Climbing auxiliary service optimization scheduling method and system based on hybrid lstm model

Abstract

The application discloses a kind of based on the method and system for optimizing scheduling of climbing auxiliary service of mixed LSTM model, the method of the present application includes: constructing new energy power generation prediction model;Calculate the weight parameter and bias parameter of mixed LSTM model;Based on power generation prediction error iterative optimization weight parameter and bias parameter of mixed LSTM model and predict new energy power generation;The new energy power generation obtained by prediction is substituted into constraint condition, considering unit start-stop cost, fuel variable cost, frequency modulation opportunity cost, standby opportunity cost and up and down climbing cost, to establish the climbing clearing model with the minimum total cost of generation in the whole market as the goal, solve the scalar in electric energy market, the scalar in climbing market, the scalar in frequency modulation market and the scalar in standby market.The present application aims to improve the prediction accuracy of new energy power generation using mixed LSTM model, and then optimize the climbing demand prediction, while considering various cost factors, so that the optimization result is closer to actual operation.

Description

Technical Field

[0001] This invention relates to the field of power system dispatching technology, specifically to a method and system for optimizing the dispatching of hill-climbing auxiliary services based on a hybrid LSTM model. Background Technology

[0002] Traditional power system dispatching methods primarily focus on generator start-up costs and fuel costs, neglecting the impact of the accuracy of renewable energy generation forecasts on demand ramp-up forecasts. Furthermore, existing technologies rarely incorporate frequency regulation opportunity costs, reserve costs, and ramp-up costs, potentially leading to optimization results that deviate from actual operating conditions. Summary of the Invention

[0003] The technical problem this invention aims to solve is to provide a method and system for optimizing the scheduling of ramp-up auxiliary services based on a hybrid LSTM model, addressing the aforementioned problems in existing technologies. This invention aims to improve the accuracy of ramp-up demand forecasting by utilizing a hybrid LSTM model to optimize the forecasting of renewable energy generation. Simultaneously, the optimization model comprehensively considers the unit's start-up cost, fuel variation cost, ramp-up cost, secondary frequency regulation, and reserve opportunity cost, thereby minimizing the total cost while considering multiple cost factors. By fully utilizing the hybrid LSTM model to improve the accuracy of renewable energy generation forecasting, the invention optimizes ramp-up demand forecasting, and by considering multiple cost factors, makes the optimization results more closely reflect actual operation.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0005] A method for optimizing the scheduling of hill-climbing auxiliary services based on a hybrid LSTM model includes:

[0006] S01. With the goal of minimizing the prediction error of new energy power generation, a new energy power generation prediction model based on the weight parameters and bias parameters of the hybrid LSTM model is established. The hybrid LSTM model integrates two new energy power generation prediction models: the meteorological numerical model NWP and the LSTM model.

[0007] S02, calculate the weight parameters and bias parameters of the input gate, forget gate, output gate, and unit state of the LSTM model in the hybrid LSTM model respectively;

[0008] S03, the stochastic gradient descent algorithm is used to iteratively calculate the prediction error of new energy power generation, and the weight parameters and bias parameters of the hybrid LSTM model are updated to reduce the prediction error of new energy power generation in order to obtain the optimal hybrid LSTM model; the optimal hybrid LSTM model is used to predict the power generation of new energy.

[0009] S04. The predicted renewable energy generation is added as an input parameter to the constraints of the ramp-up clearing model. At the same time, the unit start-up and shutdown costs, fuel variation costs, frequency regulation opportunity costs, reserve opportunity costs, and ramp-up and ramp-down costs are considered. The ramp-up clearing model is established with the goal of minimizing the total power generation cost of the entire market. The winning bids in the power market, ramp-up market, frequency regulation market, and reserve market are solved for the ramp-up clearing model.

[0010] Optionally, the functional expression of the new energy power generation prediction model established in step S01 is:

[0011]

[0012] In the above formula, W represents the weight parameters of the input gate, forget gate, output gate, and cell state of the LSTM model in the hybrid LSTM model; b represents the bias parameters of the input gate, forget gate, output gate, and cell state of the LSTM model in the hybrid LSTM model; P t This represents the actual amount of new energy power generated at time t; The value represents the predicted renewable energy generation at time t. The hybrid LSTM model uses a weighted average of the renewable energy generation predicted by the meteorological numerical model (NWP) and the LSTM model at time t, based on preset weights, as the predicted renewable energy generation at time t.

[0013] Optionally, the meteorological data may include some or all of the following: temperature, relative humidity, wind speed, wind direction, air pressure, precipitation, and light intensity.

[0014] Optionally, the functional expression for iteratively calculating the power generation prediction error of new energy sources using the stochastic gradient descent algorithm is as follows:

[0015]

[0016]

[0017]

[0018]

[0019] In the above formula, W represents the weight parameters of the input gate, forget gate, output gate, and cell state of the LSTM model in the hybrid LSTM model; b represents the bias parameters of the input gate, forget gate, output gate, and cell state of the LSTM model in the hybrid LSTM model; P t This represents the actual amount of new energy power generated at time t; α represents the predicted renewable energy generation at time t; α represents the learning rate, used to control the step size of the update of both the weight matrix and the bias parameters; "←" indicates iterative calculation.

[0020] Optionally, the functional expression of the ramp-clearing model established in step S04 is:

[0021]

[0022] In the above formula, P gen R represents the generator set's power generation, S represents the generator set's standby capacity, and ΔP represents the generator set's start-up cost. reg The frequency regulation range of the generator set is represented by U, the uphill climbing capability of the generator set is represented by D, the downhill climbing capability of the generator set is represented by n, and the number of generator sets is S. i Let Δt represent the startup cost of the i-th generator set. i This represents the number of times the i-th generator set has been started. This represents the power generation of the i-th generator set. Let α represent the variable fuel cost of the i-th generator set. i This represents the frequency regulation cost of the i-th generator set. This represents the frequency regulation range of the i-th generator set. R represents the spot electrical energy revenue of the i-th generator set. i This represents the standby capacity of the i-th generator set. U represents the ramp-up cost of the i-th generator set. i and D i These represent the uphill and downhill climbing capabilities of the i-th generator set, respectively.

[0023] Optionally, the constraints of the ramp-up clearing model in step S04 include:

[0024]

[0025]

[0026]

[0027]

[0028]

[0029]

[0030] In the above formula, P load This is the load forecast value. This represents the predicted amount of new energy power generation. and Let these represent the maximum and minimum output values ​​of the i-th generator set, respectively; and These represent the maximum and minimum reserve capabilities of the i-th generator set, respectively. This represents the maximum frequency regulation capability of the i-th generator set; and These represent the maximum uphill climbing capacity and maximum downhill climbing capacity of the i-th generator set, respectively.

[0031] Optionally, in step S02, the functional expressions for calculating the input gate, forget gate, output gate, and weight parameters and bias parameters of the unit states in the hybrid LSTM model are as follows:

[0032] i t =σ(W i ·[h t-1 ,x t ]+b i ),

[0033] f t =σ(W f ·[h t-1 ,x t ]+b f ),

[0034] o t =σ(W o ·[h t-1 ,x t ]+b o ),

[0035]

[0036] Among them, i t f t o t and These represent the input gate, forget gate, output gate, and unit state of the LSTM model in the hybrid LSTM model, respectively, where σ is the activation function and W is the input gate, forget gate, output gate, and unit state. i W f W o and W C These are the weight parameters of the input gate, forget gate, output gate, and cell states in the hybrid LSTM model; b i b f b o and b C These represent the bias parameters for the input gate, forget gate, output gate, and cell state, respectively.

[0037] Furthermore, this invention also provides a hill-climbing auxiliary service optimization scheduling system based on a hybrid LSTM model, comprising:

[0038] The power generation prediction model modeling program unit is used to establish a new energy power generation prediction model based on the weight parameters and bias parameters of the hybrid LSTM model with the optimization objective of minimizing the prediction error of new energy power generation.

[0039] The model parameter calculation program unit is used to calculate the weight parameters and bias parameters of the input gate, forget gate, output gate, and unit states of the LSTM model in the hybrid LSTM model, respectively.

[0040] The model parameter optimization program unit is used to iteratively calculate the prediction error of new energy power generation using the stochastic gradient descent algorithm, and update the weight parameters and bias parameters of the hybrid LSTM model to reduce the prediction error of new energy power generation and obtain the optimal hybrid LSTM model; the optimal hybrid LSTM model is then used to predict the power generation of new energy.

[0041] The ramp-clearing model modeling and solving program unit is used to add the predicted new energy power generation as input parameters to the constraints of the ramp-clearing model. It also considers the unit start-up and shutdown costs, fuel variation costs, frequency regulation opportunity costs, reserve opportunity costs, and ramp-up and ramp-down costs. The ramp-clearing model is established with the goal of minimizing the total power generation cost of the entire market. The program solves for the winning bids in the power market, ramp-up market, frequency regulation market, and reserve market for the ramp-clearing model.

[0042] Furthermore, the present invention also provides a hill-climbing auxiliary service optimization scheduling system based on a hybrid LSTM model, comprising a microprocessor and a memory interconnected thereto, wherein the microprocessor is programmed or configured to execute the hill-climbing auxiliary service optimization scheduling method based on the hybrid LSTM model.

[0043] Furthermore, the present invention also provides a computer-readable storage medium storing a computer program, the computer program being programmed or configured by a microprocessor to execute the hill-climbing assisted service optimization scheduling method based on a hybrid LSTM model.

[0044] Compared with the prior art, the present invention has the following main advantages:

[0045] 1. The hybrid LSTM model used in this invention integrates two new energy power generation prediction models: the meteorological numerical model NWP and the LSTM model. Combining the LSTM model with the meteorological numerical model can yield more accurate new energy power generation.

[0046] 2. This invention combines new energy forecasting with optimization of ramp-up market clearing. A crucial aspect of ramp-up market clearing is forecasting ramp-up capacity demand, which consists of uncertain ramp-up demand plus certain ramp-up demand. The uncertain ramp-up demand is obtained by subtracting the new energy power generation forecast from the load forecast for the next period to obtain the net load forecast. Therefore, if the accuracy of new energy power generation forecasting can be optimized, the accuracy of ramp-up capacity demand forecasting can be improved, thereby enhancing the economic efficiency of the overall market clearing outcome.

[0047] 3. This invention includes incorporating the predicted renewable energy generation as an input parameter into the constraints of the ramp-up clearing model. It also considers unit start-up and shutdown costs, fuel variation costs, frequency regulation opportunity costs, reserve opportunity costs, and ramp-up and ramp-down costs. The ramp-up clearing model is established with the goal of minimizing the total power generation cost of the entire market. It comprehensively considers the joint optimization of four levels of markets, including the spot electricity market, when ramp-up ancillary service market, frequency regulation ancillary service market, and reserve ancillary service market exist simultaneously in the spot market environment. It can effectively achieve electricity market clearing with the goal of minimizing the overall cost by adding the opportunity costs of frequency regulation and reserve to the optimization objective. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the basic process of the method in an embodiment of the present invention. Detailed Implementation

[0049] like Figure 1 As shown, the hill-climbing auxiliary service optimization scheduling method based on the hybrid LSTM model in this embodiment includes:

[0050] S01. With the goal of minimizing the prediction error of new energy power generation, a new energy power generation prediction model based on the weight parameters and bias parameters of the hybrid LSTM model is established. The hybrid LSTM model integrates two new energy power generation prediction models: the meteorological numerical model NWP and the LSTM model.

[0051] S02, calculate the weight parameters and bias parameters of the input gate, forget gate, output gate, and unit state of the LSTM model in the hybrid LSTM model respectively;

[0052] S03, the stochastic gradient descent algorithm is used to iteratively calculate the prediction error of new energy power generation, and the weight parameters and bias parameters of the hybrid LSTM model are updated to reduce the prediction error of new energy power generation in order to obtain the optimal hybrid LSTM model; the optimal hybrid LSTM model is used to predict the power generation of new energy.

[0053] S04. The predicted renewable energy generation is added as an input parameter to the constraints of the ramp-up clearing model. At the same time, the unit start-up and shutdown costs, fuel variation costs, frequency regulation opportunity costs, reserve opportunity costs, and ramp-up and ramp-down costs are considered. The ramp-up clearing model is established with the goal of minimizing the total power generation cost of the entire market. The winning bids in the power market, ramp-up market, frequency regulation market, and reserve market are solved for the ramp-up clearing model.

[0054] In this embodiment, the functional expression of the new energy power generation prediction model established in step S01 is:

[0055]

[0056] In the above formula, W represents the weight parameters of the input gate, forget gate, output gate, and cell state of the LSTM model in the hybrid LSTM model; b represents the bias parameters of the input gate, forget gate, output gate, and cell state of the LSTM model in the hybrid LSTM model; P t This represents the actual amount of new energy power generated at time t; The value represents the predicted renewable energy generation at time t. The hybrid LSTM model uses a weighted average of the renewable energy generation predicted by the meteorological numerical model (NWP) and the LSTM model at time t, based on preset weights, as the predicted renewable energy generation at time t. In this embodiment, meteorological data includes some or all of the following: temperature, relative humidity, wind speed, wind direction, air pressure, precipitation, and light intensity. The meteorological numerical model (NWP) is a well-known numerical model that simulates meteorological parameters over a future period based on atmospheric dynamics and thermodynamics principles, and then combines this with a physical model of wind or solar power generation to predict power generation. It is widely used in new energy power plants. For example, the meteorological numerical model (NWP) used in this embodiment is a short-term load forecasting model based on Attention-BiLSTM neural network and meteorological data correction (Wang Jidong, Du Chong. Short-term load forecasting model based on Attention-BiLSTM neural network and meteorological data correction [J]. Electric Power Automation Equipment, 2022, 42(4):7.). It should be noted that this embodiment only involves the application of the meteorological numerical model (NWP) and does not depend on the specific implementation method of the meteorological numerical model (NWP), therefore, it will not be described in detail here.

[0057] In this embodiment, the function expression for iteratively calculating the power generation prediction error of new energy sources using the stochastic gradient descent algorithm is as follows:

[0058]

[0059]

[0060]

[0061]

[0062] In the above formula, W represents the weight parameters of the input gate, forget gate, output gate, and cell state of the LSTM model in the hybrid LSTM model; b represents the bias parameters of the input gate, forget gate, output gate, and cell state of the LSTM model in the hybrid LSTM model; P t This represents the actual amount of new energy power generated at time t; denoted by , W represents the predicted renewable energy generation at time t; α represents the learning rate, used to control the step size of parameter updates for both the weight matrix and the bias parameters; "←" indicates iterative calculation. Through iterative optimization, a set of optimized model parameters W and bias parameters b can be obtained, thereby improving the accuracy of renewable energy generation prediction.

[0063] In this embodiment, the functional expression of the ramp-up and clearing model established in step S04 is:

[0064]

[0065] In the above formula, P gen R represents the generator set's power generation, S represents the generator set's standby capacity, and ΔP represents the generator set's start-up cost. reg The frequency regulation range of the generator set is represented by U, the uphill climbing capability of the generator set is represented by D, the downhill climbing capability of the generator set is represented by n, and the number of generator sets is S. i Let Δt represent the startup cost of the i-th generator set. i This represents the number of times the i-th generator set has been started. This represents the power generation (winning bid amount in the electricity market) of the i-th generator unit. Let α represent the variable fuel cost of the i-th generator set. i This represents the frequency regulation cost of the i-th generator set. This represents the frequency regulation range of the i-th generator set (corresponding to the winning bid in the frequency regulation market). R represents the spot electrical energy revenue of the i-th generator set. i This represents the reserve capacity of the i-th generator unit (corresponding to the amount won in the reserve market). U represents the ramp-up cost of the i-th generator set. i and D i These represent the uphill and downhill climbing capabilities of the i-th generator set (corresponding to the winning bid in the uphill market).

[0066] In this embodiment, the constraints of the ramp-up clearing model in step S04 include:

[0067]

[0068]

[0069]

[0070]

[0071]

[0072]

[0073] In the above formula, P load This is the load forecast value. This represents the predicted amount of new energy power generation. and Let these represent the maximum and minimum output values ​​of the i-th generator set, respectively; and These represent the maximum and minimum reserve capabilities of the i-th generator set, respectively. This represents the maximum frequency regulation capability of the i-th generator set; and These represent the maximum uphill climbing capacity and maximum downhill climbing capacity of the i-th generator set, respectively.

[0074] In this embodiment, the function expressions for calculating the weight parameters and bias parameters of the input gate, forget gate, output gate, and unit states of the LSTM model in step S02 are as follows:

[0075] i t =σ(W i ·[h t-1 ,x t ]+b i ),

[0076] f t =σ(W f ·[h t-1 ,x t ]+b f ),

[0077] o t =σ(W o ·[h t-1 ,x t ]+b o ),

[0078]

[0079] Among them, i t f t o t and These represent the input gate, forget gate, output gate, and unit state of the LSTM model in the hybrid LSTM model, respectively, where σ is the activation function and W is the input gate, forget gate, output gate, and unit state. i W f W o and W c These are the input gate, forget gate, output gate, and weight parameters of the unit state in the hybrid LSTM model, used to store the hidden state h from the previous time step. t-1 and the input x at the current time step t Convert to the updated values ​​for each gate and cell state; b i b f b o and b C These represent the bias parameters for the input gate, forget gate, output gate, and cell state, respectively, which are used to introduce additional nonlinear offsets for updating the state of each gate and cell.

[0080] In this embodiment, step S04, which solves for the winning bids in the power market, ramp market, frequency regulation market, and reserve market for the ramp-clearing model, uses the well-known hybrid certificate current programming algorithm with the CPLEX commercial solver. Alternatively, other well-known solution methods can be used as needed.

[0081] In summary, the hill-climbing auxiliary service optimization scheduling method based on the hybrid LSTM model in this embodiment integrates two new energy power generation prediction models: the meteorological numerical model NWP and the LSTM model. Combining the LSTM model with the meteorological numerical model yields more accurate new energy power generation. This hill-climbing auxiliary service optimization scheduling method based on the hybrid LSTM model combines new energy forecasting with hill-climbing market clearing optimization. A crucial aspect of hill-climbing market clearing is predicting hill-climbing capacity demand, which consists of uncertain hill-climbing demand plus deterministic hill-climbing demand. The uncertain hill-climbing demand is obtained by subtracting the new energy power generation forecast from the load forecast for the next period to obtain the net load forecast. Therefore, optimizing the accuracy of new energy power generation forecasting improves the accuracy of hill-climbing capacity demand forecasting, thereby enhancing the economic efficiency of the overall market clearing outcome. This embodiment of the hill-climbing ancillary service optimization scheduling method based on a hybrid LSTM model includes adding the predicted renewable energy generation as an input parameter to the constraints of the hill-climbing clearing model. It also considers unit start-up and shutdown costs, fuel variation costs, frequency regulation opportunity costs, reserve opportunity costs, and hill-climbing costs. The hill-climbing clearing model is established with the goal of minimizing the total power generation cost of the entire market. It comprehensively considers the joint optimization of four levels of markets, including the spot energy market, when hill-climbing ancillary service market, frequency regulation ancillary service market, and reserve ancillary service market exist simultaneously in the spot market environment. It can effectively achieve energy market clearing with the goal of minimizing the overall cost by adding the opportunity costs of frequency regulation and reserve to the optimization objective. This embodiment of the hill-climbing auxiliary service optimization scheduling method based on the hybrid LSTM model optimizes the prediction of renewable energy power generation by using the hybrid LSTM model, thereby improving the accuracy of hill-climbing demand prediction. At the same time, the optimization model comprehensively considers the unit start-up cost, fuel variation cost, hill-climbing cost, secondary frequency regulation, and reserve opportunity cost, so as to minimize the total cost based on considering multiple cost factors. By making full use of the hybrid LSTM model to improve the accuracy of renewable energy power generation prediction, the hill-climbing demand prediction is optimized. The consideration of multiple cost factors makes the optimization results closer to actual operation.

[0082] Furthermore, this embodiment also provides a hill-climbing auxiliary service optimization scheduling system based on a hybrid LSTM model, including:

[0083] The power generation prediction model modeling program unit is used to establish a new energy power generation prediction model based on the weight parameters and bias parameters of the hybrid LSTM model with the optimization objective of minimizing the prediction error of new energy power generation.

[0084] The model parameter calculation program unit is used to calculate the weight parameters and bias parameters of the input gate, forget gate, output gate, and unit states of the LSTM model in the hybrid LSTM model, respectively.

[0085] The model parameter optimization program unit is used to iteratively calculate the prediction error of new energy power generation using the stochastic gradient descent algorithm, and update the weight parameters and bias parameters of the hybrid LSTM model to reduce the prediction error of new energy power generation and obtain the optimal hybrid LSTM model; the optimal hybrid LSTM model is then used to predict the power generation of new energy.

[0086] The ramp-clearing model modeling and solving program unit is used to add the predicted new energy power generation as input parameters to the constraints of the ramp-clearing model. It also considers the unit start-up and shutdown costs, fuel variation costs, frequency regulation opportunity costs, reserve opportunity costs, and ramp-up and ramp-down costs. The ramp-clearing model is established with the goal of minimizing the total power generation cost of the entire market. The program solves for the winning bids in the power market, ramp-up market, frequency regulation market, and reserve market for the ramp-clearing model.

[0087] Furthermore, this embodiment also provides a hill-climbing auxiliary service optimization scheduling system based on a hybrid LSTM model, including a microprocessor and a memory interconnected, wherein the microprocessor is programmed or configured to execute the hill-climbing auxiliary service optimization scheduling method based on the hybrid LSTM model.

[0088] Furthermore, this embodiment also provides a computer-readable storage medium storing a computer program, which is used to be programmed or configured by a microprocessor to execute the hill-climbing assisted service optimization scheduling method based on a hybrid LSTM model.

[0089] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-readable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0090] The above description is merely a preferred embodiment of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for optimizing the scheduling of hill-climbing auxiliary services based on a hybrid LSTM model, characterized in that, include: S01. With the goal of minimizing the prediction error of new energy power generation, a new energy power generation prediction model based on the weight parameters and bias parameters of the hybrid LSTM model is established. The hybrid LSTM model integrates two new energy power generation prediction models: the meteorological numerical model NWP and the LSTM model. S02, calculate the weight parameters and bias parameters of the input gate, forget gate, output gate, and unit state of the LSTM model in the hybrid LSTM model respectively; S03, the stochastic gradient descent algorithm is used to iteratively calculate the prediction error of new energy power generation, and the weight parameters and bias parameters of the hybrid LSTM model are updated to reduce the prediction error of new energy power generation in order to obtain the optimal hybrid LSTM model; the optimal hybrid LSTM model is used to predict the power generation of new energy. S04. The predicted new energy power generation is added as an input parameter to the constraints of the ramp-up clearing model. At the same time, the unit start-up and shutdown costs, fuel variation costs, frequency regulation opportunity costs, reserve opportunity costs, and ramp-up and ramp-down costs are considered. The ramp-up clearing model is established with the goal of minimizing the total power generation cost of the entire market. The winning bids in the power market, ramp-up market, frequency regulation market, and reserve market are solved for the ramp-up clearing model. The functional expression of the new energy power generation prediction model established in step S01 is: ， In the above formula, The weight parameters represent the input gate, forget gate, output gate, and cell state weights of the LSTM model in the hybrid LSTM model. This represents the bias parameters of the input gate, forget gate, output gate, and cell states in the hybrid LSTM model. Indicates the length of the time series; This represents the actual amount of new energy power generated at time t; The value represents the predicted renewable energy generation at time t. The hybrid LSTM model uses a weighted average of the renewable energy generation predicted by the meteorological numerical model (NWP) and the LSTM model at time t, based on preset weights, as the predicted renewable energy generation at time t. ; The functional expression for iteratively calculating the power generation prediction error of new energy sources using the stochastic gradient descent algorithm is as follows: ， ， ， ， In the above formula, The weight parameters represent the input gate, forget gate, output gate, and cell state weights of the LSTM model in the hybrid LSTM model. This represents the bias parameters of the input gate, forget gate, output gate, and cell states in the hybrid LSTM model. Indicates the length of the time series; This represents the actual amount of new energy power generated at time t; This represents the predicted amount of new energy power generation at time t; This represents the learning rate, used to control the step size for updating both the weight matrix and the bias parameters; " indicates iterative calculation.

2. The hill-climbing auxiliary service optimization scheduling method based on a hybrid LSTM model according to claim 1, characterized in that, The meteorological data includes some or all of the following: temperature, relative humidity, wind speed, wind direction, air pressure, precipitation, and light intensity.

3. The hill-climbing auxiliary service optimization scheduling method based on a hybrid LSTM model according to claim 1, characterized in that, The functional expression of the ramp-up and clearing model established in step S04 is: ， In the above formula, This indicates the power generation of the generator set. Indicates the standby capacity of the generator set. This indicates the starting cost of the generator set. Indicates the frequency regulation range of the generator set. This indicates the generator set's climbing ability. This indicates the generator set's downhill climbing capability. For the number of generator sets, Indicates the first i The startup cost of a generator set Indicates the first i Number of times each generator set is started. Indicates the first i The power generation of each generator set Indicates the first i Variable fuel costs per generator set Indicates the first i Frequency regulation cost of a single generator set Indicates the first i Frequency regulation range of each generator set Indicates the first i Spot electricity revenue of a generator set Indicates the first i The backup capacity of each generator set Indicates the first i The cost of ramping uphill for a single generator set and They represent the first i The uphill and downhill climbing capabilities of each generator set.

4. The hill-climbing auxiliary service optimization scheduling method based on a hybrid LSTM model according to claim 3, characterized in that, The constraints of the ramp-up clearing model in step S04 include: ， ， ， ， ， ， In the above formula, This is the load forecast value. This represents the predicted amount of new energy power generation. and They represent the first i The maximum and minimum output of each generator set; and They represent the first i Maximum and minimum standby capacity of each generator set; Indicates the first i Maximum frequency regulation capability of each generator set; and They represent the first i The maximum uphill and downhill climbing capabilities of each generator set.

5. The hill-climbing auxiliary service optimization scheduling method based on a hybrid LSTM model according to claim 1, characterized in that, In step S02, the functional expressions for calculating the input gate, forget gate, output gate, and weight parameters and bias parameters of the unit states in the hybrid LSTM model are as follows: ， ， ， ， in, , , and These represent the input gate, forget gate, output gate, and cell state of the LSTM model in the hybrid LSTM model, respectively. For activation function, , , and These are the weight parameters of the input gate, forget gate, output gate, and cell state in the hybrid LSTM model. , , and These represent the bias parameters for the input gate, forget gate, output gate, and cell state, respectively.

6. A hill-climbing auxiliary service optimization scheduling system for applying the hill-climbing auxiliary service optimization scheduling method based on a hybrid LSTM model as described in any one of claims 1 to 5, characterized in that, include: The power generation prediction model modeling program unit is used to establish a new energy power generation prediction model based on the weight parameters and bias parameters of the hybrid LSTM model with the optimization objective of minimizing the prediction error of new energy power generation. The model parameter calculation program unit is used to calculate the weight parameters and bias parameters of the input gate, forget gate, output gate, and unit states of the LSTM model in the hybrid LSTM model, respectively. The model parameter optimization program unit is used to iteratively calculate the prediction error of new energy power generation using the stochastic gradient descent algorithm, and update the weight parameters and bias parameters of the hybrid LSTM model to reduce the prediction error of new energy power generation and obtain the optimal hybrid LSTM model; the optimal hybrid LSTM model is then used to predict the power generation of new energy. The ramp-clearing model modeling and solving program unit is used to add the predicted new energy power generation as input parameters to the constraints of the ramp-clearing model. It also considers the unit start-up and shutdown costs, fuel variation costs, frequency regulation opportunity costs, reserve opportunity costs, and ramp-up and ramp-down costs. The ramp-clearing model is established with the goal of minimizing the total power generation cost of the entire market. The program solves for the winning bids in the power market, ramp-up market, frequency regulation market, and reserve market for the ramp-clearing model.

7. A hill-climbing assisted service optimization scheduling system based on a hybrid LSTM model, comprising interconnected microprocessors and memory, characterized in that, The microprocessor is programmed or configured to execute the hill-climbing auxiliary service optimization scheduling method based on a hybrid LSTM model as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, The computer program is used to be programmed or configured by a microprocessor to execute the hill-climbing assisted service optimization scheduling method based on a hybrid LSTM model as described in any one of claims 1 to 5.

Images