conditions, this application obtains scenarios that more closely reflect actual operational conditions.

[0036] After the above adjustments, the net load for different typical days can be obtained, and the calculation formula is as follows: (17) In the formula This represents the sum of power in 24 hours of each typical scenario. Based on this formula, the power gap under different typical scenarios can be obtained. The calculation results are weighted by probability and multiplied by 365 to obtain the expected annual energy gap.

[0037] Subsequently, by simulating energy storage devices, pumped storage capacity planning was carried out, converting the annual expected energy deficit into pumped storage reservoir volume, and a practical reservoir capacity formula was obtained: (18) In the formula, The effective head height is the pumping head difference. For overall efficiency, the energy loss throughout the pumping and storage process is included.

[0038] The output of the pumped storage unit is calculated using half-day energy conversion, assuming that the energy storage unit can only complete one cycle per day. This half-day energy conversion energy storage strategy corresponds to the load fluctuation pattern analyzed earlier: during the day, photovoltaic power generation increases supply while demand decreases, leading to downward peak shaving; at night, with concentrated charging of electric vehicles, photovoltaic power generation drops sharply, resulting in less supply and more demand, leading to upward peak shaving. Under this strategy, the output pattern of the energy storage unit is as follows: (19) In the formula, Indicates the total energy storage capacity. These are the loss parameters. By configuring the total energy storage capacity and maximum output of the energy storage unit, the improvement in the power system output gap after installing the pumped storage unit can be simulated, and the rationality of the results can be evaluated.

[0039] Simulation Experiment To test the effectiveness of the proposed method, detailed simulation experiments were conducted. For the source-load scenario, the rated power of wind power was set at 100MW, the rated power of photovoltaic power at 50MW, the size of the electric vehicle fleet was 120, and the average battery capacity per vehicle was 60kWh. Other settings were as described above. After completing the parameter configuration, Monte Carlo sampling was performed on the data, with the number of samplings set to 100,000 to balance computational efficiency with sampling accuracy and reduce errors caused by sample bias.

[0040] Considering that typical scenarios correspond to real-world samples rather than mean centers, K-Medoids clustering analysis is performed on the sampling results. First, the silhouette coefficients are calculated for different numbers of clusters, and the cluster number with the largest silhouette coefficient is taken as the optimal value. The results are as follows: Figure 2 As shown.

[0041] In each clustering result, the central sample of that cluster is selected as the typical scenario, and the corresponding net load curve, wind power output curve, photovoltaic power output curve, and total load curve for each typical scenario are output. Finally, the source-load uncertainty of the system under normal weather conditions is obtained as follows: Figure 3 As shown.

[0042] After setting extreme weather conditions, under both heavy rain and drought conditions, the new load uncertainty of the system is as follows: Figure 4 As shown. By comparison Figure 4 (a) and Figure 4 (b) As can be seen from the two figures, extreme weather has had a significant impact on the output of new energy sources, seriously affecting the stability of the system operation.

[0043] Five extreme scenarios were randomly selected, and the maximum energy and reactive power gap of the system under different scenarios were further analyzed. The results are as follows: Figure 5 , Figure 6 As shown: Comparative analysis of different typical scenarios in the figure shows that when extreme weather occurs, the maximum energy gap and reactive power gap of the system increase significantly, and the power deficit is obvious. This indicates that the system is likely to have a serious lack of regulation capacity and potential losses under this scenario, and emergency measures need to be prepared at any time.

[0044] In addition, this application also provides planning recommendations for pumped storage reservoirs. For the scenario mentioned above, if the head of the pumped storage reservoir is 200 meters and a 10% energy margin is reserved, the effective storage capacity is recommended to be 1.318 million m³.

[0045] Based on the suggested increase in reservoir capacity, an energy storage device was added to simulate the output of a pumped reservoir (which is less than the actual required energy storage capacity). The operation revealed a new deficit situation, as follows: Figure 7 , Figure 8 As shown. By Figure 7 and Figure 8 It can be seen that when the energy storage capacity is set at 1000MWh, the maximum energy gap of the system is significantly alleviated, eliminating the power deficit at most times. The remaining gaps shown in the figure can be addressed by expanding the scale of new energy power integration or further increasing the energy storage capacity. The above results demonstrate that this method has good feasibility and practical significance, and can provide theoretical support for subsequent pumped storage planning.

[0046] In summary, the method proposed in this application provides reasonable modeling and typical scenario generation for solar energy, wind power, conventional loads, and electric vehicle access at the source and load ends. At the same time, by adding different weights, it fully considers the impact of common extreme weather in the corresponding regions on the output of new energy sources. It can effectively analyze the power system regulation capacity gap under multiple uncertainties of source and load and the power limit overrun situation under the worst case, and provides reliable theoretical and data support for the construction of subsequent pumped storage power stations and other energy storage devices.

[0047] The above embodiments are preferred implementations of this application. In addition, this application can be implemented in other ways. Any obvious substitutions without departing from the concept of this technical solution are within the protection scope of this application.

[0048] To facilitate understanding by those skilled in the art of the improvements made by this application compared to the prior art, some of the accompanying drawings and descriptions have been simplified, and for clarity, some other elements have been omitted from this application. Those skilled in the art should realize that these omitted elements may also constitute the content of this application.

Claims

1. A pumped storage planning method considering the grid regulation capacity under extreme weather conditions, characterized in that, include: Establish characteristic models for various power sources and loads in the system; Acquire typical scenarios including extreme operating conditions; Based on the aforementioned characteristic model and the aforementioned typical scenario, the volume of the pumped storage reservoir is determined.

2. The method according to claim 1, characterized in that, The system includes characteristic models for various resources and loads, such as wind power output model, photovoltaic power output model, load demand model, and electric vehicle load model.

3. The method according to claim 2, characterized in that, The process of constructing the wind power output model includes the following steps: Obtain the Weibull marginal wind speed model and the first-order autoregressive model; The first-order autoregressive model is transformed using the standard normal cumulative distribution function to generate a first-order autoregressive uniform distribution model. The first-order autoregressive uniform distribution model and the Weibull marginal wind speed model are inversely transformed to generate a time-related wind speed sequence. Based on the wind speed sequence, a wind power output model is determined.

4. The method according to claim 2, characterized in that, The process of constructing the photovoltaic power output model includes the following steps: A probability distribution model of light intensity is constructed using the beta distribution; The photovoltaic power output model is determined by performing a first-order approximation on the light intensity probability distribution model.

5. The method according to claim 2, characterized in that, The process of constructing the load demand model includes the following steps: A typical daily bimodal load curve model was constructed by using a normal distribution and superimposing a first-order autoregressive model. Peak load calibration is used to calibrate the typical daily double-peak load curve model and determine the load demand model.

6. The method according to claim 2, characterized in that, The process of constructing the electric vehicle load model includes the following steps: Using a normal distribution, we construct a first-time model of electric vehicles arriving at a charging station and a second-time model of electric vehicles leaving a charging station. Based on the first time model and the second time model, the electric vehicle load model is determined.

7. The method according to claim 1, characterized in that, The process for determining typical scenarios under extreme operating conditions includes the following steps: Monte Carlo sampling is used to generate a sufficient number of scene samples through large-scale pseudo-random sampling. Principal component analysis is used to reduce the dimensionality of the scene samples while retaining the principal component features; After further standardizing the principal component features, different weights are assigned to them to construct a clustering feature matrix; K-Medoids clustering is used to extract typical scenarios from the clustering feature matrix, generating typical scenarios that include extreme working conditions.

8. The method according to claim 1, characterized in that, The determination of the pumped storage reservoir volume based on the aforementioned characteristic model and the aforementioned typical scenario includes: Using the aforementioned typical scenario, the output of the special effects model is adjusted to determine the net load for different typical days; Based on the net load of the different typical days, the annual expected energy deficit is determined; The volume of the pumped reservoir is determined based on the expected annual energy deficit.

Images