A data model driven wind-solar-storage system capacity planning method

Abstract

The application relates to a wind-solar-storage system capacity planning method based on a data model driver, which comprises the following steps: S1, observing historical data of photovoltaic and wind power to analyze the operation characteristics of a system and extracting key performance parameters of a unit; S2, predicting the performance parameters of the unit in a planning year by using a long short-term memory network (LSTM) according to the historical key performance parameters; S3, establishing an output ratio model through operation characteristics, obtaining the approximate output curve of wind and light in the planning year by using power generation output ratio, and correcting the output curve by using the unit performance parameters to obtain an optimized wind and light output curve; S4, firstly, a planning model considering economy and reliability is established; secondly, relevant basic data is input into the model, including load in a target level year and hourly characteristics of the unit; and finally, the optimal wind-solar-storage capacity planning result is obtained by solving through cplex.

Description

Technical Field

[0001] This invention relates to the field of system capacity planning for future wind and solar power generation, and specifically to a data model-driven method for wind-solar-storage system capacity planning. Background Technology

[0002] Electricity is an indispensable secondary energy source for human life and production in modern society. As of now, my country's total installed power capacity is 2.42 billion kilowatts, of which coal-fired power generation accounts for the vast majority. However, with the continuous exploitation of conventional energy resources leading to resource depletion and increasing environmental pressure, the use of renewable energy sources such as wind and solar power has become an inevitable trend.

[0003] However, due to the randomness, volatility and intermittency of renewable energy, the connection of large-scale renewable energy power generation such as wind and solar power to the power grid will have a certain impact on the operation of the power system, mainly in the following aspects: (1) impact on power quality; (2) impact on power system dispatch and operation; (3) impact on the power system's ability to accept renewable energy power generation; and (4) impact on the utilization rate of transmission and transformation equipment.

[0004] To mitigate the adverse effects of new energy sources, the current power grid employs energy storage technology. Energy storage devices can compensate for power shortages when the system is undersupply, and absorb excess power when the system is oversupplied. As energy storage technology becomes increasingly sophisticated, the speed of transition between these two states has greatly improved. Because energy storage devices can achieve state transitions and control output, they possess power regulation capabilities. Rationally planning the capacity ratio of new energy sources and energy storage units can fully leverage the advantages of both new energy power generation systems and energy storage devices. Therefore, optimizing the power grid's dispatching methods and formulating an optimal wind, solar, and energy storage capacity ratio plan that meets actual dispatching conditions plays a crucial role in ensuring the stable and safe operation of the power grid. Due to the rapid development of renewable energy in recent years driven by policy encouragement, the conventional wind and solar turbine models used for planning are complex and require a large amount of data, resulting in significant computational resources needed to update the planning results. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a data model-driven capacity planning method for wind-solar-storage systems. The method for obtaining future wind and solar power output curves is applicable to capacity planning under conditions of rapid development of wind-solar-storage systems. It uses the Kendall rank correlation coefficient to determine the correlation between wind and solar power output and installed capacity, and calculates key performance parameters of relevant wind and solar turbines according to formulas. Based on a trained LSTM, it obtains the turbine performance parameters for the target year. A model for the wind power output ratio and photovoltaic power output ratio is established to obtain approximate wind and solar power output curves. These curves are then corrected using predicted turbine performance parameters to obtain future output curves, which are then incorporated into the planning model to accelerate the capacity planning process.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A data model-driven capacity planning method for wind, solar, and energy storage systems includes the following steps:

[0008] Step S1: Observe the operating characteristics of the historical data analysis system for photovoltaic and wind power, and extract the key historical performance parameters of the units;

[0009] Step S2: Based on historical key performance parameters, use a long short-term memory network to predict the performance parameters of the unit in the planning year;

[0010] Step S3: Establish a power output ratio model through operating characteristics, obtain the wind and solar power output curve for the planned year using the power generation output ratio, and correct the power output curve by predicting the unit performance parameters to obtain the optimized wind and solar power output curve;

[0011] Step S4: Establish a planning model that considers economy and reliability; input relevant basic data into the planning model, including the load of the system in the target year and the hourly characteristics of the units; obtain the optimal wind, solar and storage capacity planning results through CPLEX solution.

[0012] Further, step S1 specifically includes:

[0013] Step S11: Determine the correlation between wind and solar power output and the correlation between wind and solar power output and installed capacity based on the Kendall rank correlation coefficient. The Kendall rank correlation coefficient is calculated as shown in formula (1):

[0014] (1)

[0015] in, This is a sample consisting of N sets of observations of the object vector (X, Y). Clearly... ,when When X and Y change in perfect agreement, it indicates that X and Y are positively correlated; when X and Y change in perfect agreement, it indicates that X and Y are positively correlated. When X and Y change in perfect unison, it indicates that X and Y are negatively correlated; when When the time interval is equal, it indicates that the changes in X and Y are half consistent and half inconsistent, and it is impossible to determine whether X and Y are related. The logarithm of wind and solar power output and installed capacity in the same order; Let n be the logarithm of the inverse order of wind and solar power output and installed capacity; n is the number of objects.

[0016] Step S12: Extract key performance parameters of wind and solar turbines from historical data based on the formulas for equivalent utilization hours and maximum output; the calculation of equivalent utilization hours, maximum output, and annual power generation is shown in formulas (2) and (3):

[0017] (2)

[0018] (3)

[0019] in, To contribute to the daily output of the generator set; This is the rated output of the unit; The equivalent use of decimals for the unit; This is the maximum output of the generator unit.

[0020] Further, step S2 specifically includes:

[0021] Step S21: Normalize the data input to the Long Short-Term Memory network and map it to the [0,1] interval. The normalization method is shown in formula (4).

[0022] (4)

[0023] in, and This represents the maximum and minimum values ​​of the i-th data set. This represents the data after normalization.

[0024] Step S22: Input the normalized historical data into the Long Short-Term Memory (LSTM) network for training, and use the trained LSTM network to obtain the unit performance parameters for the target year;

[0025] The computation process of the Long Short-Term Memory (LSTM) network is illustrated using time t as an example, showing the cell state at time t. Subject to the cell state of the previous time step and the Gate of Oblivion The impact of new information is shown in formula (5):

[0026] (5)

[0027] In the formula, , The calculation formulas are shown in formulas (6) and (7). This indicates information that needs to be updated;

[0028] (6)

[0029] (7)

[0030] For the softmax function, Implicit state information at time t-1 Forgotten Gate The weighting coefficients, Forgotten Gate The weighting coefficients, For input gate The weighting coefficients, For input gate The weighting coefficients, For the offset parameter of the forget gate, This is the offset parameter for the input gate.

[0031] After processing by the tanh function, it is then compared with the output gate. The information is multiplied point by point to obtain the final output. As shown in formula (8):

[0032] (8)

[0033] In the formula, The calculation formula is shown in formula (9).

[0034] (9)

[0035] For the output gate The weighting coefficients, For the output gate The weighting coefficients, This is the offset parameter for the output gate.

[0036] Furthermore, step S3 specifically includes:

[0037] Step S31: The wind power output and photovoltaic power output models are established as shown in formulas (10) and (11):

[0038] (10)

[0039] in, This refers to the sequential output ratio of the wind turbine generator set. For the first wind turbine Year Constant effort ; For the first wind turbine Year Installed capacity at any given time t represents the unit's operating time, and N represents the total number of years in the statistics.

[0040] (11)

[0041] in, The power output ratio of the photovoltaic unit; For the output of the photovoltaic unit at time t in year i, ; Let be the installed capacity of the photovoltaic unit at time t in year i. ;

[0042] Step S32: Extract the key performance parameters of the wind and solar turbine units from the output curve and compare them with the prediction results of step S22 to obtain the maximum output difference and equivalent utilization hours conversion factor. Correct the output at each time point to obtain the final unit output curve. The correction calculation is shown in formulas (12), (13), (14), and (15):

[0043] (12)

[0044] (13)

[0045] (14)

[0046] (15)

[0047] in, To predict the maximum output value, The output is the output after correction for the maximum output difference. To predict equivalent utilization hours, This is the equivalent utilization hours after correction for the maximum output difference. The conversion factor is used for equivalent utilization hours; This represents the maximum output difference.

[0048] Further, step S4 specifically includes:

[0049] Step S41: Establish a planning model. The planning model aims to minimize the economic efficiency of the power generation system throughout its entire life cycle. The constraints include: start-up and shutdown of thermal power units, ramp-up constraints, maximum output constraints of new energy sources, operation constraints of various types of energy storage, power balance constraints, and curtailment rate constraints.

[0050] The optimization objective of this planning model is shown in formula (16):

[0051] (16)

[0052] Among them, C wA C represents the investment cost of wind power in the target year. pvA C represents the investment cost of photovoltaic power in the target year. hyA C represents the investment cost of hydropower in the target year. bA C represents the investment cost of biomass energy in the target year. thA C represents the investment cost of thermal power in the target year;wOM C represents the operation and maintenance cost of wind power in the target year. pvOM C represents the operation and maintenance cost of photovoltaic power in the target year. hyOM C represents the operation and maintenance cost of hydropower in the target year. bOM C represents the operation and maintenance costs of biomass energy in the target year. thOM The operation and maintenance cost of thermal power plants for the target year;

[0053] At the system operation level, the constraints on power generation load balance and new energy curtailment rate are guaranteed, as shown in formulas (17) and (18). At the equipment operation level, the constraints on the output of new energy units, the charging and discharging constraints of energy storage equipment, and the state of charge constraints of remaining electricity are guaranteed, as shown in formulas (19), (20), and (21).

[0054] (17)

[0055] in, Let be the output of the thermal power plant at time t; Let t be the output power of the hydroelectric power source at time t; Let t be the output of the wind power at time t; Let be the output of the photovoltaic system at time t; Let be the output power of the stored energy at time t; Let be the load magnitude at time t; Let t be the output of biomass energy.

[0056] (18)

[0057] in, Let t be the theoretical maximum wind power output. The theoretical maximum output of the photovoltaic system at time t;

[0058] (19)

[0059] in, This represents the minimum output of wind power during time period t; This represents the maximum output of the wind power during time period t. This represents the minimum output of the photovoltaic system during time period t. This represents the maximum output of the photovoltaic system during time period t. This refers to all operating times of the unit.

[0060] (20)

[0061] Among them, P cha (t) represents the charging power of the energy storage device during time period t; P dis (t) represents the discharge power of the energy storage device during time period t, which is the decision variable; XE chaThis refers to the charging status of an energy storage device at a certain time period; XE dis This represents the discharge state of the energy storage device at a certain time period, and it is a 0-1 variable; The maximum charging power of the energy storage device; This represents the maximum discharge power of the energy storage device; the formula requires that the charging and discharging states of the energy storage device are mutually exclusive.

[0062] (twenty one)

[0063] Among them, P soc (t) represents the state of charge (SOC) of the remaining charge at time t; P soc (1) represents the initial state of charge (SOC) of the energy storage device; N represents the energy storage duration.

[0064] Step S42: Input relevant basic data into the model. The data includes the electricity load and hourly characteristics of the system in the target year, the power generation characteristics of the adjustable units, and the hourly power generation characteristics of the wind power and photovoltaic power generation bases. The optimal wind, solar and storage capacity planning results are obtained by solving the cplex problem.

[0065] Step S43: Establish an evaluation system to quantify the quality of the planning results. The evaluation index models are shown in formulas (22) and (23):

[0066] (1) Average annual power shortage rate of the power system:

[0067] (twenty two)

[0068] in, The annual average power shortage rate of the power system; The system power shortage at time t is in megawatts. Let t be the load operating power at time t, in megawatts.

[0069] (2) Average annual power shortage rate of the power system:

[0070] (twenty three)

[0071] in, The annual average curtailment rate of new energy sources; Wind power curtailment at time t, in megawatts; Photovoltaic power curtailment at time t, megawatts; Let t be the wind power generation at time t, in megawatts; Photovoltaic power generation at time t, in megawatts.

[0072] Compared with the prior art, the present invention has the following advantages:

[0073] 1. This invention discloses a capacity planning method for wind, solar and storage systems based on a data model. Currently, the mainstream method mainly involves predicting future weather conditions and inputting the obtained data into the principle feature model. This method requires the algorithm to estimate the real-time weather conditions of the target year. The large estimation dimension will affect the time required for each capacity planning. This model mainly estimates the key performance parameters of several units, which can reduce the time for training the machine learning network and thus improve the speed of capacity planning.

[0074] 2. This data model-driven method considers more comprehensive factors than conventional models when calculating wind and solar power output, resulting in more accurate output values. Conventional principle-based models mainly consider installed capacity and weather conditions. However, in practical applications, wind and solar technologies are constantly advancing, leading to improvements in equivalent utilization hours and output under the same weather conditions. Therefore, considering changes in wind and solar turbine performance parameters is crucial for predicting future output. This method modifies the approximate output curve after predicting future turbine performance parameters to obtain a more accurate output curve. Attached Figure Description

[0075] Figure 1 This is a flowchart of a data model-driven capacity planning method for wind, solar and energy storage systems in an embodiment of the present invention.

[0076] Figure 2 This is a structural diagram of the wind-solar-storage system in an embodiment of the present invention;

[0077] Figure 3 This is a structural diagram of the long short-term memory network in an embodiment of the present invention. Detailed Implementation

[0078] This invention provides a data model-driven capacity planning method for wind, solar and energy storage systems. In scenarios where renewable energy units bear the main power output of the system, the data model-driven method is used to obtain future wind and solar power output curves, simplifying conventional wind and solar models used for system planning and reducing the time required for system planning results.

[0079] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below through specific implementations and in conjunction with the accompanying drawings.

[0080] like Figure 1 As shown in the figure, an embodiment of the present invention provides a capacity planning method for wind-solar-storage systems based on a data model, comprising the following steps:

[0081] Step S1: Observe the operating characteristics of the historical data analysis system for photovoltaic and wind power, and extract the key performance parameters of the units;

[0082] Step S2: Predict the performance parameters of the unit in the planning year using a Long Short-Term Memory (LSTM) network based on historical key performance parameters;

[0083] Step S3: Establish a power output ratio model through operating characteristics, use the power generation output ratio to obtain the approximate power output curve of wind and solar power in the planned year, and predict the unit performance parameters to correct the power output curve to obtain the optimized wind and solar power output curve;

[0084] Step S4: First, establish a planning model that considers economy and reliability; second, input relevant basic data into the model, including the load of the system in the target year and the hourly characteristics of the units; finally, obtain the optimal wind, solar and storage capacity planning results through CPLEX.

[0085] In this embodiment, step S1 above, which involves observing the operational characteristics of the historical data analysis system for photovoltaic and wind power and extracting key performance parameters of the unit, specifically includes:

[0086] Step S11: Determine the correlation between wind and solar power output and the correlation between wind and solar power output and installed capacity based on the Kendall rank correlation coefficient. The Kendall rank correlation coefficient is calculated as shown in formula (1):

[0087] (1)

[0088] in, This is a sample consisting of N sets of observations of the object vector (X, Y). Clearly... ,when When X and Y change in perfect agreement, it indicates that X and Y are positively correlated; when X and Y change in perfect agreement, it indicates that X and Y are positively correlated. When X and Y change in perfect unison, it indicates that X and Y are negatively correlated; when When the time interval is equal, it indicates that the changes in X and Y are half consistent and half inconsistent, and it is impossible to determine whether X and Y are related. The logarithm of wind and solar power output and installed capacity in the same order; Let n be the logarithm of the inverse order of wind and solar power output and installed capacity; n is the number of objects.

[0089] Step S12: Extract key performance parameters of wind and solar turbines from historical data based on the formulas for equivalent utilization hours and maximum output. The calculations for equivalent utilization hours, maximum output, and annual power generation are shown in formulas (2) and (3):

[0090] (2)

[0091] (3)

[0092] in, To contribute to the daily output of the generator set; This is the rated output of the unit; The equivalent use of decimals for the unit; This is the maximum output of the generator unit.

[0093] After step S1 above, the correlation between wind and solar power output and installed capacity is determined, and the equivalent utilization hours and maximum output of wind and solar turbines are extracted from historical data.

[0094] In this embodiment, step S2 above, which involves predicting the performance parameters of the generating unit in the planning year using a Long Short-Term Memory (LSTM) network based on historical key performance parameters, specifically includes:

[0095] Step S21: Since the data input to the LSTM (capacity, technology maturity, wind speed / total solar radiation) are different physical quantities with large differences in dimensions, it is necessary to normalize these input data and map them to the [0,1] interval. The normalization method is shown in formula (4).

[0096] (4)

[0097] in, and This represents the maximum and minimum values ​​of the i-th data set. This represents the data after normalization.

[0098] Step S22: Input the normalized historical data into the LSTM for training. Use the trained LSTM to obtain the unit performance parameters for the target year. The structure diagram of the LSTM is shown below. Figure 2 As shown;

[0099] The LSTM computation process is illustrated using time t as an example, showing the cell state at time t. Subject to the cell state of the previous time step and the Gate of Oblivion The impact of new information is shown in formula (5):

[0100] (5)

[0101] In the formula, , The calculation formulas are shown in formulas (6) and (7). This indicates information that needs to be updated.

[0102] (6)

[0103] (7)

[0104] For the softmax function, Implicit state information at time t-1 Forgotten Gate The weighting coefficients, Forgotten Gate The weighting coefficients, For input gate The weighting coefficients, For input gate The weighting coefficients, For the offset parameter of the forget gate, This is the offset parameter for the input gate.

[0105] After processing by the tanh function, it is then compared with the output gate. The information is multiplied point by point to obtain the final output. As shown in formula (8):

[0106] (8)

[0107] In the formula, The calculation formula is shown in formula (9).

[0108] (9)

[0109] For the output gate The weighting coefficients, For the output gate The weighting coefficients, This is the offset parameter for the output gate.

[0110] After step S2 above, the relationship between capacity, technology maturity, wind speed / total solar radiation and unit performance parameters is established through a long short-term memory network.

[0111] In this embodiment, step S3 above: establishing a power output ratio model based on operating characteristics, obtaining the approximate wind and solar power output curve for the planned year using the power generation output ratio, and correcting the power output curve based on the predicted unit performance parameters to obtain an optimized wind and solar power output curve, specifically including:

[0112] Step S31: Climate is correlated over long time scales, so the power output of wind and solar turbines in recent years contains relevant weather information on an annual time scale. Simultaneously, considering the correlation between installed capacity and wind power output obtained from step S11, it is found that as the installed capacity increases, the output also increases. Based on this analysis, after removing extreme data, the average output ratio is used to describe the power output pattern of wind and solar turbines. The models for wind power output and photovoltaic output are shown in formulas (10) and (11):

[0113] (10)

[0114] in, This refers to the sequential output ratio of the wind turbine generator set. For the first wind turbine Year Constant effort ; For the first wind turbine Year Installed capacity at any given time t represents the unit's operating time, and N represents the total number of years in the statistics.

[0115] (11)

[0116] in, The power output ratio of the photovoltaic unit; For the output of the photovoltaic unit at time t in year i, ; Let be the installed capacity of the photovoltaic unit at time t in year i. .

[0117] Step S32: Substitute the power ratio formulas of wind turbines and photovoltaic units into historical data to calculate the approximate power output curves for the target year. Considering that wind and solar power output is not only related to installed capacity and weather, but also to the technological maturity of new energy units, the key performance parameters of wind and solar units are extracted from the obtained power output curves and compared with the prediction results of step S22 to obtain the maximum power output difference and equivalent utilization hours conversion coefficient. The power output at each time point is then corrected to obtain the final unit power output curve. The correction calculations are shown in formulas (12), (13), (14), and (15).

[0118] (12)

[0119] (13)

[0120] (14)

[0121] (15)

[0122] in, To predict the maximum output value, The output is the output after correction for the maximum output difference. To predict equivalent utilization hours, This is the equivalent utilization hours after correction for the maximum output difference. The conversion factor is used for equivalent utilization hours; This represents the maximum output difference.

[0123] After step S3 above, the approximate power output curve of wind and solar power is obtained based on historical data and the power ratio model. The key performance parameters of wind and solar units are extracted from the approximate power output curve and compared with the predicted values. The curve is then corrected to obtain the wind and solar power output curve for the target year.

[0124] In this embodiment, step S4 above involves: first, establishing a planning model that considers both economics and reliability; second, inputting relevant basic data into the model, including the system's target annual load and the hourly characteristics of the generating units; and finally, obtaining the optimal wind-solar-storage capacity planning result through CPLEX, specifically including:

[0125] Step S41: Establish a planning model. This optimization model aims to minimize the economic efficiency of the power generation system throughout its entire life cycle (including the operation and maintenance costs of thermal power, hydropower, wind power, photovoltaic, and biomass energy). The main constraints include: start-up and shutdown of thermal power units, ramp-up constraints, maximum output constraints of new energy sources (wind and solar), operation constraints of various types of energy storage, power balance constraints, and curtailment rate constraints.

[0126] The optimization objective of this planning model is shown in formula (16):

[0127] (16)

[0128] Among them, C wA C represents the investment cost of wind power in the target year. pvA C represents the investment cost of photovoltaic power in the target year. hyA C represents the investment cost of hydropower in the target year. bA C represents the investment cost of biomass energy in the target year. thA C represents the investment cost of thermal power in the target year; wOM C represents the operation and maintenance cost of wind power in the target year. pvOM C represents the operation and maintenance cost of photovoltaic power in the target year. hyOM C represents the operation and maintenance cost of hydropower in the target year. bOM C represents the operation and maintenance costs of biomass energy in the target year. thOM The operation and maintenance cost of thermal power plants for the target year.

[0129] Constraints are established from two aspects: system operation and equipment operation. At the system operation level, the constraints of power generation load balance and new energy curtailment rate need to be guaranteed as shown in formulas (17) and (18). At the equipment operation level, the constraints of new energy unit output, energy storage equipment charging and discharging, and remaining power charge state constraints need to be guaranteed as shown in formulas (19), (20), and (21).

[0130] (17)

[0131] in, Let be the output of the thermal power plant at time t; Let t be the output power of the hydroelectric power source at time t; Let t be the output of the wind power at time t; Let be the output of the photovoltaic system at time t; Let be the output power of the stored energy at time t; Let be the load magnitude at time t; Let t be the output of biomass energy.

[0132] (18)

[0133] in, Let t be the theoretical maximum wind power output. The theoretical maximum output of photovoltaic power at time t.

[0134] (19)

[0135] in, This represents the minimum output of wind power during time period t; This represents the maximum output of the wind power during time period t. This represents the minimum output of the photovoltaic system during time period t. This represents the maximum output of the photovoltaic system during time period t. This refers to all operating times of the unit.

[0136] (20)

[0137] Among them, P cha (t) represents the charging power of the energy storage device during time period t; P dis (t) represents the discharge power of the energy storage device during time period t, which is the decision variable; XE cha This refers to the charging status of an energy storage device at a certain time period; XE dis This represents the discharge state of the energy storage device at a certain time period, and it is a 0-1 variable; The maximum charging power of the energy storage device; The maximum discharge power of the energy storage device is given by the formula, which requires that the charging and discharging states of the energy storage device be mutually exclusive.

[0138] (twenty one)

[0139] Among them, P soc (t) represents the state of charge (SOC) of the remaining charge at time t; P soc (1) is the initial state of charge (SOC) of the energy storage device; N is the energy storage duration.

[0140] Step S42: Input relevant basic data into the model. This data includes the electricity load and hourly characteristics throughout the year for the system's target year, the power generation characteristics of adjustable units (thermal power, adjustable hydropower, etc.), and the hourly power generation characteristics of major wind and photovoltaic power generation bases throughout the year. The optimal wind-solar-storage capacity planning result is obtained through CPL (Continuous Processing Polymerization). The power system in the wind-solar-storage system includes photovoltaic, wind, thermal, biomass, hydropower, and pumped storage units. Power output reaches the load end through transmission, transformation, and distribution. To ensure a smoother output of new energy, an energy storage system is added. Figure 3 As shown.

[0141] Step S43: Only when the optimal capacity planning for wind, solar, and energy storage meets certain conditions is it practically meaningful. Therefore, it is necessary to establish an evaluation system to quantify the quality of the planning results. This step is of guiding significance for the planning of wind, solar, and energy storage. The models for each evaluation index are shown in formulas (22) and (23):

[0142] (1) Average annual power shortage rate of the power system:

[0143] (twenty two)

[0144] in, The annual average power shortage rate of the power system; The system power shortage at time t is in megawatts. Let t be the load operating power at time t, in megawatts.

[0145] (2) Average annual power shortage rate of the power system:

[0146] (twenty three)

[0147] in, The annual average curtailment rate of new energy sources; Wind power curtailment at time t, in megawatts; Photovoltaic power curtailment at time t, megawatts; Let t be the wind power generation at time t, in megawatts; Photovoltaic power generation at time t, in megawatts.

[0148] After step S4 above, the new energy output of the target year is input into the optimization model, and the capacity planning of the relevant units in the target year is obtained by solving the cplex problem. The evaluation system is then used to judge the quality of the planning results.

[0149] In summary, this invention discloses a data model-driven capacity planning method for wind, solar, and energy storage systems. To address the problem of long planning times due to complex unit models in current capacity planning, this invention proposes a method that establishes a wind-solar output ratio model by analyzing the relationship between wind and solar power output and unit capacity, outputting an approximate output curve. This output curve is then corrected by predicting key performance parameters of the wind and solar units before being input into the planning model to obtain the capacity configuration result. This invention also proposes a Kendall rank correlation coefficient method to determine the correlation between wind and solar power output and the correlation between wind and solar power output and installed capacity. Furthermore, to obtain information on key unit performance, a Long Short-Term Memory (LSTM) network is used to derive the relationship between unit capacity, technology maturity, wind speed / total solar radiation, and key unit performance parameters. In addition, an output ratio model is established based on the correlation between wind and solar power output and installed capacity to initially obtain the output curve. The optimized wind and solar power output curve, obtained by predicting future key performance parameters, is then input into the capacity planning model and solved using CPLEX to obtain the capacity planning result. The quality of the planning result is judged based on an evaluation system, and the capacity of conventional units is optimized based on the evaluation results.

Claims

1. A data model-driven capacity planning method for wind-solar-storage systems, characterized in that, Includes the following steps: Step S1: Observe the operating characteristics of the historical data analysis system for photovoltaic and wind power, and extract the key historical performance parameters of the units; Step S2: Based on historical key performance parameters, use a long short-term memory network to predict the unit's performance parameters for the planning year, including: Step S21: Normalize the data input to the Long Short-Term Memory network and map it to the [0,1] interval. The normalization method is shown in formula (4). (4) in, and This represents the maximum and minimum values ​​of the i-th data set. This represents the data after normalization. Step S22: Input the normalized historical data into the Long Short-Term Memory (LSTM) network for training, and use the trained LSTM network to obtain the unit performance parameters for the target year; The computation process of the Long Short-Term Memory (LSTM) network is illustrated using time t as an example, showing the cell state at time t. Subject to the cell state of the previous time step and the Gate of Oblivion The impact of new information is shown in formula (5): (5) In the formula, , The calculation formulas are shown in formulas (6) and (7). This indicates information that needs to be updated; (6) (7) For the softmax function, Implicit state information at time t-1 Forgotten Gate The weighting coefficients, Forgotten Gate The weighting coefficients, For input gate The weighting coefficients, For input gate The weighting coefficients, For the offset parameter of the forget gate, This is the offset parameter for the input gate; After processing by the tanh function, it is then compared with the output gate. The information is multiplied point by point to obtain the final output. As shown in formula (8): (8) In the formula, The calculation formula is shown in formula (9); (9) For the output gate The weighting coefficients, For the output gate The weighting coefficients, This is the offset parameter for the output gate; Step S3: Establish a power output ratio model based on operating characteristics, obtain the wind and solar power output curves for the planned year using the power generation output ratio, and correct the power output curves based on the predicted unit performance parameters to obtain optimized wind and solar power output curves, including: Step S31: The wind power output and photovoltaic power output models are established as shown in formulas (10) and (11): (10) in, This refers to the sequential output ratio of the wind turbine generator set. For the first wind turbine Year Constant effort ; For the first wind turbine Year Installed capacity at any given time ; t represents the unit's operating time, and N represents the total number of years in the statistics; (11) in, The power output ratio of the photovoltaic unit; For the output of the photovoltaic unit at time t in year i, ; Let be the installed capacity of the photovoltaic unit at time t in year i. ; Step S32: Extract the key performance parameters of the wind and solar turbine units from the output curve and compare them with the prediction results of step S22 to obtain the maximum output difference and equivalent utilization hours conversion factor. Correct the output at each time point to obtain the final unit output curve. The correction calculation is shown in formulas (12), (13), (14), and (15): (12) (13) (14) (15) in, To predict the maximum output value, The output is the output after correction for the maximum output difference. To predict equivalent utilization hours, This is the equivalent utilization hours after correction for the maximum output difference. The conversion factor is used for equivalent utilization hours; This is the maximum output difference; Step S4: Establish a planning model that considers economy and reliability; input relevant basic data into the planning model, including the load of the system in the target year and the hourly characteristics of the units; obtain the optimal wind, solar and storage capacity planning results through CPLEX solution.

2. The data model-driven capacity planning method for wind-solar-storage systems according to claim 1, characterized in that, Step S1 specifically includes: Step S11: Determine the correlation between wind and solar power output and the correlation between wind and solar power output and installed capacity based on the Kendall rank correlation coefficient. The Kendall rank correlation coefficient is calculated as shown in formula (1): (1) in, To observe a sample consisting of N sets of observations of the object vector (X, Y), ,when When X and Y change in perfect agreement, it indicates that X and Y are positively correlated; when X and Y change in perfect agreement, it indicates that X and Y are positively correlated. When X and Y change in perfect unison, it indicates that X and Y are negatively correlated; when When the time interval is equal, it indicates that the changes in X and Y are half consistent and half inconsistent, and it is impossible to determine whether X and Y are related. The logarithm of wind and solar power output and installed capacity in the same order; is the logarithm of the inverse order of wind and solar power output and installed capacity; n is the number of research objects; Step S12: Extract key performance parameters of wind and solar turbines from historical data based on the formulas for equivalent utilization hours and maximum output; the calculation of equivalent utilization hours, maximum output, and annual power generation is shown in formulas (2) and (3): (2) (3) in, To contribute to the daily output of the generator set; This is the rated output of the unit; The equivalent use of decimals for the unit; This is the maximum output of the generator unit.

3. The data model-driven capacity planning method for wind-solar-storage systems according to claim 1, characterized in that, Step S4 specifically includes: Step S41: Establish a planning model. The planning model aims to minimize the economic efficiency of the power generation system throughout its entire life cycle. The constraints include: start-up and shutdown of thermal power units, ramp-up constraints, maximum output constraints of new energy sources, operation constraints of various types of energy storage, power balance constraints, and curtailment rate constraints. The optimization objective of this planning model is shown in formula (16): (16) Among them, C wA C represents the investment cost of wind power in the target year. pvA C represents the investment cost of photovoltaic power in the target year. hyA C represents the investment cost of hydropower in the target year. bA C represents the investment cost of biomass energy in the target year. thA C represents the investment cost of thermal power in the target year; wOM C represents the operation and maintenance cost of wind power in the target year. pvOM C represents the operation and maintenance cost of photovoltaic power in the target year. hyOM C represents the operation and maintenance cost of hydropower in the target year. bOM C represents the operation and maintenance costs of biomass energy in the target year. thOM The operation and maintenance cost of thermal power plants for the target year; At the system operation level, the constraints on power generation load balance and new energy curtailment rate are guaranteed, as shown in formulas (17) and (18). At the equipment operation level, the constraints on the output of new energy units, the charging and discharging constraints of energy storage equipment, and the state of charge constraints of remaining electricity are guaranteed, as shown in formulas (19), (20), and (21). (17) in, Let be the output of the thermal power plant at time t; Let t be the output power of the hydroelectric power source at time t; Let t be the output of the wind power at time t; Let be the output of the photovoltaic system at time t; Let be the output power of the stored energy at time t; Let be the load magnitude at time t; Let t be the magnitude of biomass energy output at time t; (18) in, Let t be the theoretical maximum wind power output. The theoretical maximum output of the photovoltaic system at time t; (19) in, This represents the minimum output of wind power during time period t; This represents the maximum output of the wind power during time period t. This represents the minimum output of the photovoltaic system during time period t. This represents the maximum output of the photovoltaic system during time period t. This refers to all operating times of the unit; (20) Among them, P cha (t) represents the charging power of the energy storage device during time period t; P dis (t) represents the discharge power of the energy storage device during time period t, which is the decision variable; XE cha This refers to the charging status of an energy storage device at a certain time period; XE dis This represents the discharge state of the energy storage device at a certain time period, and it is a 0-1 variable; The maximum charging power of the energy storage device; This represents the maximum discharge power of the energy storage device; the formula requires that the charging and discharging states of the energy storage device are mutually exclusive. (21) Among them, P soc (t) represents the state of charge (SOC) of the remaining charge at time t; P soc (1) represents the initial state of charge (SOC) of the energy storage device; N represents the energy storage duration. Step S42: Input relevant basic data into the model. The data includes the electricity load and hourly characteristics of the system in the target year, the power generation characteristics of the adjustable units, and the hourly power generation characteristics of the wind power and photovoltaic power generation bases. The optimal wind, solar and storage capacity planning results are obtained by solving the cplex problem. Step S43: Establish an evaluation system to quantify the quality of the planning results. The evaluation index models are shown in formulas (22) and (23): (1) Average annual power shortage rate of the power system: (22) in, The annual average power shortage rate of the power system; The system power shortage at time t is in megawatts. Let be the load operating power at time t, in megawatts; (2) Average annual power shortage rate of the power system: (23) in, The annual average curtailment rate of new energy sources; Wind power curtailment at time t, in megawatts; Photovoltaic power curtailment at time t, megawatts; Let t be the wind power generation at time t, in megawatts; Photovoltaic power generation at time t, in megawatts.

Images